Cost-Efficient Sensor Deploymentin Indoor Space with Obstacles
Nara Institute of Science and Technology*Tokyo University of Science, Yamaguchi
Nanan Marc Thierry Kouakou, Keiichi Yasumoto, Shinya Yamamoto*, and Minoru Ito
2
Overview
Indoor Wireless Sensor Network (indoor WSN) Monitor/Collect various information of indoor space
Human position, temperature, humidity, illuminance, etc Application
Human activity prediction, energy-saving appliance control, security, etc
ChallengesCoverage of target 3D spaceConnectivity among sensor nodes
3
Design of Indoor WSN
Characteristics of indoor WSN Target monitoring space is three dimensional
Constraints on installing positions (cost, defiling)Ex. Easy on ceiling/wall, but not easy on floor/in the air
Many obstacles Influence on sensing and wireless communication
Requirements for indoor WSN Minimize deployment cost Guarantee full coverage and wireless connectivity
take into account shape of target space, deployment cost, influence of obstacles
4
Organization
1. Related Work2. Problem Formulation3. Deployment Algorithms4. Evaluation5. Conclusion
5
Related Work: Coverage of 2D space with Obstacles
[1] proposed a method using Delaunay triangulation First apply the contour deployment (around obstacles),
then cover the remaining space by triangles
ProblemThe deployment is only considered in 2D space, leading to some
inaccuracies when applying to 3D spaceDeployment cost depending on position is not considered
[1] Wu et al., “A Delaunay Triangulation based method for wireless sensor network deployment”, Computer Communications, 2007
Delaunay triangulation
6
Related Work:Coverage/connectivity in 3D space without obstacles
[2] Bai et al., “Full-Coverage and k-Connectivity (k=14,6) Three Dimensional Networks”, Infocom 2009[3] Bai et al., “Low-Connectivity and Full-Coverage Three Dimensional Wireless Sensor Networks”, MobiHoc 2009
[2][3] showed optimal deployment patterns guaranteeing full coverage and wireless connectivity in 3D space Several different optimal deployment patterns depending on
relationship between sensing and communication radii rs and rc
Problem Not consider influence of obstacles and position-
dependent deployment cost
7
Human Body Shadowing Problem
• No approach focusing on the indoor WSN deployment problem that takes into account the human body shadowing effects
[4][5] discussed effects of the human body and its mobility on indoor communications
[4] Klepal et al., Influence of People Shadowing on Optimal Deployment of WLAN Access Points, VTC2004-Fall.[5] Collonge et al., Influence of the human activity on wide-band characteristics of the 60 GHz indoor radio channel, IEEE Trans. Wireless Commun., 3(6), 2004.
8
Contribution of this Work
Cost-efficient deployment methods for 3D WSNs in indoor environment taking into account obstacles Coverage of 3D space with static and mobile
obstacles (human body)
9
Organization
1. Related Work2. Problem Formulation3. Deployment Algorithms4. Evaluation5. Conclusion
10
Assumptions
Sensor nodes Shape of sensing range and communication range: sphere Sensing radius rs, communication radius rc (fixed)
Target space Deployable area
Sensor can be installed Cost of each point in area
given Monitoring space
Space to be monitored Obstacles (static and mobile)
exist
Deployable area
Obstacle
Monitoring space
11
Assumptions for Obstacles
Influence on sensing Sensor can NOT sense
Information from shadow area
Influence on wireless comm. Sensors can NOT
communicate when obstacle is on the line of sight
Sensor
Sensing range
Obstacle
Shadow area
s0
s1Wireless
communication range
rc
Obstacle
12
Problem Definition
Input Target space, monitoring space Deployable area with cost of each point Sensing and communication radii rs, rc
Output Number of sensors, sensor positions
Constraints Monitoring space is k-covered Wireless connectivity between sensors
Objective Minimize overall deployment cost
This is NP-hard problem (minimum set cover)
Any point of monitoring space is covered by at least k sensors
13
Assumption for Mobile Obstacles
ms
mobile obstacle po
s
mr
s
m
mobile obstacle
ceiling
groundpos
mh
Only human body considered as mobile obstacle Represented by cylinder: radius mr, height mh
Mobile obstacle obstructs monitoring point m from some sensors sensing ranges obstructed sensors change by mobile’s position
We assume each point is affected by only one mobile obstacle at one time
XX
XX
Top view Side view
X
XSensors
14
Mobile k-Coverage Problem
