Research ArticleSelf-Organized Cell Outage Detection Architecture andApproach for 5G H-CRAN
Peng Yu Fanqin Zhou Tao Zhang Wenjing Li Lei Feng and Xuesong Qiu
State Key Laboratory of Networking and Switching Technology Beijing University of Posts and TelecommunicationsBeijing 100876 China
Correspondence should be addressed to Wenjing Li wjlibupteducn
Received 1 December 2017 Accepted 3 April 2018 Published 10 May 2018
Academic Editor Gui-Ling Sun
Copyright copy 2018 Peng Yu et al This is an open access article distributed under the Creative Commons Attribution License whichpermits unrestricted use distribution and reproduction in any medium provided the original work is properly cited
An attractive architecture called heterogeneous cloud radio access networks (H-CRAN) becomes one of the important componentsof 5G networks which can provide ubiquitous high-bandwidth services with flexible network construction However massiveaccess nodes increase the risk of cell outages leading to negative impact on user-perceived QoS (Quality of Service) and QoE(Quality of Experience)Thus cell outagemanagement (COM) became a key function proposed in SON (Self-OrganizedNetworks)use cases Based on COM cell outage detection (COD) will be resolved before cell outage compensation (COC) Currentlyfew studies concentrate on COD for 5G H-CRAN and we propose self-organized COD architecture and approach for it Wefirstly summarize current COD solutions for LTELTE-A HetNets and then introduce self-organized architecture and approachsuitable for H-CRAN which includes COD architecture and procedures and corresponding key technologies for it Based on thearchitecture we take a use case with handover data analysis usingmodified LOF (Local Outlier Factor) detection approach to detectoutage for different kinds of cells in H-CRAN Results show that the proposed approach can identify the outage cell effectively
1 Introduction
The future 5G networks will provide ubiquitous high-bandwidth network services for varieties types of hetero-geneous terminals by integrating different advanced tech-nologies such as millimeter wave large-scale Multiple-Input Multiple-Output (MIMO) and Nonorthogonal Multi-ple Access (NOMA) for wireless access and SoftwareDefinedNetwork (SDN) cloud computing for core networks [1] Inorder to make network more efficient 5G will be a combina-tion of centralized performance processing and distributeddemodulation C-RAN has become a new type of networkarchitecture because of its advantages of cloud computing anddistributed processing which can realize the efficient usageand sharing of computing resources [2] However LTELTE-A heterogeneous networks (HetNet) providing differentiateddata services will still be left to 5G networks To improvespectrum efficiency and energy efficiency for HetNet coex-isting with C-RAN a new 5G wireless network architecturecall H-CRAN is proposed [3] In addition a brief architecture
for H-CARN is shown in Figure 1 Each nodeC connects toACEs or RRHs from wired or wireless fronthaul and userscommunicate with ACE or RRH fromwireless link and D2Dlink is also adopted to improve edge user throughputs as wellNodeCs are located in cloud environment and can connect toInternet
In 5G H-CRAN Building Baseband Unit (BBU) poolsare centralized into cloud computing node called NodeC [3]These nodes can communicate with each other and controlregional access nodes such as remote radio heads (RRHs)and existing legacy communication entities (LCEs) All theNodeC are located in the cloud and connect to the Internetand they can control ACEs and RRHs with wired or wirelessfronthaulMoreover NodeC collects and processes the spatialinformation from ACEs and RRHs and performs most ofthe processing calculating and radio resource managementfunctions Here ACEs (such as eNodeB) may pay the role ofproviding seamless coverage and RRHs are response for hotspots Users in H-CRAN are composed of different kinds ofdevices such as mobile phones laptops smart meters and
HindawiWireless Communications and Mobile ComputingVolume 2018 Article ID 6201386 11 pageshttpsdoiorg10115520186201386
2 Wireless Communications and Mobile Computing
Internet
NodeC NodeC NodeC
ACE
RRH
RRH
RRH
RRH
RRH
RRH
RRH Fronthaul
D2D
Wireless communication linkACE Fronthaul
middot middot middot middot middot middot
Figure 1 A brief architecture of H-CRAN
Internet of things (IoT) devices and may adopt advancedcommunication technologies such as D2D [4]
To realize the universal plug and play function offloadnetwork traffic immediately and manage the computing andspectrum resource dynamically SON plays an importantrole to realize intelligent management of H-CRAN SON isproposed to reduce service providersrsquo operating expenses inLTELTE-A systems and HetNet [5] H-CRAN consists ofa large number of heterogeneous access nodes and cloudcomputing resource units and its resource should be virtu-alized for sharing as well Therefore a large-scale SON (LS-SON) that integrates a unified autonomic process of ultra-computing ultra-planning ultra-configuration and ultra-optimization is preferred [3]
LS-SON can reduce the complexity of cochannel interfer-ence management in H-CRAN to save the operating costs
of all RRHs and ACEs So it is used to coordinate themanagement functions of the entire network and improve theoverall operational efficiency Because NodeC needs to servemultiple RANs and cooperate with RRHs it is consideredto implement self-configuring self-optimization and self-healing by using a centralized SON architecture for LS-SON
As one of the critical functions and use cases of SONself-healing not only identifies fault events but also is able todiagnose the causes (for example deciding why it happened)and then triggers appropriate compensation mechanism toturn network to normal state [6] In H-CRAN it first has toperceive a fault that has occurred or is about to occur andthen adopts proper actions to recover the services (partiallyor wholly definitely or temporarily)
According to 3GPP standard the self-healing functioncan be further divided into multiple use cases including fault
Wireless Communications and Mobile Computing 3
diagnosis fault classification and COM COM first needs todetect the cell outage automatically and then perform a rea-sonable compensation mechanism to repair the faults so asto minimize the impact of cell outage [7] Therefore COD isan essential prerequisite for self-healing
At present the studies of COD mainly aim at tradi-tional LTELTE-AHetNet with limited data collection whichmay not be appropriate for the complex H-CRAN networkarchitecture This paper proposes a self-organized CODarchitecture and a corresponding approach for H-CRANbased on previous work and gives a use case with HO dataanalysis to evaluate them
The rest of the paper is organized as follows Section 2introduces the related work for COD Section 3 proposes thearchitecture and approach for 5GH-CRANCODmeanwhileintroducing COD procedures and related technologies InSection 4 a CODuse case forHOdata analysis withmodifiedM-LOF is introduced and conclusions are given in Section 5
2 Related Work
In wireless communication networks cell outage is mainlycaused by the software and hardware faults which bring innetwork communication interrupt thus affecting networkQoS and usersrsquo QoE [8] At present the existing CODresearches mainly focus on the LTELTE-A HetNet Severaldetection methods use data collected by drive test or sub-scriber complaints to analyze the network faults and celloutages For instance in [9] COD is analyzed autonomouslyby preprocessing the minimization drive testing (MDT)together with local outlier factor based detector (LOFD) andone class support vectormachine based detector (OCSVMD)to detect and localize anomalous network behavior Thesesolutions not only cost much time and manpower cost butalso require expert knowledge or prior experience
Several studies pay attention to COD with KPI variationsuch as handover statistics [10] and a cooperative femtocelloutage detection architecture which consists of a trigger stageand a detection stagewith RSRP is introduced in [11] Furtheran efficient discriminant function is used to complete CODwith CQI and RRC connection reestablishment informationin [12] However these approaches are just suitable fortraditional UMTS or LTELTE-A networks
Recently there have been a few studies that focus onCODwithmachine learning approaches under data collectedfrom users or cells In [13] an unsupervised data miningalgorithm with a reference signal received power (RSRP) andreference signal received quality (RSRQ) was proposed todetect cell outages Reference [14] applied theHiddenMarkovmodel for cell outage detection under RSRP and RSRQas well Further a classification-based approach named 119870-nearest neighbor (KNN) is proposed for COD in [15] andtransductive confidence machines (TCM) based COD withRSRP and SINR data is proposed in [16] Moreover ourprevious work used LOF to detect cell outage with handoverstatistics [17] These works can give suggestions for technol-ogy selection for 5G H-CRAN COD
In particular RRHs failure in 5G H-CRAN may be diffi-cult to detect because their failure may not trigger operation
andmaintenance system alarm which make that self-healingfunction unable to timely compensate for these nodes As aresult the outage may last hours or a few days before beingdiscovered unless the abnormal status is captured by DT testor feedback from users
Rapid fault discovery and localization of abnormal cellscan reduce network paralysis and deterioration caused bynode outage Due to the high density of nodes in H-CRANthe faults of a single RRH will not quickly affect the networkas users may handover or reconnect to neighbor RRHs orACEs It increases the fault tolerance rate but makes CODmore difficult as well thus increasing the instability of thenetwork
Based on above analysis we find that present CODapproaches for LTELTE-A HetNet may not be suitablefor H-CRAN due to their limited data sets and obviousoutage alarmTherefore in this paper we will establish CODarchitecture with complete procedures to handle implicitRRH outages
3 COD Architecture andApproach for 5G H-CRAN
To make it clean in 5G H-RAN we also regard the coverageof each RRH as a cell Still cell definition for ACEs may usetraditional ones As COD for ACEs can be easily resolvedby approach mentioned above in HetNet here we mainlyconcentrate on implicit cell outage for RRHs
For the convenience of illustration we consider a simplecell outage scenario in H-CRAN under a NodeC as shownin Figure 2 Here one ACE and many randomly distributedRRHs are under control of one NodeC When a RRHcell turns into outage its serving users may reconnect toACE or handover to another RRH cell This reconnectionor handover will result in signaling and communicationvariation among users RRHs ACEs and NodeC We requireall these data to execute COD if we want to achieve accurateand timely outage detection
In the above scenario a simple user reconnect procedureis shown in Figure 3 If a user has to establish a com-munication link with a RRH it first sends a ldquoConnectionRequestrdquo command to it and then this RRH will send aldquoResource Requestrdquo command to NodeC Only when NodeCreply a ldquoRequest Responserdquo command to RRH with requiredresource allocation the RRH will give a successful connectreply with ldquoConnection Setuprdquo to users Then the service willgo on It means a user keeps a connection to a RRH underthe control of NodeC with allocated resource If the RRHhas gone into outage state due to cable loss or power offuserrsquos service will fail and it will attempt to reconnect to ACEor other RRHs If NodeC has enough resource connectionbetween user andACEwill be established again as the same asRRH connection Here RRHoutagemay not report toNodeCimmediately However we can count the time of connectrequest fromuser to other RRHandACE as the two countingpoints show in Figure 3 If we store these data and analyze itwith time series fitting or prediction we may get variationfeatures and obtain abnormal points at several time intervalswhich may be useful for RRH COD
4 Wireless Communications and Mobile Computing
RRH
ACE
RRH
Outage RRH
Cloud
Node C
ConnectLost connectionReconnect
RRH FronthaulACE Fronthaul
Figure 2 Cell outage in H-CRAN
Connection requestResource request
Resource responseConnection response
Connection setup
In Service
Outage OccursConnection failure
Connection requestResource request
Resource responseConnection response
Connection setup
Service Recovery
Counting point
Counting point
User RRH ACE NodeC
Figure 3 An instance of user reconnection process
Wireless Communications and Mobile Computing 5
Data Collection
User
RRHACE
NodeC
DT
Data Analysis
Pre-processing
Machine Learning
Temporal Predication
Spatial Predication
COD ConclusionCell Outage Compensation
Input
Input
Input
Input
Guideline
AchievePerformance Evaluation
MDT Data
Network topologyCell parameters
HOConnection RequestNCL
RSRPRSRQSINRCQI
Figure 4 Overview architecture for COD in 5G H-CRAN
Based on the above analysis we want to construct an inte-grated COD architecture for 5G H-CRAN and its detailedintroduction is given below
31 COD Architecture and Procedures Firstly the proposedarchitecture for COD can be found in Figure 4
As shown in Figure 4 COD mainly consists of datacollection and data analysis stages Firstly data collectionshould store data from different sources which are as follows
(1) Data collection from users through measurementreports such as RSRP RSRQ SINR and CQI infor-mation this information is always huge and hard tobe handled synchronously as the time interval is fairlydense So we can set a sampling interval and just takethe statistics for them
(2) Data collection from RRHACE such as HOCon-nection RequestNCL these data may come from theKPI statistics from OAM system These data are per-formance indicators and can be used directly as theinterval is defined beforehand under acceptable level
(3) Data collection for NodeC which may network leveldata such as network topology and preconfiguredcell parameters (transmit power spectrum antennaheight and tilt) these data can be used to supplementthe spatial and temporal analysis of different cells
(4) Drive test data drive test is a tool to verify networkperformance afterward It can be used as a validationfor COD conclusions and thus provide correctionsuggestions for COD approach
After data collection all the data can be put to dataanalysis stage to execute COD As shown in Figure 4 withSON entities which located at NodeC the procedure will beexecuted as follows
(1) SON entities firstly should preprocess these data toimprove the data quality
(2) SON entities adopt temporal and spatial predictionmethod to obtain varying patterns for temporal dataand spatial data respectively
(3) Next SON entities choose proper machine learningapproach to identify the outage cell
Still with COD results we can give suggestions for COCAnd COC effectiveness evaluation can be obtained throughnetwork performance monitoring thus constructing a self-organized loop The critical technologies of preprocessingspatial and temporal prediction andmachine learning will beintroduced next
32 Key Technologies for COD
321 Data Processing Data processing includes two conceptssuch as Data Cleaning and Feature Engineering These twoare compulsory for achieving better accuracy and perfor-mance before machine learning and deep learning And itincludes data cleaning data integration data transformationand data reduction So in our architecture we should chooseproper data processing approach to obtain high-quality data
322 Spatial Prediction Spatial prediction technologiesmainly aim at analysis spatial traffic distributions or cap-ture user variations Currently several methods have beenadopted in this field such as log-normal or Weibull distri-bution used in [18] or traffic patterns identifying methodsproposed in [19] With spatial prediction we can obtaindifferent distribution laws
323 Temporal Prediction Temporal prediction is aimingat predicting future variation direction based on past and
6 Wireless Communications and Mobile Computing
Table 1 Cell level data
Measurements DescriptionTime Transmission time interval (TTI)Cell info Cell global identity (CGI)Cell location Longitude and latitude informationCell inHO Cell incoming handover
current temporal sequence data for instance traffic variationof one ACE or load of NodeC Temporal prediction canadopt HoltndashWinter traffic forecast [20] or S-ARIMA model[21] to construct prediction sequences which will be usefulfor abnormal traffic recognition Moreover joint temporal-spatial predication approaches deservemore attention aswell
324 Machine Learning As analyzed in [22] machine learn-ing has been used for LTELTE-A self-healing and achievesremarkable results However thesemethods aremainly basedon single data source and do not take integrated data setshown in Section 31 together To make COD more accurateand intelligent in 5Gnetworks deep learningmodel [23] suchas DRN model can be considered
4 A Use Case for 5G H-CRAN COD
According to the architecture and procedures above weintroduce a use case to achieve COD for 5G H-CRAN Herewe suppose that the outage RRH or ACE cell cannot carrytraffic anymore Typically at this point all users served bythis cell have to be offloaded to its neighbor cells This willresult in a huge number of handovers in H-CRANThe risinghandovers in neighbor cells could be easily detected bymeansof data mining So we will discuss our algorithm by using thetemporal data of inHO statistics collected from the NodeC
41 Use Case Scenario Description The scenario is shown inFigure 5 It consists of 19 ACE with 25 users under coverageAll these users are active users Each ACE has 3 cells andone RRH cell is located under each ACE cell The UEs arerandomly distributed in cells to generate handover statisticsand measurement reports
Still we mainly collect cell level information stored inNodeC as shown in Table 1
Moreover in order to make the evaluation more obviouswe will evaluate the procedure under a simulation platformdeveloped by our lab Parameter settings can be found inTable 2
In our simulation platformwe assume that rapid decreaseof cellrsquos transmit power indicates cell outage That is we canset cell outage manually to evaluate efficiency of our ap-proach Then users under the affected outage areas switch tothe neighbor cells which will increase the number of han-dovers We will use the inHO data collected from all the cellswith a machine learning methods to evaluate whether outagecells can be detected
Table 2 Simulation parameters
Simulation parameters Value (unit)ACE 19 sites with 3 cells eachRRH 1 cell per siteUser number 25 per ACEUser distribution Uniform random distributionLink direction DownlinkBS Tx power 46 dBmPath loss model Cost231-HataUE velocity 15 kmhUser distribution UniformCell selection criteria Strongest RSRP defines the target cellSimulation length 150 TTI
RRH cellACE
Figure 5 Scenario use case for 5G H-CRAN
42 COD Procedures with M-LOF Considering the situa-tions above and the advantage of LOF shown in [9] a mod-ified LOF based detector is adopted to detect the outage byextracting both spatial and temporal correlations of handoverstatistics over the neighbors of outage cells The procedure isshown in Figure 6
The COD procedure includes four steps data collectionpreprocessingmachine learning and localization and all canbe executed automatically under the control of LS-SON Inthis way the outage cell can be detected timely so we canminimize the performance degradation of the network Herewe will introduce the functions in three steps
421 Preprocessing When data collection is finished SONentities proceed to extract feature vector 119879119894 for LOF [24]
Wireless Communications and Mobile Computing 7
Cell Level Data
Filter Standardization
Database
Low-dimension
Pre-processing
Data Collection
Machine Learning
Detection
Test z scoreCalculation
Store data to database
Localization
ParameterEstimation
No
M-LOFCalculation
OutageYes
H-CRAN
COC
thresholdZ gt
Figure 6 COD procedure with modified LOF
corresponding to each temporal data Since the inHO data iscollected from different cells at the different time the datavector is designed as spatial-temporal series
In fact the inHO statistics of outage cells are zero sincethe KPIs in outage cell cannot be collected Moreover for onepossibility there may be a few zero-inHO cells because noactive users are within their coverage This will enhance thedifficulty of detection In this case we filter all inHO statisticsof zero Here each cell splits out an ordered time series ofinHO To finish length-wise normalization and carry outuniformization and standardization processing we slice thetime series to several subseries
For temporal analysis 5 TTIs data is put into one featurevector 119879119894 as shown in below where 119894 represents the order ofsubseries
119879119894 = inHO1 inHO2 inHO3 inHO4 inHO5 (1)
The feature vector 119879119894 utilizes exactly one cellrsquos inHO datawhich is used to find out the time when an outage happened
For spatial analysis one feature vector 119878119895 consists of 119897cell handover statistics at TTI time 119895 where 119895 represents theidentity of the cell
119878119895 = inHO1 inHO2 inHO3 inHO4 inHO5 inHO6inHO7 inHO8 inHOl
(2)
After slicing the data series linear functional transforma-tion criterion is used to normalize the data to eliminate errorscaused by nonuniform features For example the criterion of119878119895 is defined as
1198781015840119895 =119878119895 minusmin119895 (119878119895)
max119895 (119878119895) minusmin119895 (119878119895) (3)
422 Machine Learning LOF detection is an unsupervisedanomaly detection algorithm The local densities of targetpoints are calculated by LOF used to compare with itsneighbors The larger the difference between the sampleand its neighbors is the larger outage factor score will beassigned to the sampleThe advantage of LOF is that no priorknowledge is required in advance to detect the unknowncell outage The detailed definitions of LOF are explained asfollows
Definition 1 k-distance of cell 119901 denoted as 119889119896(119901) representsthe distance between cell 119901 and its 119896th nearest neighbor sothat for at least 119896 cell 119900119894 satisfies
119889 (119900119894 119901) le 119889 (119900119896 119901) (4)
and for at most 119896 minus 1 cells 119900119894 satisfies119889 (119900119894 119901) lt 119889 (119900119896 119901) (5)
Definition 2 119896-distance neighborhood of cell 119901 denoted as119873119889119896(119901)(119901) when 119889119896(119901)is given 119873119889119896(119901)(119901) is the set whichincludes every cell whose distance to cell 119901 is smaller than119889119896(119901)Definition 3 Reachability distance of cell 119901 with respect tocell 119900 is denoted as rd119896(119901 119900) it is the maximum one of 119889119896(119901)and the distance between 119901 and 119900
rd119896 (119901 119900) = max 119889119896 (119900) 119889 (119900 119901) (6)
Definition 4 Local reachability density of cell p denoted aslrd119896(119901) is shown as follows
lrd119896 (119901) =10038161003816100381610038161003816119873119889119896(119901) (119901)
10038161003816100381610038161003816sum119900isin119873119889119896(119901)(119901) rd119896 (119901 o)
(7)
Definition 5 Local outlier factor of cell p denoted asLOF119896(119901) can be obtained as below
LOF119896 (119901) =sum119900isin119873119889119896(119901)(119901) (lrd119896 (119900) lrd119896 (119901))10038161003816100381610038161003816119873119889119896(119901) (119901)
10038161003816100381610038161003816 (8)
The above descriptions give the definition of LOF Fromour previous work in [17] the proposed M-LOF is a properapproach for heterogeneous network so here we use themodified LOF algorithm called M-LOF to make a refinedchoice for H-CRAN Definitions of M-LOF and m-distanceare given below
8 Wireless Communications and Mobile Computing
0
0
InHo DataMDT DataAbnormal cells
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80
Cell Id
minus15minus10minus05000510152025303540455055
MLO
F Va
lue
1
23
4
5
6
7
8
9
10
11
12
13
14
1516
17
2122
24
25 27
29
30
3233
35
36
38 4143
44
4647
49
52
54
55
56
57
58
59
6062
63
64
65
66
67
69
68
7071
72
73
74
75
76
1
2
3
4
5
6
7
89
10
11
1213
14
15
1617
21
22 2425 27 29
30 32
33
35 36 38
41
43
44
46
47
49
52
54
55
56575859
6062
63
6465
66
6768
69
70717273
74
75
76
Figure 7 Spatial analysis for abnormal cells
Definition 6 119898-distance of cell 119901 given the positive integer119896 is defined as
119898minus119889119896 (119901) = 120576 + [[sum119900isin119873119889119896(119901)(119901) 119889 (119901 119900)10038161003816100381610038161003816119873119889119896(119901) (119901)
10038161003816100381610038161003816]] (9)
Here 120576 is a constant value to enhance the accuracyDefinition 7 119898-distance neighborhood of cell p denotedas 119873119898minus119889119896(119901)(119901) is the set which includes every cell whichdistance to cell 119901 is smaller thanm-distance
Definition 8 Reachability distance of m-distance of cell 119901concerning cell o denoted as 119903minus119889119898(119901 119900) is
119903minus119889119898 (119901 119900) = max 119898minus119889119896 (119900) 119889 (119900 119901) (10)
Definition 9 Local reachability density of cell p is defined as
lrd119898 (119901) =10038161003816100381610038161003816119873119898minus119889119896(119901) (119901)
10038161003816100381610038161003816sum119900isin119873119898minus119889119896(119901)(119901) 119903minus119889119898 (119901 o)
(11)
Definition 10 Modified local outlier factor of cell p is shownbelow
LOF119898 (119901) =sum119900isin119873119898minus119889119896(119901)(119901) (lrd119898 (119900) lrd119898 (119901))10038161003816100381610038161003816119873119898minus119889119896(119901) (119901)
10038161003816100381610038161003816 (12)
According to the definition of LOF119896(119901) and LOF119898(119901) thechoice of 119896 is exactly sensitive As a consequence if there is aprior experience we will choose the cross-validation methodto estimate the parameter k
423 Localization The last step is localization based on theoutput of LOF calculation Here the neighbor cell list [25] isused to search the relation between the outage cell and itsneighbor cells by geographic informationwith 119911-score In thisway the outage cells may be localized
43 Results Analysis Based on the above steps M-LOFwould detect anomalous behaviors of handover And in thesimulation the value of 119896 for the LOF based detector is foundbetween 5 and 14
For the purpose of validation we will first analyze theresults from the spatial and temporal perspectives and thendetermine outage cell locations with neighbor cell list andperformance comparison between M-LOF and LOF will begiven at last
431 Spatial Analysis For spatial analysis we focus on thetemporal data of different cells at the same time periodFigure 7 shows the M-LOF of each cell at 95th TTI It can beseen that factor values derived from six cellsrsquo inHO data arefar higher than the normal value which is usually less than15 Therefore the abnormal cells can be distinguished fromnormal cells as they have experienced many user reconnec-tions
For comparison Figure 7 also shows the detection resultusingMDTmeasurements (with RSRP and SINR) as the datasource It can be seen that the abnormal cells are hard to bedetected using MDT Because M-LOF values of cells basedon MDT data are smoother than inHO data In contrastour proposed method using the inHO data has a betterperformance for abnormal cell detection in H-CRAN
432 Temporal Analysis Figure 8 shows the temporal analy-sis result for M-LOF variations relative to TTI for outage cell13 At this time from the results we find that M-LOF valuearises between TTI 90 and TTI 100 which is in accordancewith our outage time setting As initial connections of usersoccur between 0ndash5 TTI under unstable status so we justignore these data
433 Localization After the implementation of M-LOFbased detector neighbor cell list is used to localize the outage
Wireless Communications and Mobile Computing 9
TTI
Outage Period
22
20
18
16
14
12
10
08
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140
MLO
F Va
lue 5
1015
2025 30
35 40
4550
55
60
6570
75
80
85
90
95
100 105
110
115120
125
130
135140
Figure 8 Temporal analysis for outage TTI
cells First we need to set a threshold to filter anomalous cellsthrough 119911-score calculation as
119911119896 = 119899119896 minus 120583119899120590119899 (13)
where 119899119896 is the M-LOF value of kth cell and 120583119899 120590119899 arethe mean and standard deviation of M-LOF value for theanomaly scores of the other cells The reference 119911-scorethreshold is configured with a preferred value for the abnor-mal cells Here the reference 119911-score is set to 21 according tothe computing results We conducted two sets of evaluationsto make a comparison of performance of COD based oninHO data and MDT data respectively The results canidentify outage cell 18 and cell 61 which is the right one asshown in Figure 9 as LOF values of their neighbor cells arerelatively higher than other cells
434 Performance Evaluation In this part we analyzedthe M-LOF detection performance under varying trafficconditions Since the behaviors of users have a direct effect oninHO data the diagnosis process has been tested in differentscenarios by changing the User Density (UD) and UserVelocity (UV) parameters as the baseline setup To evaluatethe impact of the variations of UD and UV on M-LOFvalues different scenarios are set up by adjusting these twoparameters The Cumulative Distribution Function (CDF) oftheM-LOF values under different UV conditions is shown inFigure 10
It can be seen that for the low-velocity scenario almost80 of the M-LOF values are less than 05 However thereis a significant reduction in the M-LOF value as the UVincreases Likewise a similar behavior is observed with theincrease of UD as shown in Figure 11 The UV and UDparameters influence the distribution and spread of the inHOdata as explained earlier and consequently the value of M-LOF This leads to a low detection performance of M-LOFsince it generates an increased number of false alarms
Table 3 Performance evaluation
Result LOF value M-LOF valueFalse negative rate 3 3False positive rate 12 6
Outage cellAffected cell
800
600
400
200
Use
r Pos
ition
Y
User Position X
100 200 300 400 500 600 700 800
1
2
3
4
5
6
7
8
9
10
11
2122
25
27
29
30
32
33
64 63
62
60
1258 71
1773
16
68
36
35
14
6113
59
2456
57
74
1875
65
76
55
0
72
49
52
54
47
38 43
6615
67
70 69
41
44
46
Figure 9 Outage cell location results
At last this paper compares LOF and M-LOF by evaluat-ing False Positive Rate (FPR) and False Negative Rate (FNR)and the final results of the simulation are shown in Table 3The FNR represents the chance that an outage cell is notrecognized from the outage cells while the FPR representsthe chance that a normal cell is recognized as the outage cellfrom all normal cells
From Table 3 we can see that FPR and FNR of LOF baseddetector are 12 and 3 respectively It means the outagecell can be almost wholly detected although a small part ofthe normal cell may be determined as an anomaly The maincause of the 3FNR is that a small number of anomalous cellshave low traffic so that seldom user data cannot be collectedwhen they are in outage So these small cells are failed to bedetected However these outages have little impact on overallnetwork performance and user experience Therefore wecan say the outage cells can be detected successfully withLOF based detector approach As for M-LOF the FPR is 6which is smaller than LOFThe reason whyM-LOF has betterperformance is that M-LOF pays more attention to localdensity This is in line with the actual use of outage detectionsince the abnormal handovers caused by the neighbor outagecell are also localized
Still asM-LOF just modified the distance computation ofLOF so for time complexity it is just the same as LOF withis 119874(1198992) Here 119899 is the TTI number or cell number shown inFigure 7 or Figure 8
10 Wireless Communications and Mobile Computing
10
08
06
04
02
00minus2 minus1 0 1 2 3 4 5
0mM lt 5 6FI=CNS lt 1ms1mM lt 5 6FI=CNS lt 3ms3mM lt 5 6FI=CNS lt 5ms
M-LOF
CDF
Figure 10 CDF of M-LOF for User Velocity
10
08
06
04
02
00minus2 minus1 0 1 2 3 4 65
M-LOF
CDF
UE Density = 30 UsersCellUE Density = 20 UsersCellUE Density = 10 UsersCell
Figure 11 CDF of M-LOF value for User Density
5 Conclusions
In this paper we propose self-organized cell outage detectionarchitecture and approach for 5G H-CRAN The previousmethods using single data source are hard to detect thehidden outage such as RRHs Aiming at this problem weintroduce architecture and corresponding approach to realizeCOD for H-CRAN and still analyze the required key tech-nologies Then we evaluate the approach under a use casewhich utilizes the collected inHO statistics and adopt M-LOF to detect the anomalous behavior outage cells Resultsshow that the use case is effective However much work stillhas to be done for this architecture as we do not considerresource allocation and COC solutions still we have to takemore intelligent learning model such as deep learning into
account as well and compare the efficiency of different algo-rithms under a embodied H-CRAN network deploymentsMoreover we will focus on a specific perspective for self-healing approaches for H-CRAN and propose more practicalsolutions
Conflicts of Interest
The authors declare that they have no conflicts of interest
Acknowledgments
This research is supported by the National Science andTechnology Major Project of the Ministry of Science andTechnology of China (no 2018ZX030110004)
References
[1] P Rost A Banchs I Berberana et al ldquoMobile network archi-tecture evolution toward 5Grdquo IEEE Communications Magazinevol 54 no 5 pp 84ndash91 2016
[2] M Peng Y Sun X Li Z Mao and C Wang ldquoRecentadvances in cloud radio access networks system architectureskey techniques and open issuesrdquo IEEECommunications Surveysamp Tutorials vol 18 no 3 pp 2282ndash2308 2016
[3] M Peng Y Li Z Zhao and C Wang ldquoSystem architectureand key technologies for 5G heterogeneous cloud radio accessnetworksrdquo IEEE Network vol 29 no 2 pp 6ndash14 2015
[4] N Chen B Rong X Zhang and M Kadoch ldquoScalable andflexible massive MIMO precoding for 5G H-CRANrdquo IEEEWireless Communications Magazine vol 24 no 1 pp 46ndash522017
[5] OGAliu A ImranMA Imran andB Evans ldquoA survey of selforganisation in future cellular networksrdquo IEEECommunicationsSurveys amp Tutorials vol 15 no 1 pp 336ndash361 2013
[6] E J Khatib R Barco P Munoz I D La Bandera and ISerrano ldquoSelf-healing in mobile networks with big datardquo IEEECommunications Magazine vol 54 no 1 pp 114ndash120 2016
[7] M Amirijoo L Jorguseski T Kurner et al ldquoCell outagemanagement in LTE networksrdquo in Proceedings of the 2009 6thInternational Symposium on Wireless Communication Systems(ISWCS 2009) pp 600ndash604 Siena Italy September 2009
[8] M YinW Li L Feng P Yu and X Qiu ldquoMulti-cell cooperativeoutage compensation in cloud-RANs based 5G public safetynetworkrdquo IEEE Access vol 5 pp 17309ndash17321 2017
[9] A Zoha A Saeed A Imran M A Imran and A Abu-DayyaldquoData-driven analytics for automated cell outage detection inSelf-Organizing Networksrdquo in Proceedings of the 2015 11th Inter-national Conference on the Design of Reliable CommunicationNetworks DRCN 2015 pp 203ndash210 usa March 2015
[10] I De-La-Bandera R Barco P Munoz and I Serrano ldquoCellOutage Detection Based onHandover Statisticsrdquo IEEE Commu-nications Letters vol 19 no 7 pp 1189ndash1192 2015
[11] W Wang Q Liao and Q Zhang ldquoCOD A cooperativecell outage detection architecture for self-organizing femtocellnetworksrdquo IEEE Transactions on Wireless Communications vol13 no 11 pp 6007ndash6014 2014
[12] Q Liao M Wiczanowski and S Stanczak ldquoToward cell outagedetection with composite hypothesis testingrdquo in Proceedingsof the 2012 IEEE International Conference on CommunicationsICC 2012 pp 4883ndash4887 can June 2012
Wireless Communications and Mobile Computing 11
[13] Y Ma M Peng W Xue and X Ji ldquoA dynamic affinity propa-gation clustering algorithm for cell outage detection in self-healing networksrdquo in Proceedings of the 2013 IEEE WirelessCommunications and Networking Conference WCNC 2013 pp2266ndash2270 chn April 2013
[14] M Alias N Saxena and A Roy ldquoEfficient Cell Outage De-tection in 5G HetNets Using Hidden Markov Modelrdquo IEEECommunications Letters vol 20 no 3 pp 562ndash565 2016
[15] W Xue M Peng Y Ma and H Zhang ldquoClassification-basedapproach for cell outage detection in self-healing heterogeneousnetworksrdquo in Proceedings of the 2014 IEEEWireless Communica-tions and Networking Conference WCNC 2014 pp 2822ndash2826tur April 2014
[16] J Wang N Q Phan Z Pan N Liu X You and T Deng ldquoAnimproved TCM-based approach for cell outage detection forself-healing in LTE HetNetsrdquo in Proceedings of the 83rd IEEEVehicular Technology Conference VTC Spring 2016 chn May2016
[17] T Zhang L Feng P Yu S Guo W Li and X Qiu ldquoA handoverstatistics based approach for Cell Outage Detection in self-organized Heterogeneous Networksrdquo in Proceedings of the 15thIFIPIEEE International Symposium on Integrated Network andService Management IM 2017 pp 628ndash631 prt May 2017
[18] D Lee S Zhou X Zhong Z Niu X Zhou and H ZhangldquoSpatial modeling of the traffic density in cellular networksrdquoIEEE Wireless Communications Magazine vol 21 no 1 pp 80ndash88 2014
[19] F Xu Y Li H Wang P Zhang and D Jin ldquoUnderstandingMobile Traffic Patterns of Large Scale Cellular Towers in UrbanEnvironmentrdquo IEEEACM Transactions on Networking vol 25no 2 pp 1147ndash1161 2017
[20] S Morosi P Piunti and E Del Re ldquoSleep mode managementin cellular networks a traffic based technique enabling energysavingrdquo Transactions on Emerging Telecommunications Tech-nologies vol 24 no 3 pp 331ndash341 2013
[21] W Y Szeto B Ghosh B Basu andM OrsquoMahony ldquoMultivariatetraffic forecasting technique using cell transmission model andSARIMA modelrdquo Journal of Transportation Engineering vol135 no 9 pp 658ndash667 2009
[22] P V KlaineM A Imran O Onireti and R D Souza ldquoA Surveyof Machine Learning Techniques Applied to Self OrganizingCellular Networksrdquo IEEE Communications Surveys amp Tutorials2017
[23] Z M Fadlulla F Tang B Mao and et al ldquoState-of-the-Art Deep Learning Evolving Machine Intelligence TowardTomorrowrsquos Intelligent Network Traffic Control Systemsrdquo IEEECommunications Surveys amp Tutorials vol 19 no 4 pp 2432ndash2455 2017
[24] M M Breuniq H Kriegel R T Ng and J Sander ldquoLOFidentifying density-based local outliersrdquo SIGMOD Record vol29 no 2 pp 93ndash104 2000
[25] C M Mueller M Kaschub C Blankenhorn and S WankeldquoA Cell Outage Detection Algorithm Using Neighbor Cell ListReportsrdquo in Self-Organizing Systems vol 5343 of Lecture Notesin Computer Science pp 218ndash229 Springer Berlin HeidelbergBerlin Heidelberg 2008
International Journal of
AerospaceEngineeringHindawiwwwhindawicom Volume 2018
RoboticsJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Active and Passive Electronic Components
VLSI Design
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Shock and Vibration
Hindawiwwwhindawicom Volume 2018
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawiwwwhindawicom
Volume 2018
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Control Scienceand Engineering
Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom
Journal ofEngineeringVolume 2018
SensorsJournal of
Hindawiwwwhindawicom Volume 2018
International Journal of
RotatingMachinery
Hindawiwwwhindawicom Volume 2018
Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Navigation and Observation
International Journal of
Hindawi
wwwhindawicom Volume 2018
Advances in
Multimedia
Submit your manuscripts atwwwhindawicom
2 Wireless Communications and Mobile Computing
Internet
NodeC NodeC NodeC
ACE
RRH
RRH
RRH
RRH
RRH
RRH
RRH Fronthaul
D2D
Wireless communication linkACE Fronthaul
middot middot middot middot middot middot
Figure 1 A brief architecture of H-CRAN
Internet of things (IoT) devices and may adopt advancedcommunication technologies such as D2D [4]
To realize the universal plug and play function offloadnetwork traffic immediately and manage the computing andspectrum resource dynamically SON plays an importantrole to realize intelligent management of H-CRAN SON isproposed to reduce service providersrsquo operating expenses inLTELTE-A systems and HetNet [5] H-CRAN consists ofa large number of heterogeneous access nodes and cloudcomputing resource units and its resource should be virtu-alized for sharing as well Therefore a large-scale SON (LS-SON) that integrates a unified autonomic process of ultra-computing ultra-planning ultra-configuration and ultra-optimization is preferred [3]
LS-SON can reduce the complexity of cochannel interfer-ence management in H-CRAN to save the operating costs
of all RRHs and ACEs So it is used to coordinate themanagement functions of the entire network and improve theoverall operational efficiency Because NodeC needs to servemultiple RANs and cooperate with RRHs it is consideredto implement self-configuring self-optimization and self-healing by using a centralized SON architecture for LS-SON
As one of the critical functions and use cases of SONself-healing not only identifies fault events but also is able todiagnose the causes (for example deciding why it happened)and then triggers appropriate compensation mechanism toturn network to normal state [6] In H-CRAN it first has toperceive a fault that has occurred or is about to occur andthen adopts proper actions to recover the services (partiallyor wholly definitely or temporarily)
According to 3GPP standard the self-healing functioncan be further divided into multiple use cases including fault
Wireless Communications and Mobile Computing 3
diagnosis fault classification and COM COM first needs todetect the cell outage automatically and then perform a rea-sonable compensation mechanism to repair the faults so asto minimize the impact of cell outage [7] Therefore COD isan essential prerequisite for self-healing
At present the studies of COD mainly aim at tradi-tional LTELTE-AHetNet with limited data collection whichmay not be appropriate for the complex H-CRAN networkarchitecture This paper proposes a self-organized CODarchitecture and a corresponding approach for H-CRANbased on previous work and gives a use case with HO dataanalysis to evaluate them
The rest of the paper is organized as follows Section 2introduces the related work for COD Section 3 proposes thearchitecture and approach for 5GH-CRANCODmeanwhileintroducing COD procedures and related technologies InSection 4 a CODuse case forHOdata analysis withmodifiedM-LOF is introduced and conclusions are given in Section 5
2 Related Work
In wireless communication networks cell outage is mainlycaused by the software and hardware faults which bring innetwork communication interrupt thus affecting networkQoS and usersrsquo QoE [8] At present the existing CODresearches mainly focus on the LTELTE-A HetNet Severaldetection methods use data collected by drive test or sub-scriber complaints to analyze the network faults and celloutages For instance in [9] COD is analyzed autonomouslyby preprocessing the minimization drive testing (MDT)together with local outlier factor based detector (LOFD) andone class support vectormachine based detector (OCSVMD)to detect and localize anomalous network behavior Thesesolutions not only cost much time and manpower cost butalso require expert knowledge or prior experience
Several studies pay attention to COD with KPI variationsuch as handover statistics [10] and a cooperative femtocelloutage detection architecture which consists of a trigger stageand a detection stagewith RSRP is introduced in [11] Furtheran efficient discriminant function is used to complete CODwith CQI and RRC connection reestablishment informationin [12] However these approaches are just suitable fortraditional UMTS or LTELTE-A networks
Recently there have been a few studies that focus onCODwithmachine learning approaches under data collectedfrom users or cells In [13] an unsupervised data miningalgorithm with a reference signal received power (RSRP) andreference signal received quality (RSRQ) was proposed todetect cell outages Reference [14] applied theHiddenMarkovmodel for cell outage detection under RSRP and RSRQas well Further a classification-based approach named 119870-nearest neighbor (KNN) is proposed for COD in [15] andtransductive confidence machines (TCM) based COD withRSRP and SINR data is proposed in [16] Moreover ourprevious work used LOF to detect cell outage with handoverstatistics [17] These works can give suggestions for technol-ogy selection for 5G H-CRAN COD
In particular RRHs failure in 5G H-CRAN may be diffi-cult to detect because their failure may not trigger operation
andmaintenance system alarm which make that self-healingfunction unable to timely compensate for these nodes As aresult the outage may last hours or a few days before beingdiscovered unless the abnormal status is captured by DT testor feedback from users
Rapid fault discovery and localization of abnormal cellscan reduce network paralysis and deterioration caused bynode outage Due to the high density of nodes in H-CRANthe faults of a single RRH will not quickly affect the networkas users may handover or reconnect to neighbor RRHs orACEs It increases the fault tolerance rate but makes CODmore difficult as well thus increasing the instability of thenetwork
Based on above analysis we find that present CODapproaches for LTELTE-A HetNet may not be suitablefor H-CRAN due to their limited data sets and obviousoutage alarmTherefore in this paper we will establish CODarchitecture with complete procedures to handle implicitRRH outages
3 COD Architecture andApproach for 5G H-CRAN
To make it clean in 5G H-RAN we also regard the coverageof each RRH as a cell Still cell definition for ACEs may usetraditional ones As COD for ACEs can be easily resolvedby approach mentioned above in HetNet here we mainlyconcentrate on implicit cell outage for RRHs
For the convenience of illustration we consider a simplecell outage scenario in H-CRAN under a NodeC as shownin Figure 2 Here one ACE and many randomly distributedRRHs are under control of one NodeC When a RRHcell turns into outage its serving users may reconnect toACE or handover to another RRH cell This reconnectionor handover will result in signaling and communicationvariation among users RRHs ACEs and NodeC We requireall these data to execute COD if we want to achieve accurateand timely outage detection
In the above scenario a simple user reconnect procedureis shown in Figure 3 If a user has to establish a com-munication link with a RRH it first sends a ldquoConnectionRequestrdquo command to it and then this RRH will send aldquoResource Requestrdquo command to NodeC Only when NodeCreply a ldquoRequest Responserdquo command to RRH with requiredresource allocation the RRH will give a successful connectreply with ldquoConnection Setuprdquo to users Then the service willgo on It means a user keeps a connection to a RRH underthe control of NodeC with allocated resource If the RRHhas gone into outage state due to cable loss or power offuserrsquos service will fail and it will attempt to reconnect to ACEor other RRHs If NodeC has enough resource connectionbetween user andACEwill be established again as the same asRRH connection Here RRHoutagemay not report toNodeCimmediately However we can count the time of connectrequest fromuser to other RRHandACE as the two countingpoints show in Figure 3 If we store these data and analyze itwith time series fitting or prediction we may get variationfeatures and obtain abnormal points at several time intervalswhich may be useful for RRH COD
4 Wireless Communications and Mobile Computing
RRH
ACE
RRH
Outage RRH
Cloud
Node C
ConnectLost connectionReconnect
RRH FronthaulACE Fronthaul
Figure 2 Cell outage in H-CRAN
Connection requestResource request
Resource responseConnection response
Connection setup
In Service
Outage OccursConnection failure
Connection requestResource request
Resource responseConnection response
Connection setup
Service Recovery
Counting point
Counting point
User RRH ACE NodeC
Figure 3 An instance of user reconnection process
Wireless Communications and Mobile Computing 5
Data Collection
User
RRHACE
NodeC
DT
Data Analysis
Pre-processing
Machine Learning
Temporal Predication
Spatial Predication
COD ConclusionCell Outage Compensation
Input
Input
Input
Input
Guideline
AchievePerformance Evaluation
MDT Data
Network topologyCell parameters
HOConnection RequestNCL
RSRPRSRQSINRCQI
Figure 4 Overview architecture for COD in 5G H-CRAN
Based on the above analysis we want to construct an inte-grated COD architecture for 5G H-CRAN and its detailedintroduction is given below
31 COD Architecture and Procedures Firstly the proposedarchitecture for COD can be found in Figure 4
As shown in Figure 4 COD mainly consists of datacollection and data analysis stages Firstly data collectionshould store data from different sources which are as follows
(1) Data collection from users through measurementreports such as RSRP RSRQ SINR and CQI infor-mation this information is always huge and hard tobe handled synchronously as the time interval is fairlydense So we can set a sampling interval and just takethe statistics for them
(2) Data collection from RRHACE such as HOCon-nection RequestNCL these data may come from theKPI statistics from OAM system These data are per-formance indicators and can be used directly as theinterval is defined beforehand under acceptable level
(3) Data collection for NodeC which may network leveldata such as network topology and preconfiguredcell parameters (transmit power spectrum antennaheight and tilt) these data can be used to supplementthe spatial and temporal analysis of different cells
(4) Drive test data drive test is a tool to verify networkperformance afterward It can be used as a validationfor COD conclusions and thus provide correctionsuggestions for COD approach
After data collection all the data can be put to dataanalysis stage to execute COD As shown in Figure 4 withSON entities which located at NodeC the procedure will beexecuted as follows
(1) SON entities firstly should preprocess these data toimprove the data quality
(2) SON entities adopt temporal and spatial predictionmethod to obtain varying patterns for temporal dataand spatial data respectively
(3) Next SON entities choose proper machine learningapproach to identify the outage cell
Still with COD results we can give suggestions for COCAnd COC effectiveness evaluation can be obtained throughnetwork performance monitoring thus constructing a self-organized loop The critical technologies of preprocessingspatial and temporal prediction andmachine learning will beintroduced next
32 Key Technologies for COD
321 Data Processing Data processing includes two conceptssuch as Data Cleaning and Feature Engineering These twoare compulsory for achieving better accuracy and perfor-mance before machine learning and deep learning And itincludes data cleaning data integration data transformationand data reduction So in our architecture we should chooseproper data processing approach to obtain high-quality data
322 Spatial Prediction Spatial prediction technologiesmainly aim at analysis spatial traffic distributions or cap-ture user variations Currently several methods have beenadopted in this field such as log-normal or Weibull distri-bution used in [18] or traffic patterns identifying methodsproposed in [19] With spatial prediction we can obtaindifferent distribution laws
323 Temporal Prediction Temporal prediction is aimingat predicting future variation direction based on past and
6 Wireless Communications and Mobile Computing
Table 1 Cell level data
Measurements DescriptionTime Transmission time interval (TTI)Cell info Cell global identity (CGI)Cell location Longitude and latitude informationCell inHO Cell incoming handover
current temporal sequence data for instance traffic variationof one ACE or load of NodeC Temporal prediction canadopt HoltndashWinter traffic forecast [20] or S-ARIMA model[21] to construct prediction sequences which will be usefulfor abnormal traffic recognition Moreover joint temporal-spatial predication approaches deservemore attention aswell
324 Machine Learning As analyzed in [22] machine learn-ing has been used for LTELTE-A self-healing and achievesremarkable results However thesemethods aremainly basedon single data source and do not take integrated data setshown in Section 31 together To make COD more accurateand intelligent in 5Gnetworks deep learningmodel [23] suchas DRN model can be considered
4 A Use Case for 5G H-CRAN COD
According to the architecture and procedures above weintroduce a use case to achieve COD for 5G H-CRAN Herewe suppose that the outage RRH or ACE cell cannot carrytraffic anymore Typically at this point all users served bythis cell have to be offloaded to its neighbor cells This willresult in a huge number of handovers in H-CRANThe risinghandovers in neighbor cells could be easily detected bymeansof data mining So we will discuss our algorithm by using thetemporal data of inHO statistics collected from the NodeC
41 Use Case Scenario Description The scenario is shown inFigure 5 It consists of 19 ACE with 25 users under coverageAll these users are active users Each ACE has 3 cells andone RRH cell is located under each ACE cell The UEs arerandomly distributed in cells to generate handover statisticsand measurement reports
Still we mainly collect cell level information stored inNodeC as shown in Table 1
Moreover in order to make the evaluation more obviouswe will evaluate the procedure under a simulation platformdeveloped by our lab Parameter settings can be found inTable 2
In our simulation platformwe assume that rapid decreaseof cellrsquos transmit power indicates cell outage That is we canset cell outage manually to evaluate efficiency of our ap-proach Then users under the affected outage areas switch tothe neighbor cells which will increase the number of han-dovers We will use the inHO data collected from all the cellswith a machine learning methods to evaluate whether outagecells can be detected
Table 2 Simulation parameters
Simulation parameters Value (unit)ACE 19 sites with 3 cells eachRRH 1 cell per siteUser number 25 per ACEUser distribution Uniform random distributionLink direction DownlinkBS Tx power 46 dBmPath loss model Cost231-HataUE velocity 15 kmhUser distribution UniformCell selection criteria Strongest RSRP defines the target cellSimulation length 150 TTI
RRH cellACE
Figure 5 Scenario use case for 5G H-CRAN
42 COD Procedures with M-LOF Considering the situa-tions above and the advantage of LOF shown in [9] a mod-ified LOF based detector is adopted to detect the outage byextracting both spatial and temporal correlations of handoverstatistics over the neighbors of outage cells The procedure isshown in Figure 6
The COD procedure includes four steps data collectionpreprocessingmachine learning and localization and all canbe executed automatically under the control of LS-SON Inthis way the outage cell can be detected timely so we canminimize the performance degradation of the network Herewe will introduce the functions in three steps
421 Preprocessing When data collection is finished SONentities proceed to extract feature vector 119879119894 for LOF [24]
Wireless Communications and Mobile Computing 7
Cell Level Data
Filter Standardization
Database
Low-dimension
Pre-processing
Data Collection
Machine Learning
Detection
Test z scoreCalculation
Store data to database
Localization
ParameterEstimation
No
M-LOFCalculation
OutageYes
H-CRAN
COC
thresholdZ gt
Figure 6 COD procedure with modified LOF
corresponding to each temporal data Since the inHO data iscollected from different cells at the different time the datavector is designed as spatial-temporal series
In fact the inHO statistics of outage cells are zero sincethe KPIs in outage cell cannot be collected Moreover for onepossibility there may be a few zero-inHO cells because noactive users are within their coverage This will enhance thedifficulty of detection In this case we filter all inHO statisticsof zero Here each cell splits out an ordered time series ofinHO To finish length-wise normalization and carry outuniformization and standardization processing we slice thetime series to several subseries
For temporal analysis 5 TTIs data is put into one featurevector 119879119894 as shown in below where 119894 represents the order ofsubseries
119879119894 = inHO1 inHO2 inHO3 inHO4 inHO5 (1)
The feature vector 119879119894 utilizes exactly one cellrsquos inHO datawhich is used to find out the time when an outage happened
For spatial analysis one feature vector 119878119895 consists of 119897cell handover statistics at TTI time 119895 where 119895 represents theidentity of the cell
119878119895 = inHO1 inHO2 inHO3 inHO4 inHO5 inHO6inHO7 inHO8 inHOl
(2)
After slicing the data series linear functional transforma-tion criterion is used to normalize the data to eliminate errorscaused by nonuniform features For example the criterion of119878119895 is defined as
1198781015840119895 =119878119895 minusmin119895 (119878119895)
max119895 (119878119895) minusmin119895 (119878119895) (3)
422 Machine Learning LOF detection is an unsupervisedanomaly detection algorithm The local densities of targetpoints are calculated by LOF used to compare with itsneighbors The larger the difference between the sampleand its neighbors is the larger outage factor score will beassigned to the sampleThe advantage of LOF is that no priorknowledge is required in advance to detect the unknowncell outage The detailed definitions of LOF are explained asfollows
Definition 1 k-distance of cell 119901 denoted as 119889119896(119901) representsthe distance between cell 119901 and its 119896th nearest neighbor sothat for at least 119896 cell 119900119894 satisfies
119889 (119900119894 119901) le 119889 (119900119896 119901) (4)
and for at most 119896 minus 1 cells 119900119894 satisfies119889 (119900119894 119901) lt 119889 (119900119896 119901) (5)
Definition 2 119896-distance neighborhood of cell 119901 denoted as119873119889119896(119901)(119901) when 119889119896(119901)is given 119873119889119896(119901)(119901) is the set whichincludes every cell whose distance to cell 119901 is smaller than119889119896(119901)Definition 3 Reachability distance of cell 119901 with respect tocell 119900 is denoted as rd119896(119901 119900) it is the maximum one of 119889119896(119901)and the distance between 119901 and 119900
rd119896 (119901 119900) = max 119889119896 (119900) 119889 (119900 119901) (6)
Definition 4 Local reachability density of cell p denoted aslrd119896(119901) is shown as follows
lrd119896 (119901) =10038161003816100381610038161003816119873119889119896(119901) (119901)
10038161003816100381610038161003816sum119900isin119873119889119896(119901)(119901) rd119896 (119901 o)
(7)
Definition 5 Local outlier factor of cell p denoted asLOF119896(119901) can be obtained as below
LOF119896 (119901) =sum119900isin119873119889119896(119901)(119901) (lrd119896 (119900) lrd119896 (119901))10038161003816100381610038161003816119873119889119896(119901) (119901)
10038161003816100381610038161003816 (8)
The above descriptions give the definition of LOF Fromour previous work in [17] the proposed M-LOF is a properapproach for heterogeneous network so here we use themodified LOF algorithm called M-LOF to make a refinedchoice for H-CRAN Definitions of M-LOF and m-distanceare given below
8 Wireless Communications and Mobile Computing
0
0
InHo DataMDT DataAbnormal cells
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80
Cell Id
minus15minus10minus05000510152025303540455055
MLO
F Va
lue
1
23
4
5
6
7
8
9
10
11
12
13
14
1516
17
2122
24
25 27
29
30
3233
35
36
38 4143
44
4647
49
52
54
55
56
57
58
59
6062
63
64
65
66
67
69
68
7071
72
73
74
75
76
1
2
3
4
5
6
7
89
10
11
1213
14
15
1617
21
22 2425 27 29
30 32
33
35 36 38
41
43
44
46
47
49
52
54
55
56575859
6062
63
6465
66
6768
69
70717273
74
75
76
Figure 7 Spatial analysis for abnormal cells
Definition 6 119898-distance of cell 119901 given the positive integer119896 is defined as
119898minus119889119896 (119901) = 120576 + [[sum119900isin119873119889119896(119901)(119901) 119889 (119901 119900)10038161003816100381610038161003816119873119889119896(119901) (119901)
10038161003816100381610038161003816]] (9)
Here 120576 is a constant value to enhance the accuracyDefinition 7 119898-distance neighborhood of cell p denotedas 119873119898minus119889119896(119901)(119901) is the set which includes every cell whichdistance to cell 119901 is smaller thanm-distance
Definition 8 Reachability distance of m-distance of cell 119901concerning cell o denoted as 119903minus119889119898(119901 119900) is
119903minus119889119898 (119901 119900) = max 119898minus119889119896 (119900) 119889 (119900 119901) (10)
Definition 9 Local reachability density of cell p is defined as
lrd119898 (119901) =10038161003816100381610038161003816119873119898minus119889119896(119901) (119901)
10038161003816100381610038161003816sum119900isin119873119898minus119889119896(119901)(119901) 119903minus119889119898 (119901 o)
(11)
Definition 10 Modified local outlier factor of cell p is shownbelow
LOF119898 (119901) =sum119900isin119873119898minus119889119896(119901)(119901) (lrd119898 (119900) lrd119898 (119901))10038161003816100381610038161003816119873119898minus119889119896(119901) (119901)
10038161003816100381610038161003816 (12)
According to the definition of LOF119896(119901) and LOF119898(119901) thechoice of 119896 is exactly sensitive As a consequence if there is aprior experience we will choose the cross-validation methodto estimate the parameter k
423 Localization The last step is localization based on theoutput of LOF calculation Here the neighbor cell list [25] isused to search the relation between the outage cell and itsneighbor cells by geographic informationwith 119911-score In thisway the outage cells may be localized
43 Results Analysis Based on the above steps M-LOFwould detect anomalous behaviors of handover And in thesimulation the value of 119896 for the LOF based detector is foundbetween 5 and 14
For the purpose of validation we will first analyze theresults from the spatial and temporal perspectives and thendetermine outage cell locations with neighbor cell list andperformance comparison between M-LOF and LOF will begiven at last
431 Spatial Analysis For spatial analysis we focus on thetemporal data of different cells at the same time periodFigure 7 shows the M-LOF of each cell at 95th TTI It can beseen that factor values derived from six cellsrsquo inHO data arefar higher than the normal value which is usually less than15 Therefore the abnormal cells can be distinguished fromnormal cells as they have experienced many user reconnec-tions
For comparison Figure 7 also shows the detection resultusingMDTmeasurements (with RSRP and SINR) as the datasource It can be seen that the abnormal cells are hard to bedetected using MDT Because M-LOF values of cells basedon MDT data are smoother than inHO data In contrastour proposed method using the inHO data has a betterperformance for abnormal cell detection in H-CRAN
432 Temporal Analysis Figure 8 shows the temporal analy-sis result for M-LOF variations relative to TTI for outage cell13 At this time from the results we find that M-LOF valuearises between TTI 90 and TTI 100 which is in accordancewith our outage time setting As initial connections of usersoccur between 0ndash5 TTI under unstable status so we justignore these data
433 Localization After the implementation of M-LOFbased detector neighbor cell list is used to localize the outage
Wireless Communications and Mobile Computing 9
TTI
Outage Period
22
20
18
16
14
12
10
08
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140
MLO
F Va
lue 5
1015
2025 30
35 40
4550
55
60
6570
75
80
85
90
95
100 105
110
115120
125
130
135140
Figure 8 Temporal analysis for outage TTI
cells First we need to set a threshold to filter anomalous cellsthrough 119911-score calculation as
119911119896 = 119899119896 minus 120583119899120590119899 (13)
where 119899119896 is the M-LOF value of kth cell and 120583119899 120590119899 arethe mean and standard deviation of M-LOF value for theanomaly scores of the other cells The reference 119911-scorethreshold is configured with a preferred value for the abnor-mal cells Here the reference 119911-score is set to 21 according tothe computing results We conducted two sets of evaluationsto make a comparison of performance of COD based oninHO data and MDT data respectively The results canidentify outage cell 18 and cell 61 which is the right one asshown in Figure 9 as LOF values of their neighbor cells arerelatively higher than other cells
434 Performance Evaluation In this part we analyzedthe M-LOF detection performance under varying trafficconditions Since the behaviors of users have a direct effect oninHO data the diagnosis process has been tested in differentscenarios by changing the User Density (UD) and UserVelocity (UV) parameters as the baseline setup To evaluatethe impact of the variations of UD and UV on M-LOFvalues different scenarios are set up by adjusting these twoparameters The Cumulative Distribution Function (CDF) oftheM-LOF values under different UV conditions is shown inFigure 10
It can be seen that for the low-velocity scenario almost80 of the M-LOF values are less than 05 However thereis a significant reduction in the M-LOF value as the UVincreases Likewise a similar behavior is observed with theincrease of UD as shown in Figure 11 The UV and UDparameters influence the distribution and spread of the inHOdata as explained earlier and consequently the value of M-LOF This leads to a low detection performance of M-LOFsince it generates an increased number of false alarms
Table 3 Performance evaluation
Result LOF value M-LOF valueFalse negative rate 3 3False positive rate 12 6
Outage cellAffected cell
800
600
400
200
Use
r Pos
ition
Y
User Position X
100 200 300 400 500 600 700 800
1
2
3
4
5
6
7
8
9
10
11
2122
25
27
29
30
32
33
64 63
62
60
1258 71
1773
16
68
36
35
14
6113
59
2456
57
74
1875
65
76
55
0
72
49
52
54
47
38 43
6615
67
70 69
41
44
46
Figure 9 Outage cell location results
At last this paper compares LOF and M-LOF by evaluat-ing False Positive Rate (FPR) and False Negative Rate (FNR)and the final results of the simulation are shown in Table 3The FNR represents the chance that an outage cell is notrecognized from the outage cells while the FPR representsthe chance that a normal cell is recognized as the outage cellfrom all normal cells
From Table 3 we can see that FPR and FNR of LOF baseddetector are 12 and 3 respectively It means the outagecell can be almost wholly detected although a small part ofthe normal cell may be determined as an anomaly The maincause of the 3FNR is that a small number of anomalous cellshave low traffic so that seldom user data cannot be collectedwhen they are in outage So these small cells are failed to bedetected However these outages have little impact on overallnetwork performance and user experience Therefore wecan say the outage cells can be detected successfully withLOF based detector approach As for M-LOF the FPR is 6which is smaller than LOFThe reason whyM-LOF has betterperformance is that M-LOF pays more attention to localdensity This is in line with the actual use of outage detectionsince the abnormal handovers caused by the neighbor outagecell are also localized
Still asM-LOF just modified the distance computation ofLOF so for time complexity it is just the same as LOF withis 119874(1198992) Here 119899 is the TTI number or cell number shown inFigure 7 or Figure 8
10 Wireless Communications and Mobile Computing
10
08
06
04
02
00minus2 minus1 0 1 2 3 4 5
0mM lt 5 6FI=CNS lt 1ms1mM lt 5 6FI=CNS lt 3ms3mM lt 5 6FI=CNS lt 5ms
M-LOF
CDF
Figure 10 CDF of M-LOF for User Velocity
10
08
06
04
02
00minus2 minus1 0 1 2 3 4 65
M-LOF
CDF
UE Density = 30 UsersCellUE Density = 20 UsersCellUE Density = 10 UsersCell
Figure 11 CDF of M-LOF value for User Density
5 Conclusions
In this paper we propose self-organized cell outage detectionarchitecture and approach for 5G H-CRAN The previousmethods using single data source are hard to detect thehidden outage such as RRHs Aiming at this problem weintroduce architecture and corresponding approach to realizeCOD for H-CRAN and still analyze the required key tech-nologies Then we evaluate the approach under a use casewhich utilizes the collected inHO statistics and adopt M-LOF to detect the anomalous behavior outage cells Resultsshow that the use case is effective However much work stillhas to be done for this architecture as we do not considerresource allocation and COC solutions still we have to takemore intelligent learning model such as deep learning into
account as well and compare the efficiency of different algo-rithms under a embodied H-CRAN network deploymentsMoreover we will focus on a specific perspective for self-healing approaches for H-CRAN and propose more practicalsolutions
Conflicts of Interest
The authors declare that they have no conflicts of interest
Acknowledgments
This research is supported by the National Science andTechnology Major Project of the Ministry of Science andTechnology of China (no 2018ZX030110004)
References
[1] P Rost A Banchs I Berberana et al ldquoMobile network archi-tecture evolution toward 5Grdquo IEEE Communications Magazinevol 54 no 5 pp 84ndash91 2016
[2] M Peng Y Sun X Li Z Mao and C Wang ldquoRecentadvances in cloud radio access networks system architectureskey techniques and open issuesrdquo IEEECommunications Surveysamp Tutorials vol 18 no 3 pp 2282ndash2308 2016
[3] M Peng Y Li Z Zhao and C Wang ldquoSystem architectureand key technologies for 5G heterogeneous cloud radio accessnetworksrdquo IEEE Network vol 29 no 2 pp 6ndash14 2015
[4] N Chen B Rong X Zhang and M Kadoch ldquoScalable andflexible massive MIMO precoding for 5G H-CRANrdquo IEEEWireless Communications Magazine vol 24 no 1 pp 46ndash522017
[5] OGAliu A ImranMA Imran andB Evans ldquoA survey of selforganisation in future cellular networksrdquo IEEECommunicationsSurveys amp Tutorials vol 15 no 1 pp 336ndash361 2013
[6] E J Khatib R Barco P Munoz I D La Bandera and ISerrano ldquoSelf-healing in mobile networks with big datardquo IEEECommunications Magazine vol 54 no 1 pp 114ndash120 2016
[7] M Amirijoo L Jorguseski T Kurner et al ldquoCell outagemanagement in LTE networksrdquo in Proceedings of the 2009 6thInternational Symposium on Wireless Communication Systems(ISWCS 2009) pp 600ndash604 Siena Italy September 2009
[8] M YinW Li L Feng P Yu and X Qiu ldquoMulti-cell cooperativeoutage compensation in cloud-RANs based 5G public safetynetworkrdquo IEEE Access vol 5 pp 17309ndash17321 2017
[9] A Zoha A Saeed A Imran M A Imran and A Abu-DayyaldquoData-driven analytics for automated cell outage detection inSelf-Organizing Networksrdquo in Proceedings of the 2015 11th Inter-national Conference on the Design of Reliable CommunicationNetworks DRCN 2015 pp 203ndash210 usa March 2015
[10] I De-La-Bandera R Barco P Munoz and I Serrano ldquoCellOutage Detection Based onHandover Statisticsrdquo IEEE Commu-nications Letters vol 19 no 7 pp 1189ndash1192 2015
[11] W Wang Q Liao and Q Zhang ldquoCOD A cooperativecell outage detection architecture for self-organizing femtocellnetworksrdquo IEEE Transactions on Wireless Communications vol13 no 11 pp 6007ndash6014 2014
[12] Q Liao M Wiczanowski and S Stanczak ldquoToward cell outagedetection with composite hypothesis testingrdquo in Proceedingsof the 2012 IEEE International Conference on CommunicationsICC 2012 pp 4883ndash4887 can June 2012
Wireless Communications and Mobile Computing 11
[13] Y Ma M Peng W Xue and X Ji ldquoA dynamic affinity propa-gation clustering algorithm for cell outage detection in self-healing networksrdquo in Proceedings of the 2013 IEEE WirelessCommunications and Networking Conference WCNC 2013 pp2266ndash2270 chn April 2013
[14] M Alias N Saxena and A Roy ldquoEfficient Cell Outage De-tection in 5G HetNets Using Hidden Markov Modelrdquo IEEECommunications Letters vol 20 no 3 pp 562ndash565 2016
[15] W Xue M Peng Y Ma and H Zhang ldquoClassification-basedapproach for cell outage detection in self-healing heterogeneousnetworksrdquo in Proceedings of the 2014 IEEEWireless Communica-tions and Networking Conference WCNC 2014 pp 2822ndash2826tur April 2014
[16] J Wang N Q Phan Z Pan N Liu X You and T Deng ldquoAnimproved TCM-based approach for cell outage detection forself-healing in LTE HetNetsrdquo in Proceedings of the 83rd IEEEVehicular Technology Conference VTC Spring 2016 chn May2016
[17] T Zhang L Feng P Yu S Guo W Li and X Qiu ldquoA handoverstatistics based approach for Cell Outage Detection in self-organized Heterogeneous Networksrdquo in Proceedings of the 15thIFIPIEEE International Symposium on Integrated Network andService Management IM 2017 pp 628ndash631 prt May 2017
[18] D Lee S Zhou X Zhong Z Niu X Zhou and H ZhangldquoSpatial modeling of the traffic density in cellular networksrdquoIEEE Wireless Communications Magazine vol 21 no 1 pp 80ndash88 2014
[19] F Xu Y Li H Wang P Zhang and D Jin ldquoUnderstandingMobile Traffic Patterns of Large Scale Cellular Towers in UrbanEnvironmentrdquo IEEEACM Transactions on Networking vol 25no 2 pp 1147ndash1161 2017
[20] S Morosi P Piunti and E Del Re ldquoSleep mode managementin cellular networks a traffic based technique enabling energysavingrdquo Transactions on Emerging Telecommunications Tech-nologies vol 24 no 3 pp 331ndash341 2013
[21] W Y Szeto B Ghosh B Basu andM OrsquoMahony ldquoMultivariatetraffic forecasting technique using cell transmission model andSARIMA modelrdquo Journal of Transportation Engineering vol135 no 9 pp 658ndash667 2009
[22] P V KlaineM A Imran O Onireti and R D Souza ldquoA Surveyof Machine Learning Techniques Applied to Self OrganizingCellular Networksrdquo IEEE Communications Surveys amp Tutorials2017
[23] Z M Fadlulla F Tang B Mao and et al ldquoState-of-the-Art Deep Learning Evolving Machine Intelligence TowardTomorrowrsquos Intelligent Network Traffic Control Systemsrdquo IEEECommunications Surveys amp Tutorials vol 19 no 4 pp 2432ndash2455 2017
[24] M M Breuniq H Kriegel R T Ng and J Sander ldquoLOFidentifying density-based local outliersrdquo SIGMOD Record vol29 no 2 pp 93ndash104 2000
[25] C M Mueller M Kaschub C Blankenhorn and S WankeldquoA Cell Outage Detection Algorithm Using Neighbor Cell ListReportsrdquo in Self-Organizing Systems vol 5343 of Lecture Notesin Computer Science pp 218ndash229 Springer Berlin HeidelbergBerlin Heidelberg 2008
International Journal of
AerospaceEngineeringHindawiwwwhindawicom Volume 2018
RoboticsJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Active and Passive Electronic Components
VLSI Design
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Shock and Vibration
Hindawiwwwhindawicom Volume 2018
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawiwwwhindawicom
Volume 2018
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Control Scienceand Engineering
Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom
Journal ofEngineeringVolume 2018
SensorsJournal of
Hindawiwwwhindawicom Volume 2018
International Journal of
RotatingMachinery
Hindawiwwwhindawicom Volume 2018
Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Navigation and Observation
International Journal of
Hindawi
wwwhindawicom Volume 2018
Advances in
Multimedia
Submit your manuscripts atwwwhindawicom
Wireless Communications and Mobile Computing 3
diagnosis fault classification and COM COM first needs todetect the cell outage automatically and then perform a rea-sonable compensation mechanism to repair the faults so asto minimize the impact of cell outage [7] Therefore COD isan essential prerequisite for self-healing
At present the studies of COD mainly aim at tradi-tional LTELTE-AHetNet with limited data collection whichmay not be appropriate for the complex H-CRAN networkarchitecture This paper proposes a self-organized CODarchitecture and a corresponding approach for H-CRANbased on previous work and gives a use case with HO dataanalysis to evaluate them
The rest of the paper is organized as follows Section 2introduces the related work for COD Section 3 proposes thearchitecture and approach for 5GH-CRANCODmeanwhileintroducing COD procedures and related technologies InSection 4 a CODuse case forHOdata analysis withmodifiedM-LOF is introduced and conclusions are given in Section 5
2 Related Work
In wireless communication networks cell outage is mainlycaused by the software and hardware faults which bring innetwork communication interrupt thus affecting networkQoS and usersrsquo QoE [8] At present the existing CODresearches mainly focus on the LTELTE-A HetNet Severaldetection methods use data collected by drive test or sub-scriber complaints to analyze the network faults and celloutages For instance in [9] COD is analyzed autonomouslyby preprocessing the minimization drive testing (MDT)together with local outlier factor based detector (LOFD) andone class support vectormachine based detector (OCSVMD)to detect and localize anomalous network behavior Thesesolutions not only cost much time and manpower cost butalso require expert knowledge or prior experience
Several studies pay attention to COD with KPI variationsuch as handover statistics [10] and a cooperative femtocelloutage detection architecture which consists of a trigger stageand a detection stagewith RSRP is introduced in [11] Furtheran efficient discriminant function is used to complete CODwith CQI and RRC connection reestablishment informationin [12] However these approaches are just suitable fortraditional UMTS or LTELTE-A networks
Recently there have been a few studies that focus onCODwithmachine learning approaches under data collectedfrom users or cells In [13] an unsupervised data miningalgorithm with a reference signal received power (RSRP) andreference signal received quality (RSRQ) was proposed todetect cell outages Reference [14] applied theHiddenMarkovmodel for cell outage detection under RSRP and RSRQas well Further a classification-based approach named 119870-nearest neighbor (KNN) is proposed for COD in [15] andtransductive confidence machines (TCM) based COD withRSRP and SINR data is proposed in [16] Moreover ourprevious work used LOF to detect cell outage with handoverstatistics [17] These works can give suggestions for technol-ogy selection for 5G H-CRAN COD
In particular RRHs failure in 5G H-CRAN may be diffi-cult to detect because their failure may not trigger operation
andmaintenance system alarm which make that self-healingfunction unable to timely compensate for these nodes As aresult the outage may last hours or a few days before beingdiscovered unless the abnormal status is captured by DT testor feedback from users
Rapid fault discovery and localization of abnormal cellscan reduce network paralysis and deterioration caused bynode outage Due to the high density of nodes in H-CRANthe faults of a single RRH will not quickly affect the networkas users may handover or reconnect to neighbor RRHs orACEs It increases the fault tolerance rate but makes CODmore difficult as well thus increasing the instability of thenetwork
Based on above analysis we find that present CODapproaches for LTELTE-A HetNet may not be suitablefor H-CRAN due to their limited data sets and obviousoutage alarmTherefore in this paper we will establish CODarchitecture with complete procedures to handle implicitRRH outages
3 COD Architecture andApproach for 5G H-CRAN
To make it clean in 5G H-RAN we also regard the coverageof each RRH as a cell Still cell definition for ACEs may usetraditional ones As COD for ACEs can be easily resolvedby approach mentioned above in HetNet here we mainlyconcentrate on implicit cell outage for RRHs
For the convenience of illustration we consider a simplecell outage scenario in H-CRAN under a NodeC as shownin Figure 2 Here one ACE and many randomly distributedRRHs are under control of one NodeC When a RRHcell turns into outage its serving users may reconnect toACE or handover to another RRH cell This reconnectionor handover will result in signaling and communicationvariation among users RRHs ACEs and NodeC We requireall these data to execute COD if we want to achieve accurateand timely outage detection
In the above scenario a simple user reconnect procedureis shown in Figure 3 If a user has to establish a com-munication link with a RRH it first sends a ldquoConnectionRequestrdquo command to it and then this RRH will send aldquoResource Requestrdquo command to NodeC Only when NodeCreply a ldquoRequest Responserdquo command to RRH with requiredresource allocation the RRH will give a successful connectreply with ldquoConnection Setuprdquo to users Then the service willgo on It means a user keeps a connection to a RRH underthe control of NodeC with allocated resource If the RRHhas gone into outage state due to cable loss or power offuserrsquos service will fail and it will attempt to reconnect to ACEor other RRHs If NodeC has enough resource connectionbetween user andACEwill be established again as the same asRRH connection Here RRHoutagemay not report toNodeCimmediately However we can count the time of connectrequest fromuser to other RRHandACE as the two countingpoints show in Figure 3 If we store these data and analyze itwith time series fitting or prediction we may get variationfeatures and obtain abnormal points at several time intervalswhich may be useful for RRH COD
4 Wireless Communications and Mobile Computing
RRH
ACE
RRH
Outage RRH
Cloud
Node C
ConnectLost connectionReconnect
RRH FronthaulACE Fronthaul
Figure 2 Cell outage in H-CRAN
Connection requestResource request
Resource responseConnection response
Connection setup
In Service
Outage OccursConnection failure
Connection requestResource request
Resource responseConnection response
Connection setup
Service Recovery
Counting point
Counting point
User RRH ACE NodeC
Figure 3 An instance of user reconnection process
Wireless Communications and Mobile Computing 5
Data Collection
User
RRHACE
NodeC
DT
Data Analysis
Pre-processing
Machine Learning
Temporal Predication
Spatial Predication
COD ConclusionCell Outage Compensation
Input
Input
Input
Input
Guideline
AchievePerformance Evaluation
MDT Data
Network topologyCell parameters
HOConnection RequestNCL
RSRPRSRQSINRCQI
Figure 4 Overview architecture for COD in 5G H-CRAN
Based on the above analysis we want to construct an inte-grated COD architecture for 5G H-CRAN and its detailedintroduction is given below
31 COD Architecture and Procedures Firstly the proposedarchitecture for COD can be found in Figure 4
As shown in Figure 4 COD mainly consists of datacollection and data analysis stages Firstly data collectionshould store data from different sources which are as follows
(1) Data collection from users through measurementreports such as RSRP RSRQ SINR and CQI infor-mation this information is always huge and hard tobe handled synchronously as the time interval is fairlydense So we can set a sampling interval and just takethe statistics for them
(2) Data collection from RRHACE such as HOCon-nection RequestNCL these data may come from theKPI statistics from OAM system These data are per-formance indicators and can be used directly as theinterval is defined beforehand under acceptable level
(3) Data collection for NodeC which may network leveldata such as network topology and preconfiguredcell parameters (transmit power spectrum antennaheight and tilt) these data can be used to supplementthe spatial and temporal analysis of different cells
(4) Drive test data drive test is a tool to verify networkperformance afterward It can be used as a validationfor COD conclusions and thus provide correctionsuggestions for COD approach
After data collection all the data can be put to dataanalysis stage to execute COD As shown in Figure 4 withSON entities which located at NodeC the procedure will beexecuted as follows
(1) SON entities firstly should preprocess these data toimprove the data quality
(2) SON entities adopt temporal and spatial predictionmethod to obtain varying patterns for temporal dataand spatial data respectively
(3) Next SON entities choose proper machine learningapproach to identify the outage cell
Still with COD results we can give suggestions for COCAnd COC effectiveness evaluation can be obtained throughnetwork performance monitoring thus constructing a self-organized loop The critical technologies of preprocessingspatial and temporal prediction andmachine learning will beintroduced next
32 Key Technologies for COD
321 Data Processing Data processing includes two conceptssuch as Data Cleaning and Feature Engineering These twoare compulsory for achieving better accuracy and perfor-mance before machine learning and deep learning And itincludes data cleaning data integration data transformationand data reduction So in our architecture we should chooseproper data processing approach to obtain high-quality data
322 Spatial Prediction Spatial prediction technologiesmainly aim at analysis spatial traffic distributions or cap-ture user variations Currently several methods have beenadopted in this field such as log-normal or Weibull distri-bution used in [18] or traffic patterns identifying methodsproposed in [19] With spatial prediction we can obtaindifferent distribution laws
323 Temporal Prediction Temporal prediction is aimingat predicting future variation direction based on past and
6 Wireless Communications and Mobile Computing
Table 1 Cell level data
Measurements DescriptionTime Transmission time interval (TTI)Cell info Cell global identity (CGI)Cell location Longitude and latitude informationCell inHO Cell incoming handover
current temporal sequence data for instance traffic variationof one ACE or load of NodeC Temporal prediction canadopt HoltndashWinter traffic forecast [20] or S-ARIMA model[21] to construct prediction sequences which will be usefulfor abnormal traffic recognition Moreover joint temporal-spatial predication approaches deservemore attention aswell
324 Machine Learning As analyzed in [22] machine learn-ing has been used for LTELTE-A self-healing and achievesremarkable results However thesemethods aremainly basedon single data source and do not take integrated data setshown in Section 31 together To make COD more accurateand intelligent in 5Gnetworks deep learningmodel [23] suchas DRN model can be considered
4 A Use Case for 5G H-CRAN COD
According to the architecture and procedures above weintroduce a use case to achieve COD for 5G H-CRAN Herewe suppose that the outage RRH or ACE cell cannot carrytraffic anymore Typically at this point all users served bythis cell have to be offloaded to its neighbor cells This willresult in a huge number of handovers in H-CRANThe risinghandovers in neighbor cells could be easily detected bymeansof data mining So we will discuss our algorithm by using thetemporal data of inHO statistics collected from the NodeC
41 Use Case Scenario Description The scenario is shown inFigure 5 It consists of 19 ACE with 25 users under coverageAll these users are active users Each ACE has 3 cells andone RRH cell is located under each ACE cell The UEs arerandomly distributed in cells to generate handover statisticsand measurement reports
Still we mainly collect cell level information stored inNodeC as shown in Table 1
Moreover in order to make the evaluation more obviouswe will evaluate the procedure under a simulation platformdeveloped by our lab Parameter settings can be found inTable 2
In our simulation platformwe assume that rapid decreaseof cellrsquos transmit power indicates cell outage That is we canset cell outage manually to evaluate efficiency of our ap-proach Then users under the affected outage areas switch tothe neighbor cells which will increase the number of han-dovers We will use the inHO data collected from all the cellswith a machine learning methods to evaluate whether outagecells can be detected
Table 2 Simulation parameters
Simulation parameters Value (unit)ACE 19 sites with 3 cells eachRRH 1 cell per siteUser number 25 per ACEUser distribution Uniform random distributionLink direction DownlinkBS Tx power 46 dBmPath loss model Cost231-HataUE velocity 15 kmhUser distribution UniformCell selection criteria Strongest RSRP defines the target cellSimulation length 150 TTI
RRH cellACE
Figure 5 Scenario use case for 5G H-CRAN
42 COD Procedures with M-LOF Considering the situa-tions above and the advantage of LOF shown in [9] a mod-ified LOF based detector is adopted to detect the outage byextracting both spatial and temporal correlations of handoverstatistics over the neighbors of outage cells The procedure isshown in Figure 6
The COD procedure includes four steps data collectionpreprocessingmachine learning and localization and all canbe executed automatically under the control of LS-SON Inthis way the outage cell can be detected timely so we canminimize the performance degradation of the network Herewe will introduce the functions in three steps
421 Preprocessing When data collection is finished SONentities proceed to extract feature vector 119879119894 for LOF [24]
Wireless Communications and Mobile Computing 7
Cell Level Data
Filter Standardization
Database
Low-dimension
Pre-processing
Data Collection
Machine Learning
Detection
Test z scoreCalculation
Store data to database
Localization
ParameterEstimation
No
M-LOFCalculation
OutageYes
H-CRAN
COC
thresholdZ gt
Figure 6 COD procedure with modified LOF
corresponding to each temporal data Since the inHO data iscollected from different cells at the different time the datavector is designed as spatial-temporal series
In fact the inHO statistics of outage cells are zero sincethe KPIs in outage cell cannot be collected Moreover for onepossibility there may be a few zero-inHO cells because noactive users are within their coverage This will enhance thedifficulty of detection In this case we filter all inHO statisticsof zero Here each cell splits out an ordered time series ofinHO To finish length-wise normalization and carry outuniformization and standardization processing we slice thetime series to several subseries
For temporal analysis 5 TTIs data is put into one featurevector 119879119894 as shown in below where 119894 represents the order ofsubseries
119879119894 = inHO1 inHO2 inHO3 inHO4 inHO5 (1)
The feature vector 119879119894 utilizes exactly one cellrsquos inHO datawhich is used to find out the time when an outage happened
For spatial analysis one feature vector 119878119895 consists of 119897cell handover statistics at TTI time 119895 where 119895 represents theidentity of the cell
119878119895 = inHO1 inHO2 inHO3 inHO4 inHO5 inHO6inHO7 inHO8 inHOl
(2)
After slicing the data series linear functional transforma-tion criterion is used to normalize the data to eliminate errorscaused by nonuniform features For example the criterion of119878119895 is defined as
1198781015840119895 =119878119895 minusmin119895 (119878119895)
max119895 (119878119895) minusmin119895 (119878119895) (3)
422 Machine Learning LOF detection is an unsupervisedanomaly detection algorithm The local densities of targetpoints are calculated by LOF used to compare with itsneighbors The larger the difference between the sampleand its neighbors is the larger outage factor score will beassigned to the sampleThe advantage of LOF is that no priorknowledge is required in advance to detect the unknowncell outage The detailed definitions of LOF are explained asfollows
Definition 1 k-distance of cell 119901 denoted as 119889119896(119901) representsthe distance between cell 119901 and its 119896th nearest neighbor sothat for at least 119896 cell 119900119894 satisfies
119889 (119900119894 119901) le 119889 (119900119896 119901) (4)
and for at most 119896 minus 1 cells 119900119894 satisfies119889 (119900119894 119901) lt 119889 (119900119896 119901) (5)
Definition 2 119896-distance neighborhood of cell 119901 denoted as119873119889119896(119901)(119901) when 119889119896(119901)is given 119873119889119896(119901)(119901) is the set whichincludes every cell whose distance to cell 119901 is smaller than119889119896(119901)Definition 3 Reachability distance of cell 119901 with respect tocell 119900 is denoted as rd119896(119901 119900) it is the maximum one of 119889119896(119901)and the distance between 119901 and 119900
rd119896 (119901 119900) = max 119889119896 (119900) 119889 (119900 119901) (6)
Definition 4 Local reachability density of cell p denoted aslrd119896(119901) is shown as follows
lrd119896 (119901) =10038161003816100381610038161003816119873119889119896(119901) (119901)
10038161003816100381610038161003816sum119900isin119873119889119896(119901)(119901) rd119896 (119901 o)
(7)
Definition 5 Local outlier factor of cell p denoted asLOF119896(119901) can be obtained as below
LOF119896 (119901) =sum119900isin119873119889119896(119901)(119901) (lrd119896 (119900) lrd119896 (119901))10038161003816100381610038161003816119873119889119896(119901) (119901)
10038161003816100381610038161003816 (8)
The above descriptions give the definition of LOF Fromour previous work in [17] the proposed M-LOF is a properapproach for heterogeneous network so here we use themodified LOF algorithm called M-LOF to make a refinedchoice for H-CRAN Definitions of M-LOF and m-distanceare given below
8 Wireless Communications and Mobile Computing
0
0
InHo DataMDT DataAbnormal cells
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80
Cell Id
minus15minus10minus05000510152025303540455055
MLO
F Va
lue
1
23
4
5
6
7
8
9
10
11
12
13
14
1516
17
2122
24
25 27
29
30
3233
35
36
38 4143
44
4647
49
52
54
55
56
57
58
59
6062
63
64
65
66
67
69
68
7071
72
73
74
75
76
1
2
3
4
5
6
7
89
10
11
1213
14
15
1617
21
22 2425 27 29
30 32
33
35 36 38
41
43
44
46
47
49
52
54
55
56575859
6062
63
6465
66
6768
69
70717273
74
75
76
Figure 7 Spatial analysis for abnormal cells
Definition 6 119898-distance of cell 119901 given the positive integer119896 is defined as
119898minus119889119896 (119901) = 120576 + [[sum119900isin119873119889119896(119901)(119901) 119889 (119901 119900)10038161003816100381610038161003816119873119889119896(119901) (119901)
10038161003816100381610038161003816]] (9)
Here 120576 is a constant value to enhance the accuracyDefinition 7 119898-distance neighborhood of cell p denotedas 119873119898minus119889119896(119901)(119901) is the set which includes every cell whichdistance to cell 119901 is smaller thanm-distance
Definition 8 Reachability distance of m-distance of cell 119901concerning cell o denoted as 119903minus119889119898(119901 119900) is
119903minus119889119898 (119901 119900) = max 119898minus119889119896 (119900) 119889 (119900 119901) (10)
Definition 9 Local reachability density of cell p is defined as
lrd119898 (119901) =10038161003816100381610038161003816119873119898minus119889119896(119901) (119901)
10038161003816100381610038161003816sum119900isin119873119898minus119889119896(119901)(119901) 119903minus119889119898 (119901 o)
(11)
Definition 10 Modified local outlier factor of cell p is shownbelow
LOF119898 (119901) =sum119900isin119873119898minus119889119896(119901)(119901) (lrd119898 (119900) lrd119898 (119901))10038161003816100381610038161003816119873119898minus119889119896(119901) (119901)
10038161003816100381610038161003816 (12)
According to the definition of LOF119896(119901) and LOF119898(119901) thechoice of 119896 is exactly sensitive As a consequence if there is aprior experience we will choose the cross-validation methodto estimate the parameter k
423 Localization The last step is localization based on theoutput of LOF calculation Here the neighbor cell list [25] isused to search the relation between the outage cell and itsneighbor cells by geographic informationwith 119911-score In thisway the outage cells may be localized
43 Results Analysis Based on the above steps M-LOFwould detect anomalous behaviors of handover And in thesimulation the value of 119896 for the LOF based detector is foundbetween 5 and 14
For the purpose of validation we will first analyze theresults from the spatial and temporal perspectives and thendetermine outage cell locations with neighbor cell list andperformance comparison between M-LOF and LOF will begiven at last
431 Spatial Analysis For spatial analysis we focus on thetemporal data of different cells at the same time periodFigure 7 shows the M-LOF of each cell at 95th TTI It can beseen that factor values derived from six cellsrsquo inHO data arefar higher than the normal value which is usually less than15 Therefore the abnormal cells can be distinguished fromnormal cells as they have experienced many user reconnec-tions
For comparison Figure 7 also shows the detection resultusingMDTmeasurements (with RSRP and SINR) as the datasource It can be seen that the abnormal cells are hard to bedetected using MDT Because M-LOF values of cells basedon MDT data are smoother than inHO data In contrastour proposed method using the inHO data has a betterperformance for abnormal cell detection in H-CRAN
432 Temporal Analysis Figure 8 shows the temporal analy-sis result for M-LOF variations relative to TTI for outage cell13 At this time from the results we find that M-LOF valuearises between TTI 90 and TTI 100 which is in accordancewith our outage time setting As initial connections of usersoccur between 0ndash5 TTI under unstable status so we justignore these data
433 Localization After the implementation of M-LOFbased detector neighbor cell list is used to localize the outage
Wireless Communications and Mobile Computing 9
TTI
Outage Period
22
20
18
16
14
12
10
08
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140
MLO
F Va
lue 5
1015
2025 30
35 40
4550
55
60
6570
75
80
85
90
95
100 105
110
115120
125
130
135140
Figure 8 Temporal analysis for outage TTI
cells First we need to set a threshold to filter anomalous cellsthrough 119911-score calculation as
119911119896 = 119899119896 minus 120583119899120590119899 (13)
where 119899119896 is the M-LOF value of kth cell and 120583119899 120590119899 arethe mean and standard deviation of M-LOF value for theanomaly scores of the other cells The reference 119911-scorethreshold is configured with a preferred value for the abnor-mal cells Here the reference 119911-score is set to 21 according tothe computing results We conducted two sets of evaluationsto make a comparison of performance of COD based oninHO data and MDT data respectively The results canidentify outage cell 18 and cell 61 which is the right one asshown in Figure 9 as LOF values of their neighbor cells arerelatively higher than other cells
434 Performance Evaluation In this part we analyzedthe M-LOF detection performance under varying trafficconditions Since the behaviors of users have a direct effect oninHO data the diagnosis process has been tested in differentscenarios by changing the User Density (UD) and UserVelocity (UV) parameters as the baseline setup To evaluatethe impact of the variations of UD and UV on M-LOFvalues different scenarios are set up by adjusting these twoparameters The Cumulative Distribution Function (CDF) oftheM-LOF values under different UV conditions is shown inFigure 10
It can be seen that for the low-velocity scenario almost80 of the M-LOF values are less than 05 However thereis a significant reduction in the M-LOF value as the UVincreases Likewise a similar behavior is observed with theincrease of UD as shown in Figure 11 The UV and UDparameters influence the distribution and spread of the inHOdata as explained earlier and consequently the value of M-LOF This leads to a low detection performance of M-LOFsince it generates an increased number of false alarms
Table 3 Performance evaluation
Result LOF value M-LOF valueFalse negative rate 3 3False positive rate 12 6
Outage cellAffected cell
800
600
400
200
Use
r Pos
ition
Y
User Position X
100 200 300 400 500 600 700 800
1
2
3
4
5
6
7
8
9
10
11
2122
25
27
29
30
32
33
64 63
62
60
1258 71
1773
16
68
36
35
14
6113
59
2456
57
74
1875
65
76
55
0
72
49
52
54
47
38 43
6615
67
70 69
41
44
46
Figure 9 Outage cell location results
At last this paper compares LOF and M-LOF by evaluat-ing False Positive Rate (FPR) and False Negative Rate (FNR)and the final results of the simulation are shown in Table 3The FNR represents the chance that an outage cell is notrecognized from the outage cells while the FPR representsthe chance that a normal cell is recognized as the outage cellfrom all normal cells
From Table 3 we can see that FPR and FNR of LOF baseddetector are 12 and 3 respectively It means the outagecell can be almost wholly detected although a small part ofthe normal cell may be determined as an anomaly The maincause of the 3FNR is that a small number of anomalous cellshave low traffic so that seldom user data cannot be collectedwhen they are in outage So these small cells are failed to bedetected However these outages have little impact on overallnetwork performance and user experience Therefore wecan say the outage cells can be detected successfully withLOF based detector approach As for M-LOF the FPR is 6which is smaller than LOFThe reason whyM-LOF has betterperformance is that M-LOF pays more attention to localdensity This is in line with the actual use of outage detectionsince the abnormal handovers caused by the neighbor outagecell are also localized
Still asM-LOF just modified the distance computation ofLOF so for time complexity it is just the same as LOF withis 119874(1198992) Here 119899 is the TTI number or cell number shown inFigure 7 or Figure 8
10 Wireless Communications and Mobile Computing
10
08
06
04
02
00minus2 minus1 0 1 2 3 4 5
0mM lt 5 6FI=CNS lt 1ms1mM lt 5 6FI=CNS lt 3ms3mM lt 5 6FI=CNS lt 5ms
M-LOF
CDF
Figure 10 CDF of M-LOF for User Velocity
10
08
06
04
02
00minus2 minus1 0 1 2 3 4 65
M-LOF
CDF
UE Density = 30 UsersCellUE Density = 20 UsersCellUE Density = 10 UsersCell
Figure 11 CDF of M-LOF value for User Density
5 Conclusions
In this paper we propose self-organized cell outage detectionarchitecture and approach for 5G H-CRAN The previousmethods using single data source are hard to detect thehidden outage such as RRHs Aiming at this problem weintroduce architecture and corresponding approach to realizeCOD for H-CRAN and still analyze the required key tech-nologies Then we evaluate the approach under a use casewhich utilizes the collected inHO statistics and adopt M-LOF to detect the anomalous behavior outage cells Resultsshow that the use case is effective However much work stillhas to be done for this architecture as we do not considerresource allocation and COC solutions still we have to takemore intelligent learning model such as deep learning into
account as well and compare the efficiency of different algo-rithms under a embodied H-CRAN network deploymentsMoreover we will focus on a specific perspective for self-healing approaches for H-CRAN and propose more practicalsolutions
Conflicts of Interest
The authors declare that they have no conflicts of interest
Acknowledgments
This research is supported by the National Science andTechnology Major Project of the Ministry of Science andTechnology of China (no 2018ZX030110004)
References
[1] P Rost A Banchs I Berberana et al ldquoMobile network archi-tecture evolution toward 5Grdquo IEEE Communications Magazinevol 54 no 5 pp 84ndash91 2016
[2] M Peng Y Sun X Li Z Mao and C Wang ldquoRecentadvances in cloud radio access networks system architectureskey techniques and open issuesrdquo IEEECommunications Surveysamp Tutorials vol 18 no 3 pp 2282ndash2308 2016
[3] M Peng Y Li Z Zhao and C Wang ldquoSystem architectureand key technologies for 5G heterogeneous cloud radio accessnetworksrdquo IEEE Network vol 29 no 2 pp 6ndash14 2015
[4] N Chen B Rong X Zhang and M Kadoch ldquoScalable andflexible massive MIMO precoding for 5G H-CRANrdquo IEEEWireless Communications Magazine vol 24 no 1 pp 46ndash522017
[5] OGAliu A ImranMA Imran andB Evans ldquoA survey of selforganisation in future cellular networksrdquo IEEECommunicationsSurveys amp Tutorials vol 15 no 1 pp 336ndash361 2013
[6] E J Khatib R Barco P Munoz I D La Bandera and ISerrano ldquoSelf-healing in mobile networks with big datardquo IEEECommunications Magazine vol 54 no 1 pp 114ndash120 2016
[7] M Amirijoo L Jorguseski T Kurner et al ldquoCell outagemanagement in LTE networksrdquo in Proceedings of the 2009 6thInternational Symposium on Wireless Communication Systems(ISWCS 2009) pp 600ndash604 Siena Italy September 2009
[8] M YinW Li L Feng P Yu and X Qiu ldquoMulti-cell cooperativeoutage compensation in cloud-RANs based 5G public safetynetworkrdquo IEEE Access vol 5 pp 17309ndash17321 2017
[9] A Zoha A Saeed A Imran M A Imran and A Abu-DayyaldquoData-driven analytics for automated cell outage detection inSelf-Organizing Networksrdquo in Proceedings of the 2015 11th Inter-national Conference on the Design of Reliable CommunicationNetworks DRCN 2015 pp 203ndash210 usa March 2015
[10] I De-La-Bandera R Barco P Munoz and I Serrano ldquoCellOutage Detection Based onHandover Statisticsrdquo IEEE Commu-nications Letters vol 19 no 7 pp 1189ndash1192 2015
[11] W Wang Q Liao and Q Zhang ldquoCOD A cooperativecell outage detection architecture for self-organizing femtocellnetworksrdquo IEEE Transactions on Wireless Communications vol13 no 11 pp 6007ndash6014 2014
[12] Q Liao M Wiczanowski and S Stanczak ldquoToward cell outagedetection with composite hypothesis testingrdquo in Proceedingsof the 2012 IEEE International Conference on CommunicationsICC 2012 pp 4883ndash4887 can June 2012
Wireless Communications and Mobile Computing 11
[13] Y Ma M Peng W Xue and X Ji ldquoA dynamic affinity propa-gation clustering algorithm for cell outage detection in self-healing networksrdquo in Proceedings of the 2013 IEEE WirelessCommunications and Networking Conference WCNC 2013 pp2266ndash2270 chn April 2013
[14] M Alias N Saxena and A Roy ldquoEfficient Cell Outage De-tection in 5G HetNets Using Hidden Markov Modelrdquo IEEECommunications Letters vol 20 no 3 pp 562ndash565 2016
[15] W Xue M Peng Y Ma and H Zhang ldquoClassification-basedapproach for cell outage detection in self-healing heterogeneousnetworksrdquo in Proceedings of the 2014 IEEEWireless Communica-tions and Networking Conference WCNC 2014 pp 2822ndash2826tur April 2014
[16] J Wang N Q Phan Z Pan N Liu X You and T Deng ldquoAnimproved TCM-based approach for cell outage detection forself-healing in LTE HetNetsrdquo in Proceedings of the 83rd IEEEVehicular Technology Conference VTC Spring 2016 chn May2016
[17] T Zhang L Feng P Yu S Guo W Li and X Qiu ldquoA handoverstatistics based approach for Cell Outage Detection in self-organized Heterogeneous Networksrdquo in Proceedings of the 15thIFIPIEEE International Symposium on Integrated Network andService Management IM 2017 pp 628ndash631 prt May 2017
[18] D Lee S Zhou X Zhong Z Niu X Zhou and H ZhangldquoSpatial modeling of the traffic density in cellular networksrdquoIEEE Wireless Communications Magazine vol 21 no 1 pp 80ndash88 2014
[19] F Xu Y Li H Wang P Zhang and D Jin ldquoUnderstandingMobile Traffic Patterns of Large Scale Cellular Towers in UrbanEnvironmentrdquo IEEEACM Transactions on Networking vol 25no 2 pp 1147ndash1161 2017
[20] S Morosi P Piunti and E Del Re ldquoSleep mode managementin cellular networks a traffic based technique enabling energysavingrdquo Transactions on Emerging Telecommunications Tech-nologies vol 24 no 3 pp 331ndash341 2013
[21] W Y Szeto B Ghosh B Basu andM OrsquoMahony ldquoMultivariatetraffic forecasting technique using cell transmission model andSARIMA modelrdquo Journal of Transportation Engineering vol135 no 9 pp 658ndash667 2009
[22] P V KlaineM A Imran O Onireti and R D Souza ldquoA Surveyof Machine Learning Techniques Applied to Self OrganizingCellular Networksrdquo IEEE Communications Surveys amp Tutorials2017
[23] Z M Fadlulla F Tang B Mao and et al ldquoState-of-the-Art Deep Learning Evolving Machine Intelligence TowardTomorrowrsquos Intelligent Network Traffic Control Systemsrdquo IEEECommunications Surveys amp Tutorials vol 19 no 4 pp 2432ndash2455 2017
[24] M M Breuniq H Kriegel R T Ng and J Sander ldquoLOFidentifying density-based local outliersrdquo SIGMOD Record vol29 no 2 pp 93ndash104 2000
[25] C M Mueller M Kaschub C Blankenhorn and S WankeldquoA Cell Outage Detection Algorithm Using Neighbor Cell ListReportsrdquo in Self-Organizing Systems vol 5343 of Lecture Notesin Computer Science pp 218ndash229 Springer Berlin HeidelbergBerlin Heidelberg 2008
International Journal of
AerospaceEngineeringHindawiwwwhindawicom Volume 2018
RoboticsJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Active and Passive Electronic Components
VLSI Design
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Shock and Vibration
Hindawiwwwhindawicom Volume 2018
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawiwwwhindawicom
Volume 2018
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Control Scienceand Engineering
Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom
Journal ofEngineeringVolume 2018
SensorsJournal of
Hindawiwwwhindawicom Volume 2018
International Journal of
RotatingMachinery
Hindawiwwwhindawicom Volume 2018
Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Navigation and Observation
International Journal of
Hindawi
wwwhindawicom Volume 2018
Advances in
Multimedia
Submit your manuscripts atwwwhindawicom
4 Wireless Communications and Mobile Computing
RRH
ACE
RRH
Outage RRH
Cloud
Node C
ConnectLost connectionReconnect
RRH FronthaulACE Fronthaul
Figure 2 Cell outage in H-CRAN
Connection requestResource request
Resource responseConnection response
Connection setup
In Service
Outage OccursConnection failure
Connection requestResource request
Resource responseConnection response
Connection setup
Service Recovery
Counting point
Counting point
User RRH ACE NodeC
Figure 3 An instance of user reconnection process
Wireless Communications and Mobile Computing 5
Data Collection
User
RRHACE
NodeC
DT
Data Analysis
Pre-processing
Machine Learning
Temporal Predication
Spatial Predication
COD ConclusionCell Outage Compensation
Input
Input
Input
Input
Guideline
AchievePerformance Evaluation
MDT Data
Network topologyCell parameters
HOConnection RequestNCL
RSRPRSRQSINRCQI
Figure 4 Overview architecture for COD in 5G H-CRAN
Based on the above analysis we want to construct an inte-grated COD architecture for 5G H-CRAN and its detailedintroduction is given below
31 COD Architecture and Procedures Firstly the proposedarchitecture for COD can be found in Figure 4
As shown in Figure 4 COD mainly consists of datacollection and data analysis stages Firstly data collectionshould store data from different sources which are as follows
(1) Data collection from users through measurementreports such as RSRP RSRQ SINR and CQI infor-mation this information is always huge and hard tobe handled synchronously as the time interval is fairlydense So we can set a sampling interval and just takethe statistics for them
(2) Data collection from RRHACE such as HOCon-nection RequestNCL these data may come from theKPI statistics from OAM system These data are per-formance indicators and can be used directly as theinterval is defined beforehand under acceptable level
(3) Data collection for NodeC which may network leveldata such as network topology and preconfiguredcell parameters (transmit power spectrum antennaheight and tilt) these data can be used to supplementthe spatial and temporal analysis of different cells
(4) Drive test data drive test is a tool to verify networkperformance afterward It can be used as a validationfor COD conclusions and thus provide correctionsuggestions for COD approach
After data collection all the data can be put to dataanalysis stage to execute COD As shown in Figure 4 withSON entities which located at NodeC the procedure will beexecuted as follows
(1) SON entities firstly should preprocess these data toimprove the data quality
(2) SON entities adopt temporal and spatial predictionmethod to obtain varying patterns for temporal dataand spatial data respectively
(3) Next SON entities choose proper machine learningapproach to identify the outage cell
Still with COD results we can give suggestions for COCAnd COC effectiveness evaluation can be obtained throughnetwork performance monitoring thus constructing a self-organized loop The critical technologies of preprocessingspatial and temporal prediction andmachine learning will beintroduced next
32 Key Technologies for COD
321 Data Processing Data processing includes two conceptssuch as Data Cleaning and Feature Engineering These twoare compulsory for achieving better accuracy and perfor-mance before machine learning and deep learning And itincludes data cleaning data integration data transformationand data reduction So in our architecture we should chooseproper data processing approach to obtain high-quality data
322 Spatial Prediction Spatial prediction technologiesmainly aim at analysis spatial traffic distributions or cap-ture user variations Currently several methods have beenadopted in this field such as log-normal or Weibull distri-bution used in [18] or traffic patterns identifying methodsproposed in [19] With spatial prediction we can obtaindifferent distribution laws
323 Temporal Prediction Temporal prediction is aimingat predicting future variation direction based on past and
6 Wireless Communications and Mobile Computing
Table 1 Cell level data
Measurements DescriptionTime Transmission time interval (TTI)Cell info Cell global identity (CGI)Cell location Longitude and latitude informationCell inHO Cell incoming handover
current temporal sequence data for instance traffic variationof one ACE or load of NodeC Temporal prediction canadopt HoltndashWinter traffic forecast [20] or S-ARIMA model[21] to construct prediction sequences which will be usefulfor abnormal traffic recognition Moreover joint temporal-spatial predication approaches deservemore attention aswell
324 Machine Learning As analyzed in [22] machine learn-ing has been used for LTELTE-A self-healing and achievesremarkable results However thesemethods aremainly basedon single data source and do not take integrated data setshown in Section 31 together To make COD more accurateand intelligent in 5Gnetworks deep learningmodel [23] suchas DRN model can be considered
4 A Use Case for 5G H-CRAN COD
According to the architecture and procedures above weintroduce a use case to achieve COD for 5G H-CRAN Herewe suppose that the outage RRH or ACE cell cannot carrytraffic anymore Typically at this point all users served bythis cell have to be offloaded to its neighbor cells This willresult in a huge number of handovers in H-CRANThe risinghandovers in neighbor cells could be easily detected bymeansof data mining So we will discuss our algorithm by using thetemporal data of inHO statistics collected from the NodeC
41 Use Case Scenario Description The scenario is shown inFigure 5 It consists of 19 ACE with 25 users under coverageAll these users are active users Each ACE has 3 cells andone RRH cell is located under each ACE cell The UEs arerandomly distributed in cells to generate handover statisticsand measurement reports
Still we mainly collect cell level information stored inNodeC as shown in Table 1
Moreover in order to make the evaluation more obviouswe will evaluate the procedure under a simulation platformdeveloped by our lab Parameter settings can be found inTable 2
In our simulation platformwe assume that rapid decreaseof cellrsquos transmit power indicates cell outage That is we canset cell outage manually to evaluate efficiency of our ap-proach Then users under the affected outage areas switch tothe neighbor cells which will increase the number of han-dovers We will use the inHO data collected from all the cellswith a machine learning methods to evaluate whether outagecells can be detected
Table 2 Simulation parameters
Simulation parameters Value (unit)ACE 19 sites with 3 cells eachRRH 1 cell per siteUser number 25 per ACEUser distribution Uniform random distributionLink direction DownlinkBS Tx power 46 dBmPath loss model Cost231-HataUE velocity 15 kmhUser distribution UniformCell selection criteria Strongest RSRP defines the target cellSimulation length 150 TTI
RRH cellACE
Figure 5 Scenario use case for 5G H-CRAN
42 COD Procedures with M-LOF Considering the situa-tions above and the advantage of LOF shown in [9] a mod-ified LOF based detector is adopted to detect the outage byextracting both spatial and temporal correlations of handoverstatistics over the neighbors of outage cells The procedure isshown in Figure 6
The COD procedure includes four steps data collectionpreprocessingmachine learning and localization and all canbe executed automatically under the control of LS-SON Inthis way the outage cell can be detected timely so we canminimize the performance degradation of the network Herewe will introduce the functions in three steps
421 Preprocessing When data collection is finished SONentities proceed to extract feature vector 119879119894 for LOF [24]
Wireless Communications and Mobile Computing 7
Cell Level Data
Filter Standardization
Database
Low-dimension
Pre-processing
Data Collection
Machine Learning
Detection
Test z scoreCalculation
Store data to database
Localization
ParameterEstimation
No
M-LOFCalculation
OutageYes
H-CRAN
COC
thresholdZ gt
Figure 6 COD procedure with modified LOF
corresponding to each temporal data Since the inHO data iscollected from different cells at the different time the datavector is designed as spatial-temporal series
In fact the inHO statistics of outage cells are zero sincethe KPIs in outage cell cannot be collected Moreover for onepossibility there may be a few zero-inHO cells because noactive users are within their coverage This will enhance thedifficulty of detection In this case we filter all inHO statisticsof zero Here each cell splits out an ordered time series ofinHO To finish length-wise normalization and carry outuniformization and standardization processing we slice thetime series to several subseries
For temporal analysis 5 TTIs data is put into one featurevector 119879119894 as shown in below where 119894 represents the order ofsubseries
119879119894 = inHO1 inHO2 inHO3 inHO4 inHO5 (1)
The feature vector 119879119894 utilizes exactly one cellrsquos inHO datawhich is used to find out the time when an outage happened
For spatial analysis one feature vector 119878119895 consists of 119897cell handover statistics at TTI time 119895 where 119895 represents theidentity of the cell
119878119895 = inHO1 inHO2 inHO3 inHO4 inHO5 inHO6inHO7 inHO8 inHOl
(2)
After slicing the data series linear functional transforma-tion criterion is used to normalize the data to eliminate errorscaused by nonuniform features For example the criterion of119878119895 is defined as
1198781015840119895 =119878119895 minusmin119895 (119878119895)
max119895 (119878119895) minusmin119895 (119878119895) (3)
422 Machine Learning LOF detection is an unsupervisedanomaly detection algorithm The local densities of targetpoints are calculated by LOF used to compare with itsneighbors The larger the difference between the sampleand its neighbors is the larger outage factor score will beassigned to the sampleThe advantage of LOF is that no priorknowledge is required in advance to detect the unknowncell outage The detailed definitions of LOF are explained asfollows
Definition 1 k-distance of cell 119901 denoted as 119889119896(119901) representsthe distance between cell 119901 and its 119896th nearest neighbor sothat for at least 119896 cell 119900119894 satisfies
119889 (119900119894 119901) le 119889 (119900119896 119901) (4)
and for at most 119896 minus 1 cells 119900119894 satisfies119889 (119900119894 119901) lt 119889 (119900119896 119901) (5)
Definition 2 119896-distance neighborhood of cell 119901 denoted as119873119889119896(119901)(119901) when 119889119896(119901)is given 119873119889119896(119901)(119901) is the set whichincludes every cell whose distance to cell 119901 is smaller than119889119896(119901)Definition 3 Reachability distance of cell 119901 with respect tocell 119900 is denoted as rd119896(119901 119900) it is the maximum one of 119889119896(119901)and the distance between 119901 and 119900
rd119896 (119901 119900) = max 119889119896 (119900) 119889 (119900 119901) (6)
Definition 4 Local reachability density of cell p denoted aslrd119896(119901) is shown as follows
lrd119896 (119901) =10038161003816100381610038161003816119873119889119896(119901) (119901)
10038161003816100381610038161003816sum119900isin119873119889119896(119901)(119901) rd119896 (119901 o)
(7)
Definition 5 Local outlier factor of cell p denoted asLOF119896(119901) can be obtained as below
LOF119896 (119901) =sum119900isin119873119889119896(119901)(119901) (lrd119896 (119900) lrd119896 (119901))10038161003816100381610038161003816119873119889119896(119901) (119901)
10038161003816100381610038161003816 (8)
The above descriptions give the definition of LOF Fromour previous work in [17] the proposed M-LOF is a properapproach for heterogeneous network so here we use themodified LOF algorithm called M-LOF to make a refinedchoice for H-CRAN Definitions of M-LOF and m-distanceare given below
8 Wireless Communications and Mobile Computing
0
0
InHo DataMDT DataAbnormal cells
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80
Cell Id
minus15minus10minus05000510152025303540455055
MLO
F Va
lue
1
23
4
5
6
7
8
9
10
11
12
13
14
1516
17
2122
24
25 27
29
30
3233
35
36
38 4143
44
4647
49
52
54
55
56
57
58
59
6062
63
64
65
66
67
69
68
7071
72
73
74
75
76
1
2
3
4
5
6
7
89
10
11
1213
14
15
1617
21
22 2425 27 29
30 32
33
35 36 38
41
43
44
46
47
49
52
54
55
56575859
6062
63
6465
66
6768
69
70717273
74
75
76
Figure 7 Spatial analysis for abnormal cells
Definition 6 119898-distance of cell 119901 given the positive integer119896 is defined as
119898minus119889119896 (119901) = 120576 + [[sum119900isin119873119889119896(119901)(119901) 119889 (119901 119900)10038161003816100381610038161003816119873119889119896(119901) (119901)
10038161003816100381610038161003816]] (9)
Here 120576 is a constant value to enhance the accuracyDefinition 7 119898-distance neighborhood of cell p denotedas 119873119898minus119889119896(119901)(119901) is the set which includes every cell whichdistance to cell 119901 is smaller thanm-distance
Definition 8 Reachability distance of m-distance of cell 119901concerning cell o denoted as 119903minus119889119898(119901 119900) is
119903minus119889119898 (119901 119900) = max 119898minus119889119896 (119900) 119889 (119900 119901) (10)
Definition 9 Local reachability density of cell p is defined as
lrd119898 (119901) =10038161003816100381610038161003816119873119898minus119889119896(119901) (119901)
10038161003816100381610038161003816sum119900isin119873119898minus119889119896(119901)(119901) 119903minus119889119898 (119901 o)
(11)
Definition 10 Modified local outlier factor of cell p is shownbelow
LOF119898 (119901) =sum119900isin119873119898minus119889119896(119901)(119901) (lrd119898 (119900) lrd119898 (119901))10038161003816100381610038161003816119873119898minus119889119896(119901) (119901)
10038161003816100381610038161003816 (12)
According to the definition of LOF119896(119901) and LOF119898(119901) thechoice of 119896 is exactly sensitive As a consequence if there is aprior experience we will choose the cross-validation methodto estimate the parameter k
423 Localization The last step is localization based on theoutput of LOF calculation Here the neighbor cell list [25] isused to search the relation between the outage cell and itsneighbor cells by geographic informationwith 119911-score In thisway the outage cells may be localized
43 Results Analysis Based on the above steps M-LOFwould detect anomalous behaviors of handover And in thesimulation the value of 119896 for the LOF based detector is foundbetween 5 and 14
For the purpose of validation we will first analyze theresults from the spatial and temporal perspectives and thendetermine outage cell locations with neighbor cell list andperformance comparison between M-LOF and LOF will begiven at last
431 Spatial Analysis For spatial analysis we focus on thetemporal data of different cells at the same time periodFigure 7 shows the M-LOF of each cell at 95th TTI It can beseen that factor values derived from six cellsrsquo inHO data arefar higher than the normal value which is usually less than15 Therefore the abnormal cells can be distinguished fromnormal cells as they have experienced many user reconnec-tions
For comparison Figure 7 also shows the detection resultusingMDTmeasurements (with RSRP and SINR) as the datasource It can be seen that the abnormal cells are hard to bedetected using MDT Because M-LOF values of cells basedon MDT data are smoother than inHO data In contrastour proposed method using the inHO data has a betterperformance for abnormal cell detection in H-CRAN
432 Temporal Analysis Figure 8 shows the temporal analy-sis result for M-LOF variations relative to TTI for outage cell13 At this time from the results we find that M-LOF valuearises between TTI 90 and TTI 100 which is in accordancewith our outage time setting As initial connections of usersoccur between 0ndash5 TTI under unstable status so we justignore these data
433 Localization After the implementation of M-LOFbased detector neighbor cell list is used to localize the outage
Wireless Communications and Mobile Computing 9
TTI
Outage Period
22
20
18
16
14
12
10
08
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140
MLO
F Va
lue 5
1015
2025 30
35 40
4550
55
60
6570
75
80
85
90
95
100 105
110
115120
125
130
135140
Figure 8 Temporal analysis for outage TTI
cells First we need to set a threshold to filter anomalous cellsthrough 119911-score calculation as
119911119896 = 119899119896 minus 120583119899120590119899 (13)
where 119899119896 is the M-LOF value of kth cell and 120583119899 120590119899 arethe mean and standard deviation of M-LOF value for theanomaly scores of the other cells The reference 119911-scorethreshold is configured with a preferred value for the abnor-mal cells Here the reference 119911-score is set to 21 according tothe computing results We conducted two sets of evaluationsto make a comparison of performance of COD based oninHO data and MDT data respectively The results canidentify outage cell 18 and cell 61 which is the right one asshown in Figure 9 as LOF values of their neighbor cells arerelatively higher than other cells
434 Performance Evaluation In this part we analyzedthe M-LOF detection performance under varying trafficconditions Since the behaviors of users have a direct effect oninHO data the diagnosis process has been tested in differentscenarios by changing the User Density (UD) and UserVelocity (UV) parameters as the baseline setup To evaluatethe impact of the variations of UD and UV on M-LOFvalues different scenarios are set up by adjusting these twoparameters The Cumulative Distribution Function (CDF) oftheM-LOF values under different UV conditions is shown inFigure 10
It can be seen that for the low-velocity scenario almost80 of the M-LOF values are less than 05 However thereis a significant reduction in the M-LOF value as the UVincreases Likewise a similar behavior is observed with theincrease of UD as shown in Figure 11 The UV and UDparameters influence the distribution and spread of the inHOdata as explained earlier and consequently the value of M-LOF This leads to a low detection performance of M-LOFsince it generates an increased number of false alarms
Table 3 Performance evaluation
Result LOF value M-LOF valueFalse negative rate 3 3False positive rate 12 6
Outage cellAffected cell
800
600
400
200
Use
r Pos
ition
Y
User Position X
100 200 300 400 500 600 700 800
1
2
3
4
5
6
7
8
9
10
11
2122
25
27
29
30
32
33
64 63
62
60
1258 71
1773
16
68
36
35
14
6113
59
2456
57
74
1875
65
76
55
0
72
49
52
54
47
38 43
6615
67
70 69
41
44
46
Figure 9 Outage cell location results
At last this paper compares LOF and M-LOF by evaluat-ing False Positive Rate (FPR) and False Negative Rate (FNR)and the final results of the simulation are shown in Table 3The FNR represents the chance that an outage cell is notrecognized from the outage cells while the FPR representsthe chance that a normal cell is recognized as the outage cellfrom all normal cells
From Table 3 we can see that FPR and FNR of LOF baseddetector are 12 and 3 respectively It means the outagecell can be almost wholly detected although a small part ofthe normal cell may be determined as an anomaly The maincause of the 3FNR is that a small number of anomalous cellshave low traffic so that seldom user data cannot be collectedwhen they are in outage So these small cells are failed to bedetected However these outages have little impact on overallnetwork performance and user experience Therefore wecan say the outage cells can be detected successfully withLOF based detector approach As for M-LOF the FPR is 6which is smaller than LOFThe reason whyM-LOF has betterperformance is that M-LOF pays more attention to localdensity This is in line with the actual use of outage detectionsince the abnormal handovers caused by the neighbor outagecell are also localized
Still asM-LOF just modified the distance computation ofLOF so for time complexity it is just the same as LOF withis 119874(1198992) Here 119899 is the TTI number or cell number shown inFigure 7 or Figure 8
10 Wireless Communications and Mobile Computing
10
08
06
04
02
00minus2 minus1 0 1 2 3 4 5
0mM lt 5 6FI=CNS lt 1ms1mM lt 5 6FI=CNS lt 3ms3mM lt 5 6FI=CNS lt 5ms
M-LOF
CDF
Figure 10 CDF of M-LOF for User Velocity
10
08
06
04
02
00minus2 minus1 0 1 2 3 4 65
M-LOF
CDF
UE Density = 30 UsersCellUE Density = 20 UsersCellUE Density = 10 UsersCell
Figure 11 CDF of M-LOF value for User Density
5 Conclusions
In this paper we propose self-organized cell outage detectionarchitecture and approach for 5G H-CRAN The previousmethods using single data source are hard to detect thehidden outage such as RRHs Aiming at this problem weintroduce architecture and corresponding approach to realizeCOD for H-CRAN and still analyze the required key tech-nologies Then we evaluate the approach under a use casewhich utilizes the collected inHO statistics and adopt M-LOF to detect the anomalous behavior outage cells Resultsshow that the use case is effective However much work stillhas to be done for this architecture as we do not considerresource allocation and COC solutions still we have to takemore intelligent learning model such as deep learning into
account as well and compare the efficiency of different algo-rithms under a embodied H-CRAN network deploymentsMoreover we will focus on a specific perspective for self-healing approaches for H-CRAN and propose more practicalsolutions
Conflicts of Interest
The authors declare that they have no conflicts of interest
Acknowledgments
This research is supported by the National Science andTechnology Major Project of the Ministry of Science andTechnology of China (no 2018ZX030110004)
References
[1] P Rost A Banchs I Berberana et al ldquoMobile network archi-tecture evolution toward 5Grdquo IEEE Communications Magazinevol 54 no 5 pp 84ndash91 2016
[2] M Peng Y Sun X Li Z Mao and C Wang ldquoRecentadvances in cloud radio access networks system architectureskey techniques and open issuesrdquo IEEECommunications Surveysamp Tutorials vol 18 no 3 pp 2282ndash2308 2016
[3] M Peng Y Li Z Zhao and C Wang ldquoSystem architectureand key technologies for 5G heterogeneous cloud radio accessnetworksrdquo IEEE Network vol 29 no 2 pp 6ndash14 2015
[4] N Chen B Rong X Zhang and M Kadoch ldquoScalable andflexible massive MIMO precoding for 5G H-CRANrdquo IEEEWireless Communications Magazine vol 24 no 1 pp 46ndash522017
[5] OGAliu A ImranMA Imran andB Evans ldquoA survey of selforganisation in future cellular networksrdquo IEEECommunicationsSurveys amp Tutorials vol 15 no 1 pp 336ndash361 2013
[6] E J Khatib R Barco P Munoz I D La Bandera and ISerrano ldquoSelf-healing in mobile networks with big datardquo IEEECommunications Magazine vol 54 no 1 pp 114ndash120 2016
[7] M Amirijoo L Jorguseski T Kurner et al ldquoCell outagemanagement in LTE networksrdquo in Proceedings of the 2009 6thInternational Symposium on Wireless Communication Systems(ISWCS 2009) pp 600ndash604 Siena Italy September 2009
[8] M YinW Li L Feng P Yu and X Qiu ldquoMulti-cell cooperativeoutage compensation in cloud-RANs based 5G public safetynetworkrdquo IEEE Access vol 5 pp 17309ndash17321 2017
[9] A Zoha A Saeed A Imran M A Imran and A Abu-DayyaldquoData-driven analytics for automated cell outage detection inSelf-Organizing Networksrdquo in Proceedings of the 2015 11th Inter-national Conference on the Design of Reliable CommunicationNetworks DRCN 2015 pp 203ndash210 usa March 2015
[10] I De-La-Bandera R Barco P Munoz and I Serrano ldquoCellOutage Detection Based onHandover Statisticsrdquo IEEE Commu-nications Letters vol 19 no 7 pp 1189ndash1192 2015
[11] W Wang Q Liao and Q Zhang ldquoCOD A cooperativecell outage detection architecture for self-organizing femtocellnetworksrdquo IEEE Transactions on Wireless Communications vol13 no 11 pp 6007ndash6014 2014
[12] Q Liao M Wiczanowski and S Stanczak ldquoToward cell outagedetection with composite hypothesis testingrdquo in Proceedingsof the 2012 IEEE International Conference on CommunicationsICC 2012 pp 4883ndash4887 can June 2012
Wireless Communications and Mobile Computing 11
[13] Y Ma M Peng W Xue and X Ji ldquoA dynamic affinity propa-gation clustering algorithm for cell outage detection in self-healing networksrdquo in Proceedings of the 2013 IEEE WirelessCommunications and Networking Conference WCNC 2013 pp2266ndash2270 chn April 2013
[14] M Alias N Saxena and A Roy ldquoEfficient Cell Outage De-tection in 5G HetNets Using Hidden Markov Modelrdquo IEEECommunications Letters vol 20 no 3 pp 562ndash565 2016
[15] W Xue M Peng Y Ma and H Zhang ldquoClassification-basedapproach for cell outage detection in self-healing heterogeneousnetworksrdquo in Proceedings of the 2014 IEEEWireless Communica-tions and Networking Conference WCNC 2014 pp 2822ndash2826tur April 2014
[16] J Wang N Q Phan Z Pan N Liu X You and T Deng ldquoAnimproved TCM-based approach for cell outage detection forself-healing in LTE HetNetsrdquo in Proceedings of the 83rd IEEEVehicular Technology Conference VTC Spring 2016 chn May2016
[17] T Zhang L Feng P Yu S Guo W Li and X Qiu ldquoA handoverstatistics based approach for Cell Outage Detection in self-organized Heterogeneous Networksrdquo in Proceedings of the 15thIFIPIEEE International Symposium on Integrated Network andService Management IM 2017 pp 628ndash631 prt May 2017
[18] D Lee S Zhou X Zhong Z Niu X Zhou and H ZhangldquoSpatial modeling of the traffic density in cellular networksrdquoIEEE Wireless Communications Magazine vol 21 no 1 pp 80ndash88 2014
[19] F Xu Y Li H Wang P Zhang and D Jin ldquoUnderstandingMobile Traffic Patterns of Large Scale Cellular Towers in UrbanEnvironmentrdquo IEEEACM Transactions on Networking vol 25no 2 pp 1147ndash1161 2017
[20] S Morosi P Piunti and E Del Re ldquoSleep mode managementin cellular networks a traffic based technique enabling energysavingrdquo Transactions on Emerging Telecommunications Tech-nologies vol 24 no 3 pp 331ndash341 2013
[21] W Y Szeto B Ghosh B Basu andM OrsquoMahony ldquoMultivariatetraffic forecasting technique using cell transmission model andSARIMA modelrdquo Journal of Transportation Engineering vol135 no 9 pp 658ndash667 2009
[22] P V KlaineM A Imran O Onireti and R D Souza ldquoA Surveyof Machine Learning Techniques Applied to Self OrganizingCellular Networksrdquo IEEE Communications Surveys amp Tutorials2017
[23] Z M Fadlulla F Tang B Mao and et al ldquoState-of-the-Art Deep Learning Evolving Machine Intelligence TowardTomorrowrsquos Intelligent Network Traffic Control Systemsrdquo IEEECommunications Surveys amp Tutorials vol 19 no 4 pp 2432ndash2455 2017
[24] M M Breuniq H Kriegel R T Ng and J Sander ldquoLOFidentifying density-based local outliersrdquo SIGMOD Record vol29 no 2 pp 93ndash104 2000
[25] C M Mueller M Kaschub C Blankenhorn and S WankeldquoA Cell Outage Detection Algorithm Using Neighbor Cell ListReportsrdquo in Self-Organizing Systems vol 5343 of Lecture Notesin Computer Science pp 218ndash229 Springer Berlin HeidelbergBerlin Heidelberg 2008
International Journal of
AerospaceEngineeringHindawiwwwhindawicom Volume 2018
RoboticsJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Active and Passive Electronic Components
VLSI Design
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Shock and Vibration
Hindawiwwwhindawicom Volume 2018
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawiwwwhindawicom
Volume 2018
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Control Scienceand Engineering
Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom
Journal ofEngineeringVolume 2018
SensorsJournal of
Hindawiwwwhindawicom Volume 2018
International Journal of
RotatingMachinery
Hindawiwwwhindawicom Volume 2018
Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Navigation and Observation
International Journal of
Hindawi
wwwhindawicom Volume 2018
Advances in
Multimedia
Submit your manuscripts atwwwhindawicom
Wireless Communications and Mobile Computing 5
Data Collection
User
RRHACE
NodeC
DT
Data Analysis
Pre-processing
Machine Learning
Temporal Predication
Spatial Predication
COD ConclusionCell Outage Compensation
Input
Input
Input
Input
Guideline
AchievePerformance Evaluation
MDT Data
Network topologyCell parameters
HOConnection RequestNCL
RSRPRSRQSINRCQI
Figure 4 Overview architecture for COD in 5G H-CRAN
Based on the above analysis we want to construct an inte-grated COD architecture for 5G H-CRAN and its detailedintroduction is given below
31 COD Architecture and Procedures Firstly the proposedarchitecture for COD can be found in Figure 4
As shown in Figure 4 COD mainly consists of datacollection and data analysis stages Firstly data collectionshould store data from different sources which are as follows
(1) Data collection from users through measurementreports such as RSRP RSRQ SINR and CQI infor-mation this information is always huge and hard tobe handled synchronously as the time interval is fairlydense So we can set a sampling interval and just takethe statistics for them
(2) Data collection from RRHACE such as HOCon-nection RequestNCL these data may come from theKPI statistics from OAM system These data are per-formance indicators and can be used directly as theinterval is defined beforehand under acceptable level
(3) Data collection for NodeC which may network leveldata such as network topology and preconfiguredcell parameters (transmit power spectrum antennaheight and tilt) these data can be used to supplementthe spatial and temporal analysis of different cells
(4) Drive test data drive test is a tool to verify networkperformance afterward It can be used as a validationfor COD conclusions and thus provide correctionsuggestions for COD approach
After data collection all the data can be put to dataanalysis stage to execute COD As shown in Figure 4 withSON entities which located at NodeC the procedure will beexecuted as follows
(1) SON entities firstly should preprocess these data toimprove the data quality
(2) SON entities adopt temporal and spatial predictionmethod to obtain varying patterns for temporal dataand spatial data respectively
(3) Next SON entities choose proper machine learningapproach to identify the outage cell
Still with COD results we can give suggestions for COCAnd COC effectiveness evaluation can be obtained throughnetwork performance monitoring thus constructing a self-organized loop The critical technologies of preprocessingspatial and temporal prediction andmachine learning will beintroduced next
32 Key Technologies for COD
321 Data Processing Data processing includes two conceptssuch as Data Cleaning and Feature Engineering These twoare compulsory for achieving better accuracy and perfor-mance before machine learning and deep learning And itincludes data cleaning data integration data transformationand data reduction So in our architecture we should chooseproper data processing approach to obtain high-quality data
322 Spatial Prediction Spatial prediction technologiesmainly aim at analysis spatial traffic distributions or cap-ture user variations Currently several methods have beenadopted in this field such as log-normal or Weibull distri-bution used in [18] or traffic patterns identifying methodsproposed in [19] With spatial prediction we can obtaindifferent distribution laws
323 Temporal Prediction Temporal prediction is aimingat predicting future variation direction based on past and
6 Wireless Communications and Mobile Computing
Table 1 Cell level data
Measurements DescriptionTime Transmission time interval (TTI)Cell info Cell global identity (CGI)Cell location Longitude and latitude informationCell inHO Cell incoming handover
current temporal sequence data for instance traffic variationof one ACE or load of NodeC Temporal prediction canadopt HoltndashWinter traffic forecast [20] or S-ARIMA model[21] to construct prediction sequences which will be usefulfor abnormal traffic recognition Moreover joint temporal-spatial predication approaches deservemore attention aswell
324 Machine Learning As analyzed in [22] machine learn-ing has been used for LTELTE-A self-healing and achievesremarkable results However thesemethods aremainly basedon single data source and do not take integrated data setshown in Section 31 together To make COD more accurateand intelligent in 5Gnetworks deep learningmodel [23] suchas DRN model can be considered
4 A Use Case for 5G H-CRAN COD
According to the architecture and procedures above weintroduce a use case to achieve COD for 5G H-CRAN Herewe suppose that the outage RRH or ACE cell cannot carrytraffic anymore Typically at this point all users served bythis cell have to be offloaded to its neighbor cells This willresult in a huge number of handovers in H-CRANThe risinghandovers in neighbor cells could be easily detected bymeansof data mining So we will discuss our algorithm by using thetemporal data of inHO statistics collected from the NodeC
41 Use Case Scenario Description The scenario is shown inFigure 5 It consists of 19 ACE with 25 users under coverageAll these users are active users Each ACE has 3 cells andone RRH cell is located under each ACE cell The UEs arerandomly distributed in cells to generate handover statisticsand measurement reports
Still we mainly collect cell level information stored inNodeC as shown in Table 1
Moreover in order to make the evaluation more obviouswe will evaluate the procedure under a simulation platformdeveloped by our lab Parameter settings can be found inTable 2
In our simulation platformwe assume that rapid decreaseof cellrsquos transmit power indicates cell outage That is we canset cell outage manually to evaluate efficiency of our ap-proach Then users under the affected outage areas switch tothe neighbor cells which will increase the number of han-dovers We will use the inHO data collected from all the cellswith a machine learning methods to evaluate whether outagecells can be detected
Table 2 Simulation parameters
Simulation parameters Value (unit)ACE 19 sites with 3 cells eachRRH 1 cell per siteUser number 25 per ACEUser distribution Uniform random distributionLink direction DownlinkBS Tx power 46 dBmPath loss model Cost231-HataUE velocity 15 kmhUser distribution UniformCell selection criteria Strongest RSRP defines the target cellSimulation length 150 TTI
RRH cellACE
Figure 5 Scenario use case for 5G H-CRAN
42 COD Procedures with M-LOF Considering the situa-tions above and the advantage of LOF shown in [9] a mod-ified LOF based detector is adopted to detect the outage byextracting both spatial and temporal correlations of handoverstatistics over the neighbors of outage cells The procedure isshown in Figure 6
The COD procedure includes four steps data collectionpreprocessingmachine learning and localization and all canbe executed automatically under the control of LS-SON Inthis way the outage cell can be detected timely so we canminimize the performance degradation of the network Herewe will introduce the functions in three steps
421 Preprocessing When data collection is finished SONentities proceed to extract feature vector 119879119894 for LOF [24]
Wireless Communications and Mobile Computing 7
Cell Level Data
Filter Standardization
Database
Low-dimension
Pre-processing
Data Collection
Machine Learning
Detection
Test z scoreCalculation
Store data to database
Localization
ParameterEstimation
No
M-LOFCalculation
OutageYes
H-CRAN
COC
thresholdZ gt
Figure 6 COD procedure with modified LOF
corresponding to each temporal data Since the inHO data iscollected from different cells at the different time the datavector is designed as spatial-temporal series
In fact the inHO statistics of outage cells are zero sincethe KPIs in outage cell cannot be collected Moreover for onepossibility there may be a few zero-inHO cells because noactive users are within their coverage This will enhance thedifficulty of detection In this case we filter all inHO statisticsof zero Here each cell splits out an ordered time series ofinHO To finish length-wise normalization and carry outuniformization and standardization processing we slice thetime series to several subseries
For temporal analysis 5 TTIs data is put into one featurevector 119879119894 as shown in below where 119894 represents the order ofsubseries
119879119894 = inHO1 inHO2 inHO3 inHO4 inHO5 (1)
The feature vector 119879119894 utilizes exactly one cellrsquos inHO datawhich is used to find out the time when an outage happened
For spatial analysis one feature vector 119878119895 consists of 119897cell handover statistics at TTI time 119895 where 119895 represents theidentity of the cell
119878119895 = inHO1 inHO2 inHO3 inHO4 inHO5 inHO6inHO7 inHO8 inHOl
(2)
After slicing the data series linear functional transforma-tion criterion is used to normalize the data to eliminate errorscaused by nonuniform features For example the criterion of119878119895 is defined as
1198781015840119895 =119878119895 minusmin119895 (119878119895)
max119895 (119878119895) minusmin119895 (119878119895) (3)
422 Machine Learning LOF detection is an unsupervisedanomaly detection algorithm The local densities of targetpoints are calculated by LOF used to compare with itsneighbors The larger the difference between the sampleand its neighbors is the larger outage factor score will beassigned to the sampleThe advantage of LOF is that no priorknowledge is required in advance to detect the unknowncell outage The detailed definitions of LOF are explained asfollows
Definition 1 k-distance of cell 119901 denoted as 119889119896(119901) representsthe distance between cell 119901 and its 119896th nearest neighbor sothat for at least 119896 cell 119900119894 satisfies
119889 (119900119894 119901) le 119889 (119900119896 119901) (4)
and for at most 119896 minus 1 cells 119900119894 satisfies119889 (119900119894 119901) lt 119889 (119900119896 119901) (5)
Definition 2 119896-distance neighborhood of cell 119901 denoted as119873119889119896(119901)(119901) when 119889119896(119901)is given 119873119889119896(119901)(119901) is the set whichincludes every cell whose distance to cell 119901 is smaller than119889119896(119901)Definition 3 Reachability distance of cell 119901 with respect tocell 119900 is denoted as rd119896(119901 119900) it is the maximum one of 119889119896(119901)and the distance between 119901 and 119900
rd119896 (119901 119900) = max 119889119896 (119900) 119889 (119900 119901) (6)
Definition 4 Local reachability density of cell p denoted aslrd119896(119901) is shown as follows
lrd119896 (119901) =10038161003816100381610038161003816119873119889119896(119901) (119901)
10038161003816100381610038161003816sum119900isin119873119889119896(119901)(119901) rd119896 (119901 o)
(7)
Definition 5 Local outlier factor of cell p denoted asLOF119896(119901) can be obtained as below
LOF119896 (119901) =sum119900isin119873119889119896(119901)(119901) (lrd119896 (119900) lrd119896 (119901))10038161003816100381610038161003816119873119889119896(119901) (119901)
10038161003816100381610038161003816 (8)
The above descriptions give the definition of LOF Fromour previous work in [17] the proposed M-LOF is a properapproach for heterogeneous network so here we use themodified LOF algorithm called M-LOF to make a refinedchoice for H-CRAN Definitions of M-LOF and m-distanceare given below
8 Wireless Communications and Mobile Computing
0
0
InHo DataMDT DataAbnormal cells
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80
Cell Id
minus15minus10minus05000510152025303540455055
MLO
F Va
lue
1
23
4
5
6
7
8
9
10
11
12
13
14
1516
17
2122
24
25 27
29
30
3233
35
36
38 4143
44
4647
49
52
54
55
56
57
58
59
6062
63
64
65
66
67
69
68
7071
72
73
74
75
76
1
2
3
4
5
6
7
89
10
11
1213
14
15
1617
21
22 2425 27 29
30 32
33
35 36 38
41
43
44
46
47
49
52
54
55
56575859
6062
63
6465
66
6768
69
70717273
74
75
76
Figure 7 Spatial analysis for abnormal cells
Definition 6 119898-distance of cell 119901 given the positive integer119896 is defined as
119898minus119889119896 (119901) = 120576 + [[sum119900isin119873119889119896(119901)(119901) 119889 (119901 119900)10038161003816100381610038161003816119873119889119896(119901) (119901)
10038161003816100381610038161003816]] (9)
Here 120576 is a constant value to enhance the accuracyDefinition 7 119898-distance neighborhood of cell p denotedas 119873119898minus119889119896(119901)(119901) is the set which includes every cell whichdistance to cell 119901 is smaller thanm-distance
Definition 8 Reachability distance of m-distance of cell 119901concerning cell o denoted as 119903minus119889119898(119901 119900) is
119903minus119889119898 (119901 119900) = max 119898minus119889119896 (119900) 119889 (119900 119901) (10)
Definition 9 Local reachability density of cell p is defined as
lrd119898 (119901) =10038161003816100381610038161003816119873119898minus119889119896(119901) (119901)
10038161003816100381610038161003816sum119900isin119873119898minus119889119896(119901)(119901) 119903minus119889119898 (119901 o)
(11)
Definition 10 Modified local outlier factor of cell p is shownbelow
LOF119898 (119901) =sum119900isin119873119898minus119889119896(119901)(119901) (lrd119898 (119900) lrd119898 (119901))10038161003816100381610038161003816119873119898minus119889119896(119901) (119901)
10038161003816100381610038161003816 (12)
According to the definition of LOF119896(119901) and LOF119898(119901) thechoice of 119896 is exactly sensitive As a consequence if there is aprior experience we will choose the cross-validation methodto estimate the parameter k
423 Localization The last step is localization based on theoutput of LOF calculation Here the neighbor cell list [25] isused to search the relation between the outage cell and itsneighbor cells by geographic informationwith 119911-score In thisway the outage cells may be localized
43 Results Analysis Based on the above steps M-LOFwould detect anomalous behaviors of handover And in thesimulation the value of 119896 for the LOF based detector is foundbetween 5 and 14
For the purpose of validation we will first analyze theresults from the spatial and temporal perspectives and thendetermine outage cell locations with neighbor cell list andperformance comparison between M-LOF and LOF will begiven at last
431 Spatial Analysis For spatial analysis we focus on thetemporal data of different cells at the same time periodFigure 7 shows the M-LOF of each cell at 95th TTI It can beseen that factor values derived from six cellsrsquo inHO data arefar higher than the normal value which is usually less than15 Therefore the abnormal cells can be distinguished fromnormal cells as they have experienced many user reconnec-tions
For comparison Figure 7 also shows the detection resultusingMDTmeasurements (with RSRP and SINR) as the datasource It can be seen that the abnormal cells are hard to bedetected using MDT Because M-LOF values of cells basedon MDT data are smoother than inHO data In contrastour proposed method using the inHO data has a betterperformance for abnormal cell detection in H-CRAN
432 Temporal Analysis Figure 8 shows the temporal analy-sis result for M-LOF variations relative to TTI for outage cell13 At this time from the results we find that M-LOF valuearises between TTI 90 and TTI 100 which is in accordancewith our outage time setting As initial connections of usersoccur between 0ndash5 TTI under unstable status so we justignore these data
433 Localization After the implementation of M-LOFbased detector neighbor cell list is used to localize the outage
Wireless Communications and Mobile Computing 9
TTI
Outage Period
22
20
18
16
14
12
10
08
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140
MLO
F Va
lue 5
1015
2025 30
35 40
4550
55
60
6570
75
80
85
90
95
100 105
110
115120
125
130
135140
Figure 8 Temporal analysis for outage TTI
cells First we need to set a threshold to filter anomalous cellsthrough 119911-score calculation as
119911119896 = 119899119896 minus 120583119899120590119899 (13)
where 119899119896 is the M-LOF value of kth cell and 120583119899 120590119899 arethe mean and standard deviation of M-LOF value for theanomaly scores of the other cells The reference 119911-scorethreshold is configured with a preferred value for the abnor-mal cells Here the reference 119911-score is set to 21 according tothe computing results We conducted two sets of evaluationsto make a comparison of performance of COD based oninHO data and MDT data respectively The results canidentify outage cell 18 and cell 61 which is the right one asshown in Figure 9 as LOF values of their neighbor cells arerelatively higher than other cells
434 Performance Evaluation In this part we analyzedthe M-LOF detection performance under varying trafficconditions Since the behaviors of users have a direct effect oninHO data the diagnosis process has been tested in differentscenarios by changing the User Density (UD) and UserVelocity (UV) parameters as the baseline setup To evaluatethe impact of the variations of UD and UV on M-LOFvalues different scenarios are set up by adjusting these twoparameters The Cumulative Distribution Function (CDF) oftheM-LOF values under different UV conditions is shown inFigure 10
It can be seen that for the low-velocity scenario almost80 of the M-LOF values are less than 05 However thereis a significant reduction in the M-LOF value as the UVincreases Likewise a similar behavior is observed with theincrease of UD as shown in Figure 11 The UV and UDparameters influence the distribution and spread of the inHOdata as explained earlier and consequently the value of M-LOF This leads to a low detection performance of M-LOFsince it generates an increased number of false alarms
Table 3 Performance evaluation
Result LOF value M-LOF valueFalse negative rate 3 3False positive rate 12 6
Outage cellAffected cell
800
600
400
200
Use
r Pos
ition
Y
User Position X
100 200 300 400 500 600 700 800
1
2
3
4
5
6
7
8
9
10
11
2122
25
27
29
30
32
33
64 63
62
60
1258 71
1773
16
68
36
35
14
6113
59
2456
57
74
1875
65
76
55
0
72
49
52
54
47
38 43
6615
67
70 69
41
44
46
Figure 9 Outage cell location results
At last this paper compares LOF and M-LOF by evaluat-ing False Positive Rate (FPR) and False Negative Rate (FNR)and the final results of the simulation are shown in Table 3The FNR represents the chance that an outage cell is notrecognized from the outage cells while the FPR representsthe chance that a normal cell is recognized as the outage cellfrom all normal cells
From Table 3 we can see that FPR and FNR of LOF baseddetector are 12 and 3 respectively It means the outagecell can be almost wholly detected although a small part ofthe normal cell may be determined as an anomaly The maincause of the 3FNR is that a small number of anomalous cellshave low traffic so that seldom user data cannot be collectedwhen they are in outage So these small cells are failed to bedetected However these outages have little impact on overallnetwork performance and user experience Therefore wecan say the outage cells can be detected successfully withLOF based detector approach As for M-LOF the FPR is 6which is smaller than LOFThe reason whyM-LOF has betterperformance is that M-LOF pays more attention to localdensity This is in line with the actual use of outage detectionsince the abnormal handovers caused by the neighbor outagecell are also localized
Still asM-LOF just modified the distance computation ofLOF so for time complexity it is just the same as LOF withis 119874(1198992) Here 119899 is the TTI number or cell number shown inFigure 7 or Figure 8
10 Wireless Communications and Mobile Computing
10
08
06
04
02
00minus2 minus1 0 1 2 3 4 5
0mM lt 5 6FI=CNS lt 1ms1mM lt 5 6FI=CNS lt 3ms3mM lt 5 6FI=CNS lt 5ms
M-LOF
CDF
Figure 10 CDF of M-LOF for User Velocity
10
08
06
04
02
00minus2 minus1 0 1 2 3 4 65
M-LOF
CDF
UE Density = 30 UsersCellUE Density = 20 UsersCellUE Density = 10 UsersCell
Figure 11 CDF of M-LOF value for User Density
5 Conclusions
In this paper we propose self-organized cell outage detectionarchitecture and approach for 5G H-CRAN The previousmethods using single data source are hard to detect thehidden outage such as RRHs Aiming at this problem weintroduce architecture and corresponding approach to realizeCOD for H-CRAN and still analyze the required key tech-nologies Then we evaluate the approach under a use casewhich utilizes the collected inHO statistics and adopt M-LOF to detect the anomalous behavior outage cells Resultsshow that the use case is effective However much work stillhas to be done for this architecture as we do not considerresource allocation and COC solutions still we have to takemore intelligent learning model such as deep learning into
account as well and compare the efficiency of different algo-rithms under a embodied H-CRAN network deploymentsMoreover we will focus on a specific perspective for self-healing approaches for H-CRAN and propose more practicalsolutions
Conflicts of Interest
The authors declare that they have no conflicts of interest
Acknowledgments
This research is supported by the National Science andTechnology Major Project of the Ministry of Science andTechnology of China (no 2018ZX030110004)
References
[1] P Rost A Banchs I Berberana et al ldquoMobile network archi-tecture evolution toward 5Grdquo IEEE Communications Magazinevol 54 no 5 pp 84ndash91 2016
[2] M Peng Y Sun X Li Z Mao and C Wang ldquoRecentadvances in cloud radio access networks system architectureskey techniques and open issuesrdquo IEEECommunications Surveysamp Tutorials vol 18 no 3 pp 2282ndash2308 2016
[3] M Peng Y Li Z Zhao and C Wang ldquoSystem architectureand key technologies for 5G heterogeneous cloud radio accessnetworksrdquo IEEE Network vol 29 no 2 pp 6ndash14 2015
[4] N Chen B Rong X Zhang and M Kadoch ldquoScalable andflexible massive MIMO precoding for 5G H-CRANrdquo IEEEWireless Communications Magazine vol 24 no 1 pp 46ndash522017
[5] OGAliu A ImranMA Imran andB Evans ldquoA survey of selforganisation in future cellular networksrdquo IEEECommunicationsSurveys amp Tutorials vol 15 no 1 pp 336ndash361 2013
[6] E J Khatib R Barco P Munoz I D La Bandera and ISerrano ldquoSelf-healing in mobile networks with big datardquo IEEECommunications Magazine vol 54 no 1 pp 114ndash120 2016
[7] M Amirijoo L Jorguseski T Kurner et al ldquoCell outagemanagement in LTE networksrdquo in Proceedings of the 2009 6thInternational Symposium on Wireless Communication Systems(ISWCS 2009) pp 600ndash604 Siena Italy September 2009
[8] M YinW Li L Feng P Yu and X Qiu ldquoMulti-cell cooperativeoutage compensation in cloud-RANs based 5G public safetynetworkrdquo IEEE Access vol 5 pp 17309ndash17321 2017
[9] A Zoha A Saeed A Imran M A Imran and A Abu-DayyaldquoData-driven analytics for automated cell outage detection inSelf-Organizing Networksrdquo in Proceedings of the 2015 11th Inter-national Conference on the Design of Reliable CommunicationNetworks DRCN 2015 pp 203ndash210 usa March 2015
[10] I De-La-Bandera R Barco P Munoz and I Serrano ldquoCellOutage Detection Based onHandover Statisticsrdquo IEEE Commu-nications Letters vol 19 no 7 pp 1189ndash1192 2015
[11] W Wang Q Liao and Q Zhang ldquoCOD A cooperativecell outage detection architecture for self-organizing femtocellnetworksrdquo IEEE Transactions on Wireless Communications vol13 no 11 pp 6007ndash6014 2014
[12] Q Liao M Wiczanowski and S Stanczak ldquoToward cell outagedetection with composite hypothesis testingrdquo in Proceedingsof the 2012 IEEE International Conference on CommunicationsICC 2012 pp 4883ndash4887 can June 2012
Wireless Communications and Mobile Computing 11
[13] Y Ma M Peng W Xue and X Ji ldquoA dynamic affinity propa-gation clustering algorithm for cell outage detection in self-healing networksrdquo in Proceedings of the 2013 IEEE WirelessCommunications and Networking Conference WCNC 2013 pp2266ndash2270 chn April 2013
[14] M Alias N Saxena and A Roy ldquoEfficient Cell Outage De-tection in 5G HetNets Using Hidden Markov Modelrdquo IEEECommunications Letters vol 20 no 3 pp 562ndash565 2016
[15] W Xue M Peng Y Ma and H Zhang ldquoClassification-basedapproach for cell outage detection in self-healing heterogeneousnetworksrdquo in Proceedings of the 2014 IEEEWireless Communica-tions and Networking Conference WCNC 2014 pp 2822ndash2826tur April 2014
[16] J Wang N Q Phan Z Pan N Liu X You and T Deng ldquoAnimproved TCM-based approach for cell outage detection forself-healing in LTE HetNetsrdquo in Proceedings of the 83rd IEEEVehicular Technology Conference VTC Spring 2016 chn May2016
[17] T Zhang L Feng P Yu S Guo W Li and X Qiu ldquoA handoverstatistics based approach for Cell Outage Detection in self-organized Heterogeneous Networksrdquo in Proceedings of the 15thIFIPIEEE International Symposium on Integrated Network andService Management IM 2017 pp 628ndash631 prt May 2017
[18] D Lee S Zhou X Zhong Z Niu X Zhou and H ZhangldquoSpatial modeling of the traffic density in cellular networksrdquoIEEE Wireless Communications Magazine vol 21 no 1 pp 80ndash88 2014
[19] F Xu Y Li H Wang P Zhang and D Jin ldquoUnderstandingMobile Traffic Patterns of Large Scale Cellular Towers in UrbanEnvironmentrdquo IEEEACM Transactions on Networking vol 25no 2 pp 1147ndash1161 2017
[20] S Morosi P Piunti and E Del Re ldquoSleep mode managementin cellular networks a traffic based technique enabling energysavingrdquo Transactions on Emerging Telecommunications Tech-nologies vol 24 no 3 pp 331ndash341 2013
[21] W Y Szeto B Ghosh B Basu andM OrsquoMahony ldquoMultivariatetraffic forecasting technique using cell transmission model andSARIMA modelrdquo Journal of Transportation Engineering vol135 no 9 pp 658ndash667 2009
[22] P V KlaineM A Imran O Onireti and R D Souza ldquoA Surveyof Machine Learning Techniques Applied to Self OrganizingCellular Networksrdquo IEEE Communications Surveys amp Tutorials2017
[23] Z M Fadlulla F Tang B Mao and et al ldquoState-of-the-Art Deep Learning Evolving Machine Intelligence TowardTomorrowrsquos Intelligent Network Traffic Control Systemsrdquo IEEECommunications Surveys amp Tutorials vol 19 no 4 pp 2432ndash2455 2017
[24] M M Breuniq H Kriegel R T Ng and J Sander ldquoLOFidentifying density-based local outliersrdquo SIGMOD Record vol29 no 2 pp 93ndash104 2000
[25] C M Mueller M Kaschub C Blankenhorn and S WankeldquoA Cell Outage Detection Algorithm Using Neighbor Cell ListReportsrdquo in Self-Organizing Systems vol 5343 of Lecture Notesin Computer Science pp 218ndash229 Springer Berlin HeidelbergBerlin Heidelberg 2008
International Journal of
AerospaceEngineeringHindawiwwwhindawicom Volume 2018
RoboticsJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Active and Passive Electronic Components
VLSI Design
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Shock and Vibration
Hindawiwwwhindawicom Volume 2018
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawiwwwhindawicom
Volume 2018
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Control Scienceand Engineering
Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom
Journal ofEngineeringVolume 2018
SensorsJournal of
Hindawiwwwhindawicom Volume 2018
International Journal of
RotatingMachinery
Hindawiwwwhindawicom Volume 2018
Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Navigation and Observation
International Journal of
Hindawi
wwwhindawicom Volume 2018
Advances in
Multimedia
Submit your manuscripts atwwwhindawicom
6 Wireless Communications and Mobile Computing
Table 1 Cell level data
Measurements DescriptionTime Transmission time interval (TTI)Cell info Cell global identity (CGI)Cell location Longitude and latitude informationCell inHO Cell incoming handover
current temporal sequence data for instance traffic variationof one ACE or load of NodeC Temporal prediction canadopt HoltndashWinter traffic forecast [20] or S-ARIMA model[21] to construct prediction sequences which will be usefulfor abnormal traffic recognition Moreover joint temporal-spatial predication approaches deservemore attention aswell
324 Machine Learning As analyzed in [22] machine learn-ing has been used for LTELTE-A self-healing and achievesremarkable results However thesemethods aremainly basedon single data source and do not take integrated data setshown in Section 31 together To make COD more accurateand intelligent in 5Gnetworks deep learningmodel [23] suchas DRN model can be considered
4 A Use Case for 5G H-CRAN COD
According to the architecture and procedures above weintroduce a use case to achieve COD for 5G H-CRAN Herewe suppose that the outage RRH or ACE cell cannot carrytraffic anymore Typically at this point all users served bythis cell have to be offloaded to its neighbor cells This willresult in a huge number of handovers in H-CRANThe risinghandovers in neighbor cells could be easily detected bymeansof data mining So we will discuss our algorithm by using thetemporal data of inHO statistics collected from the NodeC
41 Use Case Scenario Description The scenario is shown inFigure 5 It consists of 19 ACE with 25 users under coverageAll these users are active users Each ACE has 3 cells andone RRH cell is located under each ACE cell The UEs arerandomly distributed in cells to generate handover statisticsand measurement reports
Still we mainly collect cell level information stored inNodeC as shown in Table 1
Moreover in order to make the evaluation more obviouswe will evaluate the procedure under a simulation platformdeveloped by our lab Parameter settings can be found inTable 2
In our simulation platformwe assume that rapid decreaseof cellrsquos transmit power indicates cell outage That is we canset cell outage manually to evaluate efficiency of our ap-proach Then users under the affected outage areas switch tothe neighbor cells which will increase the number of han-dovers We will use the inHO data collected from all the cellswith a machine learning methods to evaluate whether outagecells can be detected
Table 2 Simulation parameters
Simulation parameters Value (unit)ACE 19 sites with 3 cells eachRRH 1 cell per siteUser number 25 per ACEUser distribution Uniform random distributionLink direction DownlinkBS Tx power 46 dBmPath loss model Cost231-HataUE velocity 15 kmhUser distribution UniformCell selection criteria Strongest RSRP defines the target cellSimulation length 150 TTI
RRH cellACE
Figure 5 Scenario use case for 5G H-CRAN
42 COD Procedures with M-LOF Considering the situa-tions above and the advantage of LOF shown in [9] a mod-ified LOF based detector is adopted to detect the outage byextracting both spatial and temporal correlations of handoverstatistics over the neighbors of outage cells The procedure isshown in Figure 6
The COD procedure includes four steps data collectionpreprocessingmachine learning and localization and all canbe executed automatically under the control of LS-SON Inthis way the outage cell can be detected timely so we canminimize the performance degradation of the network Herewe will introduce the functions in three steps
421 Preprocessing When data collection is finished SONentities proceed to extract feature vector 119879119894 for LOF [24]
Wireless Communications and Mobile Computing 7
Cell Level Data
Filter Standardization
Database
Low-dimension
Pre-processing
Data Collection
Machine Learning
Detection
Test z scoreCalculation
Store data to database
Localization
ParameterEstimation
No
M-LOFCalculation
OutageYes
H-CRAN
COC
thresholdZ gt
Figure 6 COD procedure with modified LOF
corresponding to each temporal data Since the inHO data iscollected from different cells at the different time the datavector is designed as spatial-temporal series
In fact the inHO statistics of outage cells are zero sincethe KPIs in outage cell cannot be collected Moreover for onepossibility there may be a few zero-inHO cells because noactive users are within their coverage This will enhance thedifficulty of detection In this case we filter all inHO statisticsof zero Here each cell splits out an ordered time series ofinHO To finish length-wise normalization and carry outuniformization and standardization processing we slice thetime series to several subseries
For temporal analysis 5 TTIs data is put into one featurevector 119879119894 as shown in below where 119894 represents the order ofsubseries
119879119894 = inHO1 inHO2 inHO3 inHO4 inHO5 (1)
The feature vector 119879119894 utilizes exactly one cellrsquos inHO datawhich is used to find out the time when an outage happened
For spatial analysis one feature vector 119878119895 consists of 119897cell handover statistics at TTI time 119895 where 119895 represents theidentity of the cell
119878119895 = inHO1 inHO2 inHO3 inHO4 inHO5 inHO6inHO7 inHO8 inHOl
(2)
After slicing the data series linear functional transforma-tion criterion is used to normalize the data to eliminate errorscaused by nonuniform features For example the criterion of119878119895 is defined as
1198781015840119895 =119878119895 minusmin119895 (119878119895)
max119895 (119878119895) minusmin119895 (119878119895) (3)
422 Machine Learning LOF detection is an unsupervisedanomaly detection algorithm The local densities of targetpoints are calculated by LOF used to compare with itsneighbors The larger the difference between the sampleand its neighbors is the larger outage factor score will beassigned to the sampleThe advantage of LOF is that no priorknowledge is required in advance to detect the unknowncell outage The detailed definitions of LOF are explained asfollows
Definition 1 k-distance of cell 119901 denoted as 119889119896(119901) representsthe distance between cell 119901 and its 119896th nearest neighbor sothat for at least 119896 cell 119900119894 satisfies
119889 (119900119894 119901) le 119889 (119900119896 119901) (4)
and for at most 119896 minus 1 cells 119900119894 satisfies119889 (119900119894 119901) lt 119889 (119900119896 119901) (5)
Definition 2 119896-distance neighborhood of cell 119901 denoted as119873119889119896(119901)(119901) when 119889119896(119901)is given 119873119889119896(119901)(119901) is the set whichincludes every cell whose distance to cell 119901 is smaller than119889119896(119901)Definition 3 Reachability distance of cell 119901 with respect tocell 119900 is denoted as rd119896(119901 119900) it is the maximum one of 119889119896(119901)and the distance between 119901 and 119900
rd119896 (119901 119900) = max 119889119896 (119900) 119889 (119900 119901) (6)
Definition 4 Local reachability density of cell p denoted aslrd119896(119901) is shown as follows
lrd119896 (119901) =10038161003816100381610038161003816119873119889119896(119901) (119901)
10038161003816100381610038161003816sum119900isin119873119889119896(119901)(119901) rd119896 (119901 o)
(7)
Definition 5 Local outlier factor of cell p denoted asLOF119896(119901) can be obtained as below
LOF119896 (119901) =sum119900isin119873119889119896(119901)(119901) (lrd119896 (119900) lrd119896 (119901))10038161003816100381610038161003816119873119889119896(119901) (119901)
10038161003816100381610038161003816 (8)
The above descriptions give the definition of LOF Fromour previous work in [17] the proposed M-LOF is a properapproach for heterogeneous network so here we use themodified LOF algorithm called M-LOF to make a refinedchoice for H-CRAN Definitions of M-LOF and m-distanceare given below
8 Wireless Communications and Mobile Computing
0
0
InHo DataMDT DataAbnormal cells
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80
Cell Id
minus15minus10minus05000510152025303540455055
MLO
F Va
lue
1
23
4
5
6
7
8
9
10
11
12
13
14
1516
17
2122
24
25 27
29
30
3233
35
36
38 4143
44
4647
49
52
54
55
56
57
58
59
6062
63
64
65
66
67
69
68
7071
72
73
74
75
76
1
2
3
4
5
6
7
89
10
11
1213
14
15
1617
21
22 2425 27 29
30 32
33
35 36 38
41
43
44
46
47
49
52
54
55
56575859
6062
63
6465
66
6768
69
70717273
74
75
76
Figure 7 Spatial analysis for abnormal cells
Definition 6 119898-distance of cell 119901 given the positive integer119896 is defined as
119898minus119889119896 (119901) = 120576 + [[sum119900isin119873119889119896(119901)(119901) 119889 (119901 119900)10038161003816100381610038161003816119873119889119896(119901) (119901)
10038161003816100381610038161003816]] (9)
Here 120576 is a constant value to enhance the accuracyDefinition 7 119898-distance neighborhood of cell p denotedas 119873119898minus119889119896(119901)(119901) is the set which includes every cell whichdistance to cell 119901 is smaller thanm-distance
Definition 8 Reachability distance of m-distance of cell 119901concerning cell o denoted as 119903minus119889119898(119901 119900) is
119903minus119889119898 (119901 119900) = max 119898minus119889119896 (119900) 119889 (119900 119901) (10)
Definition 9 Local reachability density of cell p is defined as
lrd119898 (119901) =10038161003816100381610038161003816119873119898minus119889119896(119901) (119901)
10038161003816100381610038161003816sum119900isin119873119898minus119889119896(119901)(119901) 119903minus119889119898 (119901 o)
(11)
Definition 10 Modified local outlier factor of cell p is shownbelow
LOF119898 (119901) =sum119900isin119873119898minus119889119896(119901)(119901) (lrd119898 (119900) lrd119898 (119901))10038161003816100381610038161003816119873119898minus119889119896(119901) (119901)
10038161003816100381610038161003816 (12)
According to the definition of LOF119896(119901) and LOF119898(119901) thechoice of 119896 is exactly sensitive As a consequence if there is aprior experience we will choose the cross-validation methodto estimate the parameter k
423 Localization The last step is localization based on theoutput of LOF calculation Here the neighbor cell list [25] isused to search the relation between the outage cell and itsneighbor cells by geographic informationwith 119911-score In thisway the outage cells may be localized
43 Results Analysis Based on the above steps M-LOFwould detect anomalous behaviors of handover And in thesimulation the value of 119896 for the LOF based detector is foundbetween 5 and 14
For the purpose of validation we will first analyze theresults from the spatial and temporal perspectives and thendetermine outage cell locations with neighbor cell list andperformance comparison between M-LOF and LOF will begiven at last
431 Spatial Analysis For spatial analysis we focus on thetemporal data of different cells at the same time periodFigure 7 shows the M-LOF of each cell at 95th TTI It can beseen that factor values derived from six cellsrsquo inHO data arefar higher than the normal value which is usually less than15 Therefore the abnormal cells can be distinguished fromnormal cells as they have experienced many user reconnec-tions
For comparison Figure 7 also shows the detection resultusingMDTmeasurements (with RSRP and SINR) as the datasource It can be seen that the abnormal cells are hard to bedetected using MDT Because M-LOF values of cells basedon MDT data are smoother than inHO data In contrastour proposed method using the inHO data has a betterperformance for abnormal cell detection in H-CRAN
432 Temporal Analysis Figure 8 shows the temporal analy-sis result for M-LOF variations relative to TTI for outage cell13 At this time from the results we find that M-LOF valuearises between TTI 90 and TTI 100 which is in accordancewith our outage time setting As initial connections of usersoccur between 0ndash5 TTI under unstable status so we justignore these data
433 Localization After the implementation of M-LOFbased detector neighbor cell list is used to localize the outage
Wireless Communications and Mobile Computing 9
TTI
Outage Period
22
20
18
16
14
12
10
08
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140
MLO
F Va
lue 5
1015
2025 30
35 40
4550
55
60
6570
75
80
85
90
95
100 105
110
115120
125
130
135140
Figure 8 Temporal analysis for outage TTI
cells First we need to set a threshold to filter anomalous cellsthrough 119911-score calculation as
119911119896 = 119899119896 minus 120583119899120590119899 (13)
where 119899119896 is the M-LOF value of kth cell and 120583119899 120590119899 arethe mean and standard deviation of M-LOF value for theanomaly scores of the other cells The reference 119911-scorethreshold is configured with a preferred value for the abnor-mal cells Here the reference 119911-score is set to 21 according tothe computing results We conducted two sets of evaluationsto make a comparison of performance of COD based oninHO data and MDT data respectively The results canidentify outage cell 18 and cell 61 which is the right one asshown in Figure 9 as LOF values of their neighbor cells arerelatively higher than other cells
434 Performance Evaluation In this part we analyzedthe M-LOF detection performance under varying trafficconditions Since the behaviors of users have a direct effect oninHO data the diagnosis process has been tested in differentscenarios by changing the User Density (UD) and UserVelocity (UV) parameters as the baseline setup To evaluatethe impact of the variations of UD and UV on M-LOFvalues different scenarios are set up by adjusting these twoparameters The Cumulative Distribution Function (CDF) oftheM-LOF values under different UV conditions is shown inFigure 10
It can be seen that for the low-velocity scenario almost80 of the M-LOF values are less than 05 However thereis a significant reduction in the M-LOF value as the UVincreases Likewise a similar behavior is observed with theincrease of UD as shown in Figure 11 The UV and UDparameters influence the distribution and spread of the inHOdata as explained earlier and consequently the value of M-LOF This leads to a low detection performance of M-LOFsince it generates an increased number of false alarms
Table 3 Performance evaluation
Result LOF value M-LOF valueFalse negative rate 3 3False positive rate 12 6
Outage cellAffected cell
800
600
400
200
Use
r Pos
ition
Y
User Position X
100 200 300 400 500 600 700 800
1
2
3
4
5
6
7
8
9
10
11
2122
25
27
29
30
32
33
64 63
62
60
1258 71
1773
16
68
36
35
14
6113
59
2456
57
74
1875
65
76
55
0
72
49
52
54
47
38 43
6615
67
70 69
41
44
46
Figure 9 Outage cell location results
At last this paper compares LOF and M-LOF by evaluat-ing False Positive Rate (FPR) and False Negative Rate (FNR)and the final results of the simulation are shown in Table 3The FNR represents the chance that an outage cell is notrecognized from the outage cells while the FPR representsthe chance that a normal cell is recognized as the outage cellfrom all normal cells
From Table 3 we can see that FPR and FNR of LOF baseddetector are 12 and 3 respectively It means the outagecell can be almost wholly detected although a small part ofthe normal cell may be determined as an anomaly The maincause of the 3FNR is that a small number of anomalous cellshave low traffic so that seldom user data cannot be collectedwhen they are in outage So these small cells are failed to bedetected However these outages have little impact on overallnetwork performance and user experience Therefore wecan say the outage cells can be detected successfully withLOF based detector approach As for M-LOF the FPR is 6which is smaller than LOFThe reason whyM-LOF has betterperformance is that M-LOF pays more attention to localdensity This is in line with the actual use of outage detectionsince the abnormal handovers caused by the neighbor outagecell are also localized
Still asM-LOF just modified the distance computation ofLOF so for time complexity it is just the same as LOF withis 119874(1198992) Here 119899 is the TTI number or cell number shown inFigure 7 or Figure 8
10 Wireless Communications and Mobile Computing
10
08
06
04
02
00minus2 minus1 0 1 2 3 4 5
0mM lt 5 6FI=CNS lt 1ms1mM lt 5 6FI=CNS lt 3ms3mM lt 5 6FI=CNS lt 5ms
M-LOF
CDF
Figure 10 CDF of M-LOF for User Velocity
10
08
06
04
02
00minus2 minus1 0 1 2 3 4 65
M-LOF
CDF
UE Density = 30 UsersCellUE Density = 20 UsersCellUE Density = 10 UsersCell
Figure 11 CDF of M-LOF value for User Density
5 Conclusions
In this paper we propose self-organized cell outage detectionarchitecture and approach for 5G H-CRAN The previousmethods using single data source are hard to detect thehidden outage such as RRHs Aiming at this problem weintroduce architecture and corresponding approach to realizeCOD for H-CRAN and still analyze the required key tech-nologies Then we evaluate the approach under a use casewhich utilizes the collected inHO statistics and adopt M-LOF to detect the anomalous behavior outage cells Resultsshow that the use case is effective However much work stillhas to be done for this architecture as we do not considerresource allocation and COC solutions still we have to takemore intelligent learning model such as deep learning into
account as well and compare the efficiency of different algo-rithms under a embodied H-CRAN network deploymentsMoreover we will focus on a specific perspective for self-healing approaches for H-CRAN and propose more practicalsolutions
Conflicts of Interest
The authors declare that they have no conflicts of interest
Acknowledgments
This research is supported by the National Science andTechnology Major Project of the Ministry of Science andTechnology of China (no 2018ZX030110004)
References
[1] P Rost A Banchs I Berberana et al ldquoMobile network archi-tecture evolution toward 5Grdquo IEEE Communications Magazinevol 54 no 5 pp 84ndash91 2016
[2] M Peng Y Sun X Li Z Mao and C Wang ldquoRecentadvances in cloud radio access networks system architectureskey techniques and open issuesrdquo IEEECommunications Surveysamp Tutorials vol 18 no 3 pp 2282ndash2308 2016
[3] M Peng Y Li Z Zhao and C Wang ldquoSystem architectureand key technologies for 5G heterogeneous cloud radio accessnetworksrdquo IEEE Network vol 29 no 2 pp 6ndash14 2015
[4] N Chen B Rong X Zhang and M Kadoch ldquoScalable andflexible massive MIMO precoding for 5G H-CRANrdquo IEEEWireless Communications Magazine vol 24 no 1 pp 46ndash522017
[5] OGAliu A ImranMA Imran andB Evans ldquoA survey of selforganisation in future cellular networksrdquo IEEECommunicationsSurveys amp Tutorials vol 15 no 1 pp 336ndash361 2013
[6] E J Khatib R Barco P Munoz I D La Bandera and ISerrano ldquoSelf-healing in mobile networks with big datardquo IEEECommunications Magazine vol 54 no 1 pp 114ndash120 2016
[7] M Amirijoo L Jorguseski T Kurner et al ldquoCell outagemanagement in LTE networksrdquo in Proceedings of the 2009 6thInternational Symposium on Wireless Communication Systems(ISWCS 2009) pp 600ndash604 Siena Italy September 2009
[8] M YinW Li L Feng P Yu and X Qiu ldquoMulti-cell cooperativeoutage compensation in cloud-RANs based 5G public safetynetworkrdquo IEEE Access vol 5 pp 17309ndash17321 2017
[9] A Zoha A Saeed A Imran M A Imran and A Abu-DayyaldquoData-driven analytics for automated cell outage detection inSelf-Organizing Networksrdquo in Proceedings of the 2015 11th Inter-national Conference on the Design of Reliable CommunicationNetworks DRCN 2015 pp 203ndash210 usa March 2015
[10] I De-La-Bandera R Barco P Munoz and I Serrano ldquoCellOutage Detection Based onHandover Statisticsrdquo IEEE Commu-nications Letters vol 19 no 7 pp 1189ndash1192 2015
[11] W Wang Q Liao and Q Zhang ldquoCOD A cooperativecell outage detection architecture for self-organizing femtocellnetworksrdquo IEEE Transactions on Wireless Communications vol13 no 11 pp 6007ndash6014 2014
[12] Q Liao M Wiczanowski and S Stanczak ldquoToward cell outagedetection with composite hypothesis testingrdquo in Proceedingsof the 2012 IEEE International Conference on CommunicationsICC 2012 pp 4883ndash4887 can June 2012
Wireless Communications and Mobile Computing 11
[13] Y Ma M Peng W Xue and X Ji ldquoA dynamic affinity propa-gation clustering algorithm for cell outage detection in self-healing networksrdquo in Proceedings of the 2013 IEEE WirelessCommunications and Networking Conference WCNC 2013 pp2266ndash2270 chn April 2013
[14] M Alias N Saxena and A Roy ldquoEfficient Cell Outage De-tection in 5G HetNets Using Hidden Markov Modelrdquo IEEECommunications Letters vol 20 no 3 pp 562ndash565 2016
[15] W Xue M Peng Y Ma and H Zhang ldquoClassification-basedapproach for cell outage detection in self-healing heterogeneousnetworksrdquo in Proceedings of the 2014 IEEEWireless Communica-tions and Networking Conference WCNC 2014 pp 2822ndash2826tur April 2014
[16] J Wang N Q Phan Z Pan N Liu X You and T Deng ldquoAnimproved TCM-based approach for cell outage detection forself-healing in LTE HetNetsrdquo in Proceedings of the 83rd IEEEVehicular Technology Conference VTC Spring 2016 chn May2016
[17] T Zhang L Feng P Yu S Guo W Li and X Qiu ldquoA handoverstatistics based approach for Cell Outage Detection in self-organized Heterogeneous Networksrdquo in Proceedings of the 15thIFIPIEEE International Symposium on Integrated Network andService Management IM 2017 pp 628ndash631 prt May 2017
[18] D Lee S Zhou X Zhong Z Niu X Zhou and H ZhangldquoSpatial modeling of the traffic density in cellular networksrdquoIEEE Wireless Communications Magazine vol 21 no 1 pp 80ndash88 2014
[19] F Xu Y Li H Wang P Zhang and D Jin ldquoUnderstandingMobile Traffic Patterns of Large Scale Cellular Towers in UrbanEnvironmentrdquo IEEEACM Transactions on Networking vol 25no 2 pp 1147ndash1161 2017
[20] S Morosi P Piunti and E Del Re ldquoSleep mode managementin cellular networks a traffic based technique enabling energysavingrdquo Transactions on Emerging Telecommunications Tech-nologies vol 24 no 3 pp 331ndash341 2013
[21] W Y Szeto B Ghosh B Basu andM OrsquoMahony ldquoMultivariatetraffic forecasting technique using cell transmission model andSARIMA modelrdquo Journal of Transportation Engineering vol135 no 9 pp 658ndash667 2009
[22] P V KlaineM A Imran O Onireti and R D Souza ldquoA Surveyof Machine Learning Techniques Applied to Self OrganizingCellular Networksrdquo IEEE Communications Surveys amp Tutorials2017
[23] Z M Fadlulla F Tang B Mao and et al ldquoState-of-the-Art Deep Learning Evolving Machine Intelligence TowardTomorrowrsquos Intelligent Network Traffic Control Systemsrdquo IEEECommunications Surveys amp Tutorials vol 19 no 4 pp 2432ndash2455 2017
[24] M M Breuniq H Kriegel R T Ng and J Sander ldquoLOFidentifying density-based local outliersrdquo SIGMOD Record vol29 no 2 pp 93ndash104 2000
[25] C M Mueller M Kaschub C Blankenhorn and S WankeldquoA Cell Outage Detection Algorithm Using Neighbor Cell ListReportsrdquo in Self-Organizing Systems vol 5343 of Lecture Notesin Computer Science pp 218ndash229 Springer Berlin HeidelbergBerlin Heidelberg 2008
International Journal of
AerospaceEngineeringHindawiwwwhindawicom Volume 2018
RoboticsJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Active and Passive Electronic Components
VLSI Design
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Shock and Vibration
Hindawiwwwhindawicom Volume 2018
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawiwwwhindawicom
Volume 2018
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Control Scienceand Engineering
Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom
Journal ofEngineeringVolume 2018
SensorsJournal of
Hindawiwwwhindawicom Volume 2018
International Journal of
RotatingMachinery
Hindawiwwwhindawicom Volume 2018
Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Navigation and Observation
International Journal of
Hindawi
wwwhindawicom Volume 2018
Advances in
Multimedia
Submit your manuscripts atwwwhindawicom
Wireless Communications and Mobile Computing 7
Cell Level Data
Filter Standardization
Database
Low-dimension
Pre-processing
Data Collection
Machine Learning
Detection
Test z scoreCalculation
Store data to database
Localization
ParameterEstimation
No
M-LOFCalculation
OutageYes
H-CRAN
COC
thresholdZ gt
Figure 6 COD procedure with modified LOF
corresponding to each temporal data Since the inHO data iscollected from different cells at the different time the datavector is designed as spatial-temporal series
In fact the inHO statistics of outage cells are zero sincethe KPIs in outage cell cannot be collected Moreover for onepossibility there may be a few zero-inHO cells because noactive users are within their coverage This will enhance thedifficulty of detection In this case we filter all inHO statisticsof zero Here each cell splits out an ordered time series ofinHO To finish length-wise normalization and carry outuniformization and standardization processing we slice thetime series to several subseries
For temporal analysis 5 TTIs data is put into one featurevector 119879119894 as shown in below where 119894 represents the order ofsubseries
119879119894 = inHO1 inHO2 inHO3 inHO4 inHO5 (1)
The feature vector 119879119894 utilizes exactly one cellrsquos inHO datawhich is used to find out the time when an outage happened
For spatial analysis one feature vector 119878119895 consists of 119897cell handover statistics at TTI time 119895 where 119895 represents theidentity of the cell
119878119895 = inHO1 inHO2 inHO3 inHO4 inHO5 inHO6inHO7 inHO8 inHOl
(2)
After slicing the data series linear functional transforma-tion criterion is used to normalize the data to eliminate errorscaused by nonuniform features For example the criterion of119878119895 is defined as
1198781015840119895 =119878119895 minusmin119895 (119878119895)
max119895 (119878119895) minusmin119895 (119878119895) (3)
422 Machine Learning LOF detection is an unsupervisedanomaly detection algorithm The local densities of targetpoints are calculated by LOF used to compare with itsneighbors The larger the difference between the sampleand its neighbors is the larger outage factor score will beassigned to the sampleThe advantage of LOF is that no priorknowledge is required in advance to detect the unknowncell outage The detailed definitions of LOF are explained asfollows
Definition 1 k-distance of cell 119901 denoted as 119889119896(119901) representsthe distance between cell 119901 and its 119896th nearest neighbor sothat for at least 119896 cell 119900119894 satisfies
119889 (119900119894 119901) le 119889 (119900119896 119901) (4)
and for at most 119896 minus 1 cells 119900119894 satisfies119889 (119900119894 119901) lt 119889 (119900119896 119901) (5)
Definition 2 119896-distance neighborhood of cell 119901 denoted as119873119889119896(119901)(119901) when 119889119896(119901)is given 119873119889119896(119901)(119901) is the set whichincludes every cell whose distance to cell 119901 is smaller than119889119896(119901)Definition 3 Reachability distance of cell 119901 with respect tocell 119900 is denoted as rd119896(119901 119900) it is the maximum one of 119889119896(119901)and the distance between 119901 and 119900
rd119896 (119901 119900) = max 119889119896 (119900) 119889 (119900 119901) (6)
Definition 4 Local reachability density of cell p denoted aslrd119896(119901) is shown as follows
lrd119896 (119901) =10038161003816100381610038161003816119873119889119896(119901) (119901)
10038161003816100381610038161003816sum119900isin119873119889119896(119901)(119901) rd119896 (119901 o)
(7)
Definition 5 Local outlier factor of cell p denoted asLOF119896(119901) can be obtained as below
LOF119896 (119901) =sum119900isin119873119889119896(119901)(119901) (lrd119896 (119900) lrd119896 (119901))10038161003816100381610038161003816119873119889119896(119901) (119901)
10038161003816100381610038161003816 (8)
The above descriptions give the definition of LOF Fromour previous work in [17] the proposed M-LOF is a properapproach for heterogeneous network so here we use themodified LOF algorithm called M-LOF to make a refinedchoice for H-CRAN Definitions of M-LOF and m-distanceare given below
8 Wireless Communications and Mobile Computing
0
0
InHo DataMDT DataAbnormal cells
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80
Cell Id
minus15minus10minus05000510152025303540455055
MLO
F Va
lue
1
23
4
5
6
7
8
9
10
11
12
13
14
1516
17
2122
24
25 27
29
30
3233
35
36
38 4143
44
4647
49
52
54
55
56
57
58
59
6062
63
64
65
66
67
69
68
7071
72
73
74
75
76
1
2
3
4
5
6
7
89
10
11
1213
14
15
1617
21
22 2425 27 29
30 32
33
35 36 38
41
43
44
46
47
49
52
54
55
56575859
6062
63
6465
66
6768
69
70717273
74
75
76
Figure 7 Spatial analysis for abnormal cells
Definition 6 119898-distance of cell 119901 given the positive integer119896 is defined as
119898minus119889119896 (119901) = 120576 + [[sum119900isin119873119889119896(119901)(119901) 119889 (119901 119900)10038161003816100381610038161003816119873119889119896(119901) (119901)
10038161003816100381610038161003816]] (9)
Here 120576 is a constant value to enhance the accuracyDefinition 7 119898-distance neighborhood of cell p denotedas 119873119898minus119889119896(119901)(119901) is the set which includes every cell whichdistance to cell 119901 is smaller thanm-distance
Definition 8 Reachability distance of m-distance of cell 119901concerning cell o denoted as 119903minus119889119898(119901 119900) is
119903minus119889119898 (119901 119900) = max 119898minus119889119896 (119900) 119889 (119900 119901) (10)
Definition 9 Local reachability density of cell p is defined as
lrd119898 (119901) =10038161003816100381610038161003816119873119898minus119889119896(119901) (119901)
10038161003816100381610038161003816sum119900isin119873119898minus119889119896(119901)(119901) 119903minus119889119898 (119901 o)
(11)
Definition 10 Modified local outlier factor of cell p is shownbelow
LOF119898 (119901) =sum119900isin119873119898minus119889119896(119901)(119901) (lrd119898 (119900) lrd119898 (119901))10038161003816100381610038161003816119873119898minus119889119896(119901) (119901)
10038161003816100381610038161003816 (12)
According to the definition of LOF119896(119901) and LOF119898(119901) thechoice of 119896 is exactly sensitive As a consequence if there is aprior experience we will choose the cross-validation methodto estimate the parameter k
423 Localization The last step is localization based on theoutput of LOF calculation Here the neighbor cell list [25] isused to search the relation between the outage cell and itsneighbor cells by geographic informationwith 119911-score In thisway the outage cells may be localized
43 Results Analysis Based on the above steps M-LOFwould detect anomalous behaviors of handover And in thesimulation the value of 119896 for the LOF based detector is foundbetween 5 and 14
For the purpose of validation we will first analyze theresults from the spatial and temporal perspectives and thendetermine outage cell locations with neighbor cell list andperformance comparison between M-LOF and LOF will begiven at last
431 Spatial Analysis For spatial analysis we focus on thetemporal data of different cells at the same time periodFigure 7 shows the M-LOF of each cell at 95th TTI It can beseen that factor values derived from six cellsrsquo inHO data arefar higher than the normal value which is usually less than15 Therefore the abnormal cells can be distinguished fromnormal cells as they have experienced many user reconnec-tions
For comparison Figure 7 also shows the detection resultusingMDTmeasurements (with RSRP and SINR) as the datasource It can be seen that the abnormal cells are hard to bedetected using MDT Because M-LOF values of cells basedon MDT data are smoother than inHO data In contrastour proposed method using the inHO data has a betterperformance for abnormal cell detection in H-CRAN
432 Temporal Analysis Figure 8 shows the temporal analy-sis result for M-LOF variations relative to TTI for outage cell13 At this time from the results we find that M-LOF valuearises between TTI 90 and TTI 100 which is in accordancewith our outage time setting As initial connections of usersoccur between 0ndash5 TTI under unstable status so we justignore these data
433 Localization After the implementation of M-LOFbased detector neighbor cell list is used to localize the outage
Wireless Communications and Mobile Computing 9
TTI
Outage Period
22
20
18
16
14
12
10
08
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140
MLO
F Va
lue 5
1015
2025 30
35 40
4550
55
60
6570
75
80
85
90
95
100 105
110
115120
125
130
135140
Figure 8 Temporal analysis for outage TTI
cells First we need to set a threshold to filter anomalous cellsthrough 119911-score calculation as
119911119896 = 119899119896 minus 120583119899120590119899 (13)
where 119899119896 is the M-LOF value of kth cell and 120583119899 120590119899 arethe mean and standard deviation of M-LOF value for theanomaly scores of the other cells The reference 119911-scorethreshold is configured with a preferred value for the abnor-mal cells Here the reference 119911-score is set to 21 according tothe computing results We conducted two sets of evaluationsto make a comparison of performance of COD based oninHO data and MDT data respectively The results canidentify outage cell 18 and cell 61 which is the right one asshown in Figure 9 as LOF values of their neighbor cells arerelatively higher than other cells
434 Performance Evaluation In this part we analyzedthe M-LOF detection performance under varying trafficconditions Since the behaviors of users have a direct effect oninHO data the diagnosis process has been tested in differentscenarios by changing the User Density (UD) and UserVelocity (UV) parameters as the baseline setup To evaluatethe impact of the variations of UD and UV on M-LOFvalues different scenarios are set up by adjusting these twoparameters The Cumulative Distribution Function (CDF) oftheM-LOF values under different UV conditions is shown inFigure 10
It can be seen that for the low-velocity scenario almost80 of the M-LOF values are less than 05 However thereis a significant reduction in the M-LOF value as the UVincreases Likewise a similar behavior is observed with theincrease of UD as shown in Figure 11 The UV and UDparameters influence the distribution and spread of the inHOdata as explained earlier and consequently the value of M-LOF This leads to a low detection performance of M-LOFsince it generates an increased number of false alarms
Table 3 Performance evaluation
Result LOF value M-LOF valueFalse negative rate 3 3False positive rate 12 6
Outage cellAffected cell
800
600
400
200
Use
r Pos
ition
Y
User Position X
100 200 300 400 500 600 700 800
1
2
3
4
5
6
7
8
9
10
11
2122
25
27
29
30
32
33
64 63
62
60
1258 71
1773
16
68
36
35
14
6113
59
2456
57
74
1875
65
76
55
0
72
49
52
54
47
38 43
6615
67
70 69
41
44
46
Figure 9 Outage cell location results
At last this paper compares LOF and M-LOF by evaluat-ing False Positive Rate (FPR) and False Negative Rate (FNR)and the final results of the simulation are shown in Table 3The FNR represents the chance that an outage cell is notrecognized from the outage cells while the FPR representsthe chance that a normal cell is recognized as the outage cellfrom all normal cells
From Table 3 we can see that FPR and FNR of LOF baseddetector are 12 and 3 respectively It means the outagecell can be almost wholly detected although a small part ofthe normal cell may be determined as an anomaly The maincause of the 3FNR is that a small number of anomalous cellshave low traffic so that seldom user data cannot be collectedwhen they are in outage So these small cells are failed to bedetected However these outages have little impact on overallnetwork performance and user experience Therefore wecan say the outage cells can be detected successfully withLOF based detector approach As for M-LOF the FPR is 6which is smaller than LOFThe reason whyM-LOF has betterperformance is that M-LOF pays more attention to localdensity This is in line with the actual use of outage detectionsince the abnormal handovers caused by the neighbor outagecell are also localized
Still asM-LOF just modified the distance computation ofLOF so for time complexity it is just the same as LOF withis 119874(1198992) Here 119899 is the TTI number or cell number shown inFigure 7 or Figure 8
10 Wireless Communications and Mobile Computing
10
08
06
04
02
00minus2 minus1 0 1 2 3 4 5
0mM lt 5 6FI=CNS lt 1ms1mM lt 5 6FI=CNS lt 3ms3mM lt 5 6FI=CNS lt 5ms
M-LOF
CDF
Figure 10 CDF of M-LOF for User Velocity
10
08
06
04
02
00minus2 minus1 0 1 2 3 4 65
M-LOF
CDF
UE Density = 30 UsersCellUE Density = 20 UsersCellUE Density = 10 UsersCell
Figure 11 CDF of M-LOF value for User Density
5 Conclusions
In this paper we propose self-organized cell outage detectionarchitecture and approach for 5G H-CRAN The previousmethods using single data source are hard to detect thehidden outage such as RRHs Aiming at this problem weintroduce architecture and corresponding approach to realizeCOD for H-CRAN and still analyze the required key tech-nologies Then we evaluate the approach under a use casewhich utilizes the collected inHO statistics and adopt M-LOF to detect the anomalous behavior outage cells Resultsshow that the use case is effective However much work stillhas to be done for this architecture as we do not considerresource allocation and COC solutions still we have to takemore intelligent learning model such as deep learning into
account as well and compare the efficiency of different algo-rithms under a embodied H-CRAN network deploymentsMoreover we will focus on a specific perspective for self-healing approaches for H-CRAN and propose more practicalsolutions
Conflicts of Interest
The authors declare that they have no conflicts of interest
Acknowledgments
This research is supported by the National Science andTechnology Major Project of the Ministry of Science andTechnology of China (no 2018ZX030110004)
References
[1] P Rost A Banchs I Berberana et al ldquoMobile network archi-tecture evolution toward 5Grdquo IEEE Communications Magazinevol 54 no 5 pp 84ndash91 2016
[2] M Peng Y Sun X Li Z Mao and C Wang ldquoRecentadvances in cloud radio access networks system architectureskey techniques and open issuesrdquo IEEECommunications Surveysamp Tutorials vol 18 no 3 pp 2282ndash2308 2016
[3] M Peng Y Li Z Zhao and C Wang ldquoSystem architectureand key technologies for 5G heterogeneous cloud radio accessnetworksrdquo IEEE Network vol 29 no 2 pp 6ndash14 2015
[4] N Chen B Rong X Zhang and M Kadoch ldquoScalable andflexible massive MIMO precoding for 5G H-CRANrdquo IEEEWireless Communications Magazine vol 24 no 1 pp 46ndash522017
[5] OGAliu A ImranMA Imran andB Evans ldquoA survey of selforganisation in future cellular networksrdquo IEEECommunicationsSurveys amp Tutorials vol 15 no 1 pp 336ndash361 2013
[6] E J Khatib R Barco P Munoz I D La Bandera and ISerrano ldquoSelf-healing in mobile networks with big datardquo IEEECommunications Magazine vol 54 no 1 pp 114ndash120 2016
[7] M Amirijoo L Jorguseski T Kurner et al ldquoCell outagemanagement in LTE networksrdquo in Proceedings of the 2009 6thInternational Symposium on Wireless Communication Systems(ISWCS 2009) pp 600ndash604 Siena Italy September 2009
[8] M YinW Li L Feng P Yu and X Qiu ldquoMulti-cell cooperativeoutage compensation in cloud-RANs based 5G public safetynetworkrdquo IEEE Access vol 5 pp 17309ndash17321 2017
[9] A Zoha A Saeed A Imran M A Imran and A Abu-DayyaldquoData-driven analytics for automated cell outage detection inSelf-Organizing Networksrdquo in Proceedings of the 2015 11th Inter-national Conference on the Design of Reliable CommunicationNetworks DRCN 2015 pp 203ndash210 usa March 2015
[10] I De-La-Bandera R Barco P Munoz and I Serrano ldquoCellOutage Detection Based onHandover Statisticsrdquo IEEE Commu-nications Letters vol 19 no 7 pp 1189ndash1192 2015
[11] W Wang Q Liao and Q Zhang ldquoCOD A cooperativecell outage detection architecture for self-organizing femtocellnetworksrdquo IEEE Transactions on Wireless Communications vol13 no 11 pp 6007ndash6014 2014
[12] Q Liao M Wiczanowski and S Stanczak ldquoToward cell outagedetection with composite hypothesis testingrdquo in Proceedingsof the 2012 IEEE International Conference on CommunicationsICC 2012 pp 4883ndash4887 can June 2012
Wireless Communications and Mobile Computing 11
[13] Y Ma M Peng W Xue and X Ji ldquoA dynamic affinity propa-gation clustering algorithm for cell outage detection in self-healing networksrdquo in Proceedings of the 2013 IEEE WirelessCommunications and Networking Conference WCNC 2013 pp2266ndash2270 chn April 2013
[14] M Alias N Saxena and A Roy ldquoEfficient Cell Outage De-tection in 5G HetNets Using Hidden Markov Modelrdquo IEEECommunications Letters vol 20 no 3 pp 562ndash565 2016
[15] W Xue M Peng Y Ma and H Zhang ldquoClassification-basedapproach for cell outage detection in self-healing heterogeneousnetworksrdquo in Proceedings of the 2014 IEEEWireless Communica-tions and Networking Conference WCNC 2014 pp 2822ndash2826tur April 2014
[16] J Wang N Q Phan Z Pan N Liu X You and T Deng ldquoAnimproved TCM-based approach for cell outage detection forself-healing in LTE HetNetsrdquo in Proceedings of the 83rd IEEEVehicular Technology Conference VTC Spring 2016 chn May2016
[17] T Zhang L Feng P Yu S Guo W Li and X Qiu ldquoA handoverstatistics based approach for Cell Outage Detection in self-organized Heterogeneous Networksrdquo in Proceedings of the 15thIFIPIEEE International Symposium on Integrated Network andService Management IM 2017 pp 628ndash631 prt May 2017
[18] D Lee S Zhou X Zhong Z Niu X Zhou and H ZhangldquoSpatial modeling of the traffic density in cellular networksrdquoIEEE Wireless Communications Magazine vol 21 no 1 pp 80ndash88 2014
[19] F Xu Y Li H Wang P Zhang and D Jin ldquoUnderstandingMobile Traffic Patterns of Large Scale Cellular Towers in UrbanEnvironmentrdquo IEEEACM Transactions on Networking vol 25no 2 pp 1147ndash1161 2017
[20] S Morosi P Piunti and E Del Re ldquoSleep mode managementin cellular networks a traffic based technique enabling energysavingrdquo Transactions on Emerging Telecommunications Tech-nologies vol 24 no 3 pp 331ndash341 2013
[21] W Y Szeto B Ghosh B Basu andM OrsquoMahony ldquoMultivariatetraffic forecasting technique using cell transmission model andSARIMA modelrdquo Journal of Transportation Engineering vol135 no 9 pp 658ndash667 2009
[22] P V KlaineM A Imran O Onireti and R D Souza ldquoA Surveyof Machine Learning Techniques Applied to Self OrganizingCellular Networksrdquo IEEE Communications Surveys amp Tutorials2017
[23] Z M Fadlulla F Tang B Mao and et al ldquoState-of-the-Art Deep Learning Evolving Machine Intelligence TowardTomorrowrsquos Intelligent Network Traffic Control Systemsrdquo IEEECommunications Surveys amp Tutorials vol 19 no 4 pp 2432ndash2455 2017
[24] M M Breuniq H Kriegel R T Ng and J Sander ldquoLOFidentifying density-based local outliersrdquo SIGMOD Record vol29 no 2 pp 93ndash104 2000
[25] C M Mueller M Kaschub C Blankenhorn and S WankeldquoA Cell Outage Detection Algorithm Using Neighbor Cell ListReportsrdquo in Self-Organizing Systems vol 5343 of Lecture Notesin Computer Science pp 218ndash229 Springer Berlin HeidelbergBerlin Heidelberg 2008
International Journal of
AerospaceEngineeringHindawiwwwhindawicom Volume 2018
RoboticsJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Active and Passive Electronic Components
VLSI Design
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Shock and Vibration
Hindawiwwwhindawicom Volume 2018
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawiwwwhindawicom
Volume 2018
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Control Scienceand Engineering
Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom
Journal ofEngineeringVolume 2018
SensorsJournal of
Hindawiwwwhindawicom Volume 2018
International Journal of
RotatingMachinery
Hindawiwwwhindawicom Volume 2018
Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Navigation and Observation
International Journal of
Hindawi
wwwhindawicom Volume 2018
Advances in
Multimedia
Submit your manuscripts atwwwhindawicom
8 Wireless Communications and Mobile Computing
0
0
InHo DataMDT DataAbnormal cells
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80
Cell Id
minus15minus10minus05000510152025303540455055
MLO
F Va
lue
1
23
4
5
6
7
8
9
10
11
12
13
14
1516
17
2122
24
25 27
29
30
3233
35
36
38 4143
44
4647
49
52
54
55
56
57
58
59
6062
63
64
65
66
67
69
68
7071
72
73
74
75
76
1
2
3
4
5
6
7
89
10
11
1213
14
15
1617
21
22 2425 27 29
30 32
33
35 36 38
41
43
44
46
47
49
52
54
55
56575859
6062
63
6465
66
6768
69
70717273
74
75
76
Figure 7 Spatial analysis for abnormal cells
Definition 6 119898-distance of cell 119901 given the positive integer119896 is defined as
119898minus119889119896 (119901) = 120576 + [[sum119900isin119873119889119896(119901)(119901) 119889 (119901 119900)10038161003816100381610038161003816119873119889119896(119901) (119901)
10038161003816100381610038161003816]] (9)
Here 120576 is a constant value to enhance the accuracyDefinition 7 119898-distance neighborhood of cell p denotedas 119873119898minus119889119896(119901)(119901) is the set which includes every cell whichdistance to cell 119901 is smaller thanm-distance
Definition 8 Reachability distance of m-distance of cell 119901concerning cell o denoted as 119903minus119889119898(119901 119900) is
119903minus119889119898 (119901 119900) = max 119898minus119889119896 (119900) 119889 (119900 119901) (10)
Definition 9 Local reachability density of cell p is defined as
lrd119898 (119901) =10038161003816100381610038161003816119873119898minus119889119896(119901) (119901)
10038161003816100381610038161003816sum119900isin119873119898minus119889119896(119901)(119901) 119903minus119889119898 (119901 o)
(11)
Definition 10 Modified local outlier factor of cell p is shownbelow
LOF119898 (119901) =sum119900isin119873119898minus119889119896(119901)(119901) (lrd119898 (119900) lrd119898 (119901))10038161003816100381610038161003816119873119898minus119889119896(119901) (119901)
10038161003816100381610038161003816 (12)
According to the definition of LOF119896(119901) and LOF119898(119901) thechoice of 119896 is exactly sensitive As a consequence if there is aprior experience we will choose the cross-validation methodto estimate the parameter k
423 Localization The last step is localization based on theoutput of LOF calculation Here the neighbor cell list [25] isused to search the relation between the outage cell and itsneighbor cells by geographic informationwith 119911-score In thisway the outage cells may be localized
43 Results Analysis Based on the above steps M-LOFwould detect anomalous behaviors of handover And in thesimulation the value of 119896 for the LOF based detector is foundbetween 5 and 14
For the purpose of validation we will first analyze theresults from the spatial and temporal perspectives and thendetermine outage cell locations with neighbor cell list andperformance comparison between M-LOF and LOF will begiven at last
431 Spatial Analysis For spatial analysis we focus on thetemporal data of different cells at the same time periodFigure 7 shows the M-LOF of each cell at 95th TTI It can beseen that factor values derived from six cellsrsquo inHO data arefar higher than the normal value which is usually less than15 Therefore the abnormal cells can be distinguished fromnormal cells as they have experienced many user reconnec-tions
For comparison Figure 7 also shows the detection resultusingMDTmeasurements (with RSRP and SINR) as the datasource It can be seen that the abnormal cells are hard to bedetected using MDT Because M-LOF values of cells basedon MDT data are smoother than inHO data In contrastour proposed method using the inHO data has a betterperformance for abnormal cell detection in H-CRAN
432 Temporal Analysis Figure 8 shows the temporal analy-sis result for M-LOF variations relative to TTI for outage cell13 At this time from the results we find that M-LOF valuearises between TTI 90 and TTI 100 which is in accordancewith our outage time setting As initial connections of usersoccur between 0ndash5 TTI under unstable status so we justignore these data
433 Localization After the implementation of M-LOFbased detector neighbor cell list is used to localize the outage
Wireless Communications and Mobile Computing 9
TTI
Outage Period
22
20
18
16
14
12
10
08
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140
MLO
F Va
lue 5
1015
2025 30
35 40
4550
55
60
6570
75
80
85
90
95
100 105
110
115120
125
130
135140
Figure 8 Temporal analysis for outage TTI
cells First we need to set a threshold to filter anomalous cellsthrough 119911-score calculation as
119911119896 = 119899119896 minus 120583119899120590119899 (13)
where 119899119896 is the M-LOF value of kth cell and 120583119899 120590119899 arethe mean and standard deviation of M-LOF value for theanomaly scores of the other cells The reference 119911-scorethreshold is configured with a preferred value for the abnor-mal cells Here the reference 119911-score is set to 21 according tothe computing results We conducted two sets of evaluationsto make a comparison of performance of COD based oninHO data and MDT data respectively The results canidentify outage cell 18 and cell 61 which is the right one asshown in Figure 9 as LOF values of their neighbor cells arerelatively higher than other cells
434 Performance Evaluation In this part we analyzedthe M-LOF detection performance under varying trafficconditions Since the behaviors of users have a direct effect oninHO data the diagnosis process has been tested in differentscenarios by changing the User Density (UD) and UserVelocity (UV) parameters as the baseline setup To evaluatethe impact of the variations of UD and UV on M-LOFvalues different scenarios are set up by adjusting these twoparameters The Cumulative Distribution Function (CDF) oftheM-LOF values under different UV conditions is shown inFigure 10
It can be seen that for the low-velocity scenario almost80 of the M-LOF values are less than 05 However thereis a significant reduction in the M-LOF value as the UVincreases Likewise a similar behavior is observed with theincrease of UD as shown in Figure 11 The UV and UDparameters influence the distribution and spread of the inHOdata as explained earlier and consequently the value of M-LOF This leads to a low detection performance of M-LOFsince it generates an increased number of false alarms
Table 3 Performance evaluation
Result LOF value M-LOF valueFalse negative rate 3 3False positive rate 12 6
Outage cellAffected cell
800
600
400
200
Use
r Pos
ition
Y
User Position X
100 200 300 400 500 600 700 800
1
2
3
4
5
6
7
8
9
10
11
2122
25
27
29
30
32
33
64 63
62
60
1258 71
1773
16
68
36
35
14
6113
59
2456
57
74
1875
65
76
55
0
72
49
52
54
47
38 43
6615
67
70 69
41
44
46
Figure 9 Outage cell location results
At last this paper compares LOF and M-LOF by evaluat-ing False Positive Rate (FPR) and False Negative Rate (FNR)and the final results of the simulation are shown in Table 3The FNR represents the chance that an outage cell is notrecognized from the outage cells while the FPR representsthe chance that a normal cell is recognized as the outage cellfrom all normal cells
From Table 3 we can see that FPR and FNR of LOF baseddetector are 12 and 3 respectively It means the outagecell can be almost wholly detected although a small part ofthe normal cell may be determined as an anomaly The maincause of the 3FNR is that a small number of anomalous cellshave low traffic so that seldom user data cannot be collectedwhen they are in outage So these small cells are failed to bedetected However these outages have little impact on overallnetwork performance and user experience Therefore wecan say the outage cells can be detected successfully withLOF based detector approach As for M-LOF the FPR is 6which is smaller than LOFThe reason whyM-LOF has betterperformance is that M-LOF pays more attention to localdensity This is in line with the actual use of outage detectionsince the abnormal handovers caused by the neighbor outagecell are also localized
Still asM-LOF just modified the distance computation ofLOF so for time complexity it is just the same as LOF withis 119874(1198992) Here 119899 is the TTI number or cell number shown inFigure 7 or Figure 8
10 Wireless Communications and Mobile Computing
10
08
06
04
02
00minus2 minus1 0 1 2 3 4 5
0mM lt 5 6FI=CNS lt 1ms1mM lt 5 6FI=CNS lt 3ms3mM lt 5 6FI=CNS lt 5ms
M-LOF
CDF
Figure 10 CDF of M-LOF for User Velocity
10
08
06
04
02
00minus2 minus1 0 1 2 3 4 65
M-LOF
CDF
UE Density = 30 UsersCellUE Density = 20 UsersCellUE Density = 10 UsersCell
Figure 11 CDF of M-LOF value for User Density
5 Conclusions
In this paper we propose self-organized cell outage detectionarchitecture and approach for 5G H-CRAN The previousmethods using single data source are hard to detect thehidden outage such as RRHs Aiming at this problem weintroduce architecture and corresponding approach to realizeCOD for H-CRAN and still analyze the required key tech-nologies Then we evaluate the approach under a use casewhich utilizes the collected inHO statistics and adopt M-LOF to detect the anomalous behavior outage cells Resultsshow that the use case is effective However much work stillhas to be done for this architecture as we do not considerresource allocation and COC solutions still we have to takemore intelligent learning model such as deep learning into
account as well and compare the efficiency of different algo-rithms under a embodied H-CRAN network deploymentsMoreover we will focus on a specific perspective for self-healing approaches for H-CRAN and propose more practicalsolutions
Conflicts of Interest
The authors declare that they have no conflicts of interest
Acknowledgments
This research is supported by the National Science andTechnology Major Project of the Ministry of Science andTechnology of China (no 2018ZX030110004)
References
[1] P Rost A Banchs I Berberana et al ldquoMobile network archi-tecture evolution toward 5Grdquo IEEE Communications Magazinevol 54 no 5 pp 84ndash91 2016
[2] M Peng Y Sun X Li Z Mao and C Wang ldquoRecentadvances in cloud radio access networks system architectureskey techniques and open issuesrdquo IEEECommunications Surveysamp Tutorials vol 18 no 3 pp 2282ndash2308 2016
[3] M Peng Y Li Z Zhao and C Wang ldquoSystem architectureand key technologies for 5G heterogeneous cloud radio accessnetworksrdquo IEEE Network vol 29 no 2 pp 6ndash14 2015
[4] N Chen B Rong X Zhang and M Kadoch ldquoScalable andflexible massive MIMO precoding for 5G H-CRANrdquo IEEEWireless Communications Magazine vol 24 no 1 pp 46ndash522017
[5] OGAliu A ImranMA Imran andB Evans ldquoA survey of selforganisation in future cellular networksrdquo IEEECommunicationsSurveys amp Tutorials vol 15 no 1 pp 336ndash361 2013
[6] E J Khatib R Barco P Munoz I D La Bandera and ISerrano ldquoSelf-healing in mobile networks with big datardquo IEEECommunications Magazine vol 54 no 1 pp 114ndash120 2016
[7] M Amirijoo L Jorguseski T Kurner et al ldquoCell outagemanagement in LTE networksrdquo in Proceedings of the 2009 6thInternational Symposium on Wireless Communication Systems(ISWCS 2009) pp 600ndash604 Siena Italy September 2009
[8] M YinW Li L Feng P Yu and X Qiu ldquoMulti-cell cooperativeoutage compensation in cloud-RANs based 5G public safetynetworkrdquo IEEE Access vol 5 pp 17309ndash17321 2017
[9] A Zoha A Saeed A Imran M A Imran and A Abu-DayyaldquoData-driven analytics for automated cell outage detection inSelf-Organizing Networksrdquo in Proceedings of the 2015 11th Inter-national Conference on the Design of Reliable CommunicationNetworks DRCN 2015 pp 203ndash210 usa March 2015
[10] I De-La-Bandera R Barco P Munoz and I Serrano ldquoCellOutage Detection Based onHandover Statisticsrdquo IEEE Commu-nications Letters vol 19 no 7 pp 1189ndash1192 2015
[11] W Wang Q Liao and Q Zhang ldquoCOD A cooperativecell outage detection architecture for self-organizing femtocellnetworksrdquo IEEE Transactions on Wireless Communications vol13 no 11 pp 6007ndash6014 2014
[12] Q Liao M Wiczanowski and S Stanczak ldquoToward cell outagedetection with composite hypothesis testingrdquo in Proceedingsof the 2012 IEEE International Conference on CommunicationsICC 2012 pp 4883ndash4887 can June 2012
Wireless Communications and Mobile Computing 11
[13] Y Ma M Peng W Xue and X Ji ldquoA dynamic affinity propa-gation clustering algorithm for cell outage detection in self-healing networksrdquo in Proceedings of the 2013 IEEE WirelessCommunications and Networking Conference WCNC 2013 pp2266ndash2270 chn April 2013
[14] M Alias N Saxena and A Roy ldquoEfficient Cell Outage De-tection in 5G HetNets Using Hidden Markov Modelrdquo IEEECommunications Letters vol 20 no 3 pp 562ndash565 2016
[15] W Xue M Peng Y Ma and H Zhang ldquoClassification-basedapproach for cell outage detection in self-healing heterogeneousnetworksrdquo in Proceedings of the 2014 IEEEWireless Communica-tions and Networking Conference WCNC 2014 pp 2822ndash2826tur April 2014
[16] J Wang N Q Phan Z Pan N Liu X You and T Deng ldquoAnimproved TCM-based approach for cell outage detection forself-healing in LTE HetNetsrdquo in Proceedings of the 83rd IEEEVehicular Technology Conference VTC Spring 2016 chn May2016
[17] T Zhang L Feng P Yu S Guo W Li and X Qiu ldquoA handoverstatistics based approach for Cell Outage Detection in self-organized Heterogeneous Networksrdquo in Proceedings of the 15thIFIPIEEE International Symposium on Integrated Network andService Management IM 2017 pp 628ndash631 prt May 2017
[18] D Lee S Zhou X Zhong Z Niu X Zhou and H ZhangldquoSpatial modeling of the traffic density in cellular networksrdquoIEEE Wireless Communications Magazine vol 21 no 1 pp 80ndash88 2014
[19] F Xu Y Li H Wang P Zhang and D Jin ldquoUnderstandingMobile Traffic Patterns of Large Scale Cellular Towers in UrbanEnvironmentrdquo IEEEACM Transactions on Networking vol 25no 2 pp 1147ndash1161 2017
[20] S Morosi P Piunti and E Del Re ldquoSleep mode managementin cellular networks a traffic based technique enabling energysavingrdquo Transactions on Emerging Telecommunications Tech-nologies vol 24 no 3 pp 331ndash341 2013
[21] W Y Szeto B Ghosh B Basu andM OrsquoMahony ldquoMultivariatetraffic forecasting technique using cell transmission model andSARIMA modelrdquo Journal of Transportation Engineering vol135 no 9 pp 658ndash667 2009
[22] P V KlaineM A Imran O Onireti and R D Souza ldquoA Surveyof Machine Learning Techniques Applied to Self OrganizingCellular Networksrdquo IEEE Communications Surveys amp Tutorials2017
[23] Z M Fadlulla F Tang B Mao and et al ldquoState-of-the-Art Deep Learning Evolving Machine Intelligence TowardTomorrowrsquos Intelligent Network Traffic Control Systemsrdquo IEEECommunications Surveys amp Tutorials vol 19 no 4 pp 2432ndash2455 2017
[24] M M Breuniq H Kriegel R T Ng and J Sander ldquoLOFidentifying density-based local outliersrdquo SIGMOD Record vol29 no 2 pp 93ndash104 2000
[25] C M Mueller M Kaschub C Blankenhorn and S WankeldquoA Cell Outage Detection Algorithm Using Neighbor Cell ListReportsrdquo in Self-Organizing Systems vol 5343 of Lecture Notesin Computer Science pp 218ndash229 Springer Berlin HeidelbergBerlin Heidelberg 2008
International Journal of
AerospaceEngineeringHindawiwwwhindawicom Volume 2018
RoboticsJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Active and Passive Electronic Components
VLSI Design
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Shock and Vibration
Hindawiwwwhindawicom Volume 2018
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawiwwwhindawicom
Volume 2018
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Control Scienceand Engineering
Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom
Journal ofEngineeringVolume 2018
SensorsJournal of
Hindawiwwwhindawicom Volume 2018
International Journal of
RotatingMachinery
Hindawiwwwhindawicom Volume 2018
Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Navigation and Observation
International Journal of
Hindawi
wwwhindawicom Volume 2018
Advances in
Multimedia
Submit your manuscripts atwwwhindawicom
Wireless Communications and Mobile Computing 9
TTI
Outage Period
22
20
18
16
14
12
10
08
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140
MLO
F Va
lue 5
1015
2025 30
35 40
4550
55
60
6570
75
80
85
90
95
100 105
110
115120
125
130
135140
Figure 8 Temporal analysis for outage TTI
cells First we need to set a threshold to filter anomalous cellsthrough 119911-score calculation as
119911119896 = 119899119896 minus 120583119899120590119899 (13)
where 119899119896 is the M-LOF value of kth cell and 120583119899 120590119899 arethe mean and standard deviation of M-LOF value for theanomaly scores of the other cells The reference 119911-scorethreshold is configured with a preferred value for the abnor-mal cells Here the reference 119911-score is set to 21 according tothe computing results We conducted two sets of evaluationsto make a comparison of performance of COD based oninHO data and MDT data respectively The results canidentify outage cell 18 and cell 61 which is the right one asshown in Figure 9 as LOF values of their neighbor cells arerelatively higher than other cells
434 Performance Evaluation In this part we analyzedthe M-LOF detection performance under varying trafficconditions Since the behaviors of users have a direct effect oninHO data the diagnosis process has been tested in differentscenarios by changing the User Density (UD) and UserVelocity (UV) parameters as the baseline setup To evaluatethe impact of the variations of UD and UV on M-LOFvalues different scenarios are set up by adjusting these twoparameters The Cumulative Distribution Function (CDF) oftheM-LOF values under different UV conditions is shown inFigure 10
It can be seen that for the low-velocity scenario almost80 of the M-LOF values are less than 05 However thereis a significant reduction in the M-LOF value as the UVincreases Likewise a similar behavior is observed with theincrease of UD as shown in Figure 11 The UV and UDparameters influence the distribution and spread of the inHOdata as explained earlier and consequently the value of M-LOF This leads to a low detection performance of M-LOFsince it generates an increased number of false alarms
Table 3 Performance evaluation
Result LOF value M-LOF valueFalse negative rate 3 3False positive rate 12 6
Outage cellAffected cell
800
600
400
200
Use
r Pos
ition
Y
User Position X
100 200 300 400 500 600 700 800
1
2
3
4
5
6
7
8
9
10
11
2122
25
27
29
30
32
33
64 63
62
60
1258 71
1773
16
68
36
35
14
6113
59
2456
57
74
1875
65
76
55
0
72
49
52
54
47
38 43
6615
67
70 69
41
44
46
Figure 9 Outage cell location results
At last this paper compares LOF and M-LOF by evaluat-ing False Positive Rate (FPR) and False Negative Rate (FNR)and the final results of the simulation are shown in Table 3The FNR represents the chance that an outage cell is notrecognized from the outage cells while the FPR representsthe chance that a normal cell is recognized as the outage cellfrom all normal cells
From Table 3 we can see that FPR and FNR of LOF baseddetector are 12 and 3 respectively It means the outagecell can be almost wholly detected although a small part ofthe normal cell may be determined as an anomaly The maincause of the 3FNR is that a small number of anomalous cellshave low traffic so that seldom user data cannot be collectedwhen they are in outage So these small cells are failed to bedetected However these outages have little impact on overallnetwork performance and user experience Therefore wecan say the outage cells can be detected successfully withLOF based detector approach As for M-LOF the FPR is 6which is smaller than LOFThe reason whyM-LOF has betterperformance is that M-LOF pays more attention to localdensity This is in line with the actual use of outage detectionsince the abnormal handovers caused by the neighbor outagecell are also localized
Still asM-LOF just modified the distance computation ofLOF so for time complexity it is just the same as LOF withis 119874(1198992) Here 119899 is the TTI number or cell number shown inFigure 7 or Figure 8
10 Wireless Communications and Mobile Computing
10
08
06
04
02
00minus2 minus1 0 1 2 3 4 5
0mM lt 5 6FI=CNS lt 1ms1mM lt 5 6FI=CNS lt 3ms3mM lt 5 6FI=CNS lt 5ms
M-LOF
CDF
Figure 10 CDF of M-LOF for User Velocity
10
08
06
04
02
00minus2 minus1 0 1 2 3 4 65
M-LOF
CDF
UE Density = 30 UsersCellUE Density = 20 UsersCellUE Density = 10 UsersCell
Figure 11 CDF of M-LOF value for User Density
5 Conclusions
In this paper we propose self-organized cell outage detectionarchitecture and approach for 5G H-CRAN The previousmethods using single data source are hard to detect thehidden outage such as RRHs Aiming at this problem weintroduce architecture and corresponding approach to realizeCOD for H-CRAN and still analyze the required key tech-nologies Then we evaluate the approach under a use casewhich utilizes the collected inHO statistics and adopt M-LOF to detect the anomalous behavior outage cells Resultsshow that the use case is effective However much work stillhas to be done for this architecture as we do not considerresource allocation and COC solutions still we have to takemore intelligent learning model such as deep learning into
account as well and compare the efficiency of different algo-rithms under a embodied H-CRAN network deploymentsMoreover we will focus on a specific perspective for self-healing approaches for H-CRAN and propose more practicalsolutions
Conflicts of Interest
The authors declare that they have no conflicts of interest
Acknowledgments
This research is supported by the National Science andTechnology Major Project of the Ministry of Science andTechnology of China (no 2018ZX030110004)
References
[1] P Rost A Banchs I Berberana et al ldquoMobile network archi-tecture evolution toward 5Grdquo IEEE Communications Magazinevol 54 no 5 pp 84ndash91 2016
[2] M Peng Y Sun X Li Z Mao and C Wang ldquoRecentadvances in cloud radio access networks system architectureskey techniques and open issuesrdquo IEEECommunications Surveysamp Tutorials vol 18 no 3 pp 2282ndash2308 2016
[3] M Peng Y Li Z Zhao and C Wang ldquoSystem architectureand key technologies for 5G heterogeneous cloud radio accessnetworksrdquo IEEE Network vol 29 no 2 pp 6ndash14 2015
[4] N Chen B Rong X Zhang and M Kadoch ldquoScalable andflexible massive MIMO precoding for 5G H-CRANrdquo IEEEWireless Communications Magazine vol 24 no 1 pp 46ndash522017
[5] OGAliu A ImranMA Imran andB Evans ldquoA survey of selforganisation in future cellular networksrdquo IEEECommunicationsSurveys amp Tutorials vol 15 no 1 pp 336ndash361 2013
[6] E J Khatib R Barco P Munoz I D La Bandera and ISerrano ldquoSelf-healing in mobile networks with big datardquo IEEECommunications Magazine vol 54 no 1 pp 114ndash120 2016
[7] M Amirijoo L Jorguseski T Kurner et al ldquoCell outagemanagement in LTE networksrdquo in Proceedings of the 2009 6thInternational Symposium on Wireless Communication Systems(ISWCS 2009) pp 600ndash604 Siena Italy September 2009
[8] M YinW Li L Feng P Yu and X Qiu ldquoMulti-cell cooperativeoutage compensation in cloud-RANs based 5G public safetynetworkrdquo IEEE Access vol 5 pp 17309ndash17321 2017
[9] A Zoha A Saeed A Imran M A Imran and A Abu-DayyaldquoData-driven analytics for automated cell outage detection inSelf-Organizing Networksrdquo in Proceedings of the 2015 11th Inter-national Conference on the Design of Reliable CommunicationNetworks DRCN 2015 pp 203ndash210 usa March 2015
[10] I De-La-Bandera R Barco P Munoz and I Serrano ldquoCellOutage Detection Based onHandover Statisticsrdquo IEEE Commu-nications Letters vol 19 no 7 pp 1189ndash1192 2015
[11] W Wang Q Liao and Q Zhang ldquoCOD A cooperativecell outage detection architecture for self-organizing femtocellnetworksrdquo IEEE Transactions on Wireless Communications vol13 no 11 pp 6007ndash6014 2014
[12] Q Liao M Wiczanowski and S Stanczak ldquoToward cell outagedetection with composite hypothesis testingrdquo in Proceedingsof the 2012 IEEE International Conference on CommunicationsICC 2012 pp 4883ndash4887 can June 2012
Wireless Communications and Mobile Computing 11
[13] Y Ma M Peng W Xue and X Ji ldquoA dynamic affinity propa-gation clustering algorithm for cell outage detection in self-healing networksrdquo in Proceedings of the 2013 IEEE WirelessCommunications and Networking Conference WCNC 2013 pp2266ndash2270 chn April 2013
[14] M Alias N Saxena and A Roy ldquoEfficient Cell Outage De-tection in 5G HetNets Using Hidden Markov Modelrdquo IEEECommunications Letters vol 20 no 3 pp 562ndash565 2016
[15] W Xue M Peng Y Ma and H Zhang ldquoClassification-basedapproach for cell outage detection in self-healing heterogeneousnetworksrdquo in Proceedings of the 2014 IEEEWireless Communica-tions and Networking Conference WCNC 2014 pp 2822ndash2826tur April 2014
[16] J Wang N Q Phan Z Pan N Liu X You and T Deng ldquoAnimproved TCM-based approach for cell outage detection forself-healing in LTE HetNetsrdquo in Proceedings of the 83rd IEEEVehicular Technology Conference VTC Spring 2016 chn May2016
[17] T Zhang L Feng P Yu S Guo W Li and X Qiu ldquoA handoverstatistics based approach for Cell Outage Detection in self-organized Heterogeneous Networksrdquo in Proceedings of the 15thIFIPIEEE International Symposium on Integrated Network andService Management IM 2017 pp 628ndash631 prt May 2017
[18] D Lee S Zhou X Zhong Z Niu X Zhou and H ZhangldquoSpatial modeling of the traffic density in cellular networksrdquoIEEE Wireless Communications Magazine vol 21 no 1 pp 80ndash88 2014
[19] F Xu Y Li H Wang P Zhang and D Jin ldquoUnderstandingMobile Traffic Patterns of Large Scale Cellular Towers in UrbanEnvironmentrdquo IEEEACM Transactions on Networking vol 25no 2 pp 1147ndash1161 2017
[20] S Morosi P Piunti and E Del Re ldquoSleep mode managementin cellular networks a traffic based technique enabling energysavingrdquo Transactions on Emerging Telecommunications Tech-nologies vol 24 no 3 pp 331ndash341 2013
[21] W Y Szeto B Ghosh B Basu andM OrsquoMahony ldquoMultivariatetraffic forecasting technique using cell transmission model andSARIMA modelrdquo Journal of Transportation Engineering vol135 no 9 pp 658ndash667 2009
[22] P V KlaineM A Imran O Onireti and R D Souza ldquoA Surveyof Machine Learning Techniques Applied to Self OrganizingCellular Networksrdquo IEEE Communications Surveys amp Tutorials2017
[23] Z M Fadlulla F Tang B Mao and et al ldquoState-of-the-Art Deep Learning Evolving Machine Intelligence TowardTomorrowrsquos Intelligent Network Traffic Control Systemsrdquo IEEECommunications Surveys amp Tutorials vol 19 no 4 pp 2432ndash2455 2017
[24] M M Breuniq H Kriegel R T Ng and J Sander ldquoLOFidentifying density-based local outliersrdquo SIGMOD Record vol29 no 2 pp 93ndash104 2000
[25] C M Mueller M Kaschub C Blankenhorn and S WankeldquoA Cell Outage Detection Algorithm Using Neighbor Cell ListReportsrdquo in Self-Organizing Systems vol 5343 of Lecture Notesin Computer Science pp 218ndash229 Springer Berlin HeidelbergBerlin Heidelberg 2008
International Journal of
AerospaceEngineeringHindawiwwwhindawicom Volume 2018
RoboticsJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Active and Passive Electronic Components
VLSI Design
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Shock and Vibration
Hindawiwwwhindawicom Volume 2018
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawiwwwhindawicom
Volume 2018
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Control Scienceand Engineering
Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom
Journal ofEngineeringVolume 2018
SensorsJournal of
Hindawiwwwhindawicom Volume 2018
International Journal of
RotatingMachinery
Hindawiwwwhindawicom Volume 2018
Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Navigation and Observation
International Journal of
Hindawi
wwwhindawicom Volume 2018
Advances in
Multimedia
Submit your manuscripts atwwwhindawicom
10 Wireless Communications and Mobile Computing
10
08
06
04
02
00minus2 minus1 0 1 2 3 4 5
0mM lt 5 6FI=CNS lt 1ms1mM lt 5 6FI=CNS lt 3ms3mM lt 5 6FI=CNS lt 5ms
M-LOF
CDF
Figure 10 CDF of M-LOF for User Velocity
10
08
06
04
02
00minus2 minus1 0 1 2 3 4 65
M-LOF
CDF
UE Density = 30 UsersCellUE Density = 20 UsersCellUE Density = 10 UsersCell
Figure 11 CDF of M-LOF value for User Density
5 Conclusions
In this paper we propose self-organized cell outage detectionarchitecture and approach for 5G H-CRAN The previousmethods using single data source are hard to detect thehidden outage such as RRHs Aiming at this problem weintroduce architecture and corresponding approach to realizeCOD for H-CRAN and still analyze the required key tech-nologies Then we evaluate the approach under a use casewhich utilizes the collected inHO statistics and adopt M-LOF to detect the anomalous behavior outage cells Resultsshow that the use case is effective However much work stillhas to be done for this architecture as we do not considerresource allocation and COC solutions still we have to takemore intelligent learning model such as deep learning into
account as well and compare the efficiency of different algo-rithms under a embodied H-CRAN network deploymentsMoreover we will focus on a specific perspective for self-healing approaches for H-CRAN and propose more practicalsolutions
Conflicts of Interest
The authors declare that they have no conflicts of interest
Acknowledgments
This research is supported by the National Science andTechnology Major Project of the Ministry of Science andTechnology of China (no 2018ZX030110004)
References
[1] P Rost A Banchs I Berberana et al ldquoMobile network archi-tecture evolution toward 5Grdquo IEEE Communications Magazinevol 54 no 5 pp 84ndash91 2016
[2] M Peng Y Sun X Li Z Mao and C Wang ldquoRecentadvances in cloud radio access networks system architectureskey techniques and open issuesrdquo IEEECommunications Surveysamp Tutorials vol 18 no 3 pp 2282ndash2308 2016
[3] M Peng Y Li Z Zhao and C Wang ldquoSystem architectureand key technologies for 5G heterogeneous cloud radio accessnetworksrdquo IEEE Network vol 29 no 2 pp 6ndash14 2015
[4] N Chen B Rong X Zhang and M Kadoch ldquoScalable andflexible massive MIMO precoding for 5G H-CRANrdquo IEEEWireless Communications Magazine vol 24 no 1 pp 46ndash522017
[5] OGAliu A ImranMA Imran andB Evans ldquoA survey of selforganisation in future cellular networksrdquo IEEECommunicationsSurveys amp Tutorials vol 15 no 1 pp 336ndash361 2013
[6] E J Khatib R Barco P Munoz I D La Bandera and ISerrano ldquoSelf-healing in mobile networks with big datardquo IEEECommunications Magazine vol 54 no 1 pp 114ndash120 2016
[7] M Amirijoo L Jorguseski T Kurner et al ldquoCell outagemanagement in LTE networksrdquo in Proceedings of the 2009 6thInternational Symposium on Wireless Communication Systems(ISWCS 2009) pp 600ndash604 Siena Italy September 2009
[8] M YinW Li L Feng P Yu and X Qiu ldquoMulti-cell cooperativeoutage compensation in cloud-RANs based 5G public safetynetworkrdquo IEEE Access vol 5 pp 17309ndash17321 2017
[9] A Zoha A Saeed A Imran M A Imran and A Abu-DayyaldquoData-driven analytics for automated cell outage detection inSelf-Organizing Networksrdquo in Proceedings of the 2015 11th Inter-national Conference on the Design of Reliable CommunicationNetworks DRCN 2015 pp 203ndash210 usa March 2015
[10] I De-La-Bandera R Barco P Munoz and I Serrano ldquoCellOutage Detection Based onHandover Statisticsrdquo IEEE Commu-nications Letters vol 19 no 7 pp 1189ndash1192 2015
[11] W Wang Q Liao and Q Zhang ldquoCOD A cooperativecell outage detection architecture for self-organizing femtocellnetworksrdquo IEEE Transactions on Wireless Communications vol13 no 11 pp 6007ndash6014 2014
[12] Q Liao M Wiczanowski and S Stanczak ldquoToward cell outagedetection with composite hypothesis testingrdquo in Proceedingsof the 2012 IEEE International Conference on CommunicationsICC 2012 pp 4883ndash4887 can June 2012
Wireless Communications and Mobile Computing 11
[13] Y Ma M Peng W Xue and X Ji ldquoA dynamic affinity propa-gation clustering algorithm for cell outage detection in self-healing networksrdquo in Proceedings of the 2013 IEEE WirelessCommunications and Networking Conference WCNC 2013 pp2266ndash2270 chn April 2013
[14] M Alias N Saxena and A Roy ldquoEfficient Cell Outage De-tection in 5G HetNets Using Hidden Markov Modelrdquo IEEECommunications Letters vol 20 no 3 pp 562ndash565 2016
[15] W Xue M Peng Y Ma and H Zhang ldquoClassification-basedapproach for cell outage detection in self-healing heterogeneousnetworksrdquo in Proceedings of the 2014 IEEEWireless Communica-tions and Networking Conference WCNC 2014 pp 2822ndash2826tur April 2014
[16] J Wang N Q Phan Z Pan N Liu X You and T Deng ldquoAnimproved TCM-based approach for cell outage detection forself-healing in LTE HetNetsrdquo in Proceedings of the 83rd IEEEVehicular Technology Conference VTC Spring 2016 chn May2016
[17] T Zhang L Feng P Yu S Guo W Li and X Qiu ldquoA handoverstatistics based approach for Cell Outage Detection in self-organized Heterogeneous Networksrdquo in Proceedings of the 15thIFIPIEEE International Symposium on Integrated Network andService Management IM 2017 pp 628ndash631 prt May 2017
[18] D Lee S Zhou X Zhong Z Niu X Zhou and H ZhangldquoSpatial modeling of the traffic density in cellular networksrdquoIEEE Wireless Communications Magazine vol 21 no 1 pp 80ndash88 2014
[19] F Xu Y Li H Wang P Zhang and D Jin ldquoUnderstandingMobile Traffic Patterns of Large Scale Cellular Towers in UrbanEnvironmentrdquo IEEEACM Transactions on Networking vol 25no 2 pp 1147ndash1161 2017
[20] S Morosi P Piunti and E Del Re ldquoSleep mode managementin cellular networks a traffic based technique enabling energysavingrdquo Transactions on Emerging Telecommunications Tech-nologies vol 24 no 3 pp 331ndash341 2013
[21] W Y Szeto B Ghosh B Basu andM OrsquoMahony ldquoMultivariatetraffic forecasting technique using cell transmission model andSARIMA modelrdquo Journal of Transportation Engineering vol135 no 9 pp 658ndash667 2009
[22] P V KlaineM A Imran O Onireti and R D Souza ldquoA Surveyof Machine Learning Techniques Applied to Self OrganizingCellular Networksrdquo IEEE Communications Surveys amp Tutorials2017
[23] Z M Fadlulla F Tang B Mao and et al ldquoState-of-the-Art Deep Learning Evolving Machine Intelligence TowardTomorrowrsquos Intelligent Network Traffic Control Systemsrdquo IEEECommunications Surveys amp Tutorials vol 19 no 4 pp 2432ndash2455 2017
[24] M M Breuniq H Kriegel R T Ng and J Sander ldquoLOFidentifying density-based local outliersrdquo SIGMOD Record vol29 no 2 pp 93ndash104 2000
[25] C M Mueller M Kaschub C Blankenhorn and S WankeldquoA Cell Outage Detection Algorithm Using Neighbor Cell ListReportsrdquo in Self-Organizing Systems vol 5343 of Lecture Notesin Computer Science pp 218ndash229 Springer Berlin HeidelbergBerlin Heidelberg 2008
International Journal of
AerospaceEngineeringHindawiwwwhindawicom Volume 2018
RoboticsJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Active and Passive Electronic Components
VLSI Design
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Shock and Vibration
Hindawiwwwhindawicom Volume 2018
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawiwwwhindawicom
Volume 2018
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Control Scienceand Engineering
Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom
Journal ofEngineeringVolume 2018
SensorsJournal of
Hindawiwwwhindawicom Volume 2018
International Journal of
RotatingMachinery
Hindawiwwwhindawicom Volume 2018
Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Navigation and Observation
International Journal of
Hindawi
wwwhindawicom Volume 2018
Advances in
Multimedia
Submit your manuscripts atwwwhindawicom
Wireless Communications and Mobile Computing 11
[13] Y Ma M Peng W Xue and X Ji ldquoA dynamic affinity propa-gation clustering algorithm for cell outage detection in self-healing networksrdquo in Proceedings of the 2013 IEEE WirelessCommunications and Networking Conference WCNC 2013 pp2266ndash2270 chn April 2013
[14] M Alias N Saxena and A Roy ldquoEfficient Cell Outage De-tection in 5G HetNets Using Hidden Markov Modelrdquo IEEECommunications Letters vol 20 no 3 pp 562ndash565 2016
[15] W Xue M Peng Y Ma and H Zhang ldquoClassification-basedapproach for cell outage detection in self-healing heterogeneousnetworksrdquo in Proceedings of the 2014 IEEEWireless Communica-tions and Networking Conference WCNC 2014 pp 2822ndash2826tur April 2014
[16] J Wang N Q Phan Z Pan N Liu X You and T Deng ldquoAnimproved TCM-based approach for cell outage detection forself-healing in LTE HetNetsrdquo in Proceedings of the 83rd IEEEVehicular Technology Conference VTC Spring 2016 chn May2016
[17] T Zhang L Feng P Yu S Guo W Li and X Qiu ldquoA handoverstatistics based approach for Cell Outage Detection in self-organized Heterogeneous Networksrdquo in Proceedings of the 15thIFIPIEEE International Symposium on Integrated Network andService Management IM 2017 pp 628ndash631 prt May 2017
[18] D Lee S Zhou X Zhong Z Niu X Zhou and H ZhangldquoSpatial modeling of the traffic density in cellular networksrdquoIEEE Wireless Communications Magazine vol 21 no 1 pp 80ndash88 2014
[19] F Xu Y Li H Wang P Zhang and D Jin ldquoUnderstandingMobile Traffic Patterns of Large Scale Cellular Towers in UrbanEnvironmentrdquo IEEEACM Transactions on Networking vol 25no 2 pp 1147ndash1161 2017
[20] S Morosi P Piunti and E Del Re ldquoSleep mode managementin cellular networks a traffic based technique enabling energysavingrdquo Transactions on Emerging Telecommunications Tech-nologies vol 24 no 3 pp 331ndash341 2013
[21] W Y Szeto B Ghosh B Basu andM OrsquoMahony ldquoMultivariatetraffic forecasting technique using cell transmission model andSARIMA modelrdquo Journal of Transportation Engineering vol135 no 9 pp 658ndash667 2009
[22] P V KlaineM A Imran O Onireti and R D Souza ldquoA Surveyof Machine Learning Techniques Applied to Self OrganizingCellular Networksrdquo IEEE Communications Surveys amp Tutorials2017
[23] Z M Fadlulla F Tang B Mao and et al ldquoState-of-the-Art Deep Learning Evolving Machine Intelligence TowardTomorrowrsquos Intelligent Network Traffic Control Systemsrdquo IEEECommunications Surveys amp Tutorials vol 19 no 4 pp 2432ndash2455 2017
[24] M M Breuniq H Kriegel R T Ng and J Sander ldquoLOFidentifying density-based local outliersrdquo SIGMOD Record vol29 no 2 pp 93ndash104 2000
[25] C M Mueller M Kaschub C Blankenhorn and S WankeldquoA Cell Outage Detection Algorithm Using Neighbor Cell ListReportsrdquo in Self-Organizing Systems vol 5343 of Lecture Notesin Computer Science pp 218ndash229 Springer Berlin HeidelbergBerlin Heidelberg 2008
International Journal of
AerospaceEngineeringHindawiwwwhindawicom Volume 2018
RoboticsJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Active and Passive Electronic Components
VLSI Design
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Shock and Vibration
Hindawiwwwhindawicom Volume 2018
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawiwwwhindawicom
Volume 2018
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Control Scienceand Engineering
Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom
Journal ofEngineeringVolume 2018
SensorsJournal of
Hindawiwwwhindawicom Volume 2018
International Journal of
RotatingMachinery
Hindawiwwwhindawicom Volume 2018
Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Navigation and Observation
International Journal of
Hindawi
wwwhindawicom Volume 2018
Advances in
Multimedia
Submit your manuscripts atwwwhindawicom
International Journal of
AerospaceEngineeringHindawiwwwhindawicom Volume 2018
RoboticsJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Active and Passive Electronic Components
VLSI Design
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Shock and Vibration
Hindawiwwwhindawicom Volume 2018
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawiwwwhindawicom
Volume 2018
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Control Scienceand Engineering
Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom
Journal ofEngineeringVolume 2018
SensorsJournal of
Hindawiwwwhindawicom Volume 2018
International Journal of
RotatingMachinery
Hindawiwwwhindawicom Volume 2018
Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Navigation and Observation
International Journal of
Hindawi
wwwhindawicom Volume 2018
Advances in
Multimedia
Submit your manuscripts atwwwhindawicom