Target problem for mobile obstacleDetermine the number of sensor nodes and their installing positions to achieve mobile k-coverage with the minimal deployment cost
Mobile k-Coverage A monitoring point m is mobile k-covered if for
any location of the mobile obstacle, m is k-covered
15
Organization
1. Related Work2. Problem Formulation3. Deployment Algorithms4. Evaluation5. Conclusion
16
Discretization of Problem
Complexity Modified problem still NP-hard
Discretization Deployable area Deployable points Target monitoring space Monitoring
points
Heuristic algorithms to achievea near-optimal solution in a reasonable
amount of time
17
Algorithm for Minimal Costk-Coverage (only static obstacles)
per-cost volume: how many monitoring points are covered by the deployable point per unit deployment cost
Places sensor node on the grid point with the highest per-cost volume
Repeats until all the monitoring points are sufficiently covered
per-cost volume =
Number of monitoring points covered
Deployment cost of the deployable point
0.65 0.75
0.150.350.35
0.25
0.2 0.65 0.25
0.45
0.050.250.35
0.25
0.2 0.150.55
0.050.250.35
0.25 0.40
0.1 0.05
18
Influence of the Mobile Obstacle
Vertical plane Δ tangent to the monitoring point orthogonal to the mobile
obstacle Shaded area: half-space
divided by Δ that contains the mobile obstacle Nodes in the shaded area
cannot sense the monitoring point
(Δ)
rs
shaded area
monitoring point
(Δ )
Sufficient condition for mobile k-coverage:For arbitrary position of the mobile obstacle, the half-sphere that is not in the shaded area, contains at least k sensor nodes
X X
X
X
19
Sensor Placement forMobile k-Coverage (1)
Basic Idea Consider sphere with radius: rs centered at the monitoring point Divide it into 2k equivalent portions (spherical wedges) Put one sensor in each wedge
k2
2
Spherical wedge
wedge
sensor
(Δ)
(Δ)
??
sensor(Δ)
4
obstacle
(Δ)4
Dividing into 2k wedges(k=4)
Dividing into 2k+1 wedges(k=2)
20
Sensor Placement forMobile k-Coverage (2)
Covering spherical wedge Divide sphere into 2k+2
wedges angle: , radius: rs
)1(2
2
k
covering wedge
monitoring point
sensor node(Δ)
Spherical wedge Covering spherical wedge (k=3)
4
21
Heuristic Algorithm for Minimal Cost Mobile k-Coverage
per-cost volume: for a deployable point, the number of covering wedges in which it is located per unit deployment cost
For each monitoring, compute covering spherical wedges
monitoring points
deployable points
deployed nodes
1. For each monitoring point, determine its covering wedges2. Set a node on the deployable point with the highest per-cost volume3. Repeat until each wedge contains at least one node
k=1
22
Organization
1. Related Work2. Problem Formulation3. Deployment Algorithms4. Evaluation5. Conclusion
23
Evaluation
Purpose1. Understand to what extent the deployment
cost can be reduced2. Investigate the effectiveness of the computed
deployment for obstacles
24
Evaluation on Deployment Cost
Three deployable regions region 1 (cost=1): on the ceiling region 2 (cost=5): in the “air” (h = 2m)
region 3 (cost=2): on the partition walls
Target monitoring space Horizontal plane (h = 1.5m)
Side view
floor
Top view of the indoor environment
Method # of nodes
Deployment cost
Proposed Method
14 19
Triangular lattice [6]
7 35The deployment cost is 45% smaller
[6] Bai et al., “Complete optimal deployment patterns for full-coverage and k-connectivity (k≤ 6) wireless sensor networks”, 9th ACM Mobihoc, 2007
ceiling
25
Evaluation of Mobile 3-Coverage
sinktag node
sensor node
beaco
n
(node_id, rssi)
Purpose: investigate if beacon sent by tag node is received by at least 3 sensors with sufficient RSSI for arbitrary position of user
1. The tag node broadcasts a beacon at some monitoring point
2. Sensor node which receives the beacon sends the RSSI with its ID to the sink
3. The message with (node_id, rssi) is logged with the timestamp at the sink① ②
③
user
26
Coverage and Sensing Radius
rssi0 : average RSSI of a packet sent from a ZigBee device placed at a distance 5m
rssi0 (d=5m) = -60dBm
Distance (m) 3 4 5 6RSSI value (dBm)
-56 -60 -60 -63
If a sensor node receives a beacon sent from monitoring point with RSSI greater than rssi0, then this point is covered by the node.
ZigBee DeviceRSSI measurement without obstacle
27
Monitoring Area and User Position
Target monitoring area 2.5m x 2.5m, horizontal plane at height 1m above
the floor For each target point (P1…P4), the user stands at 4
positions around the tag node at distance of 5 to 10 cm
UP1
UP2
UP3
UP4tag
node
Monitoring points User’s positions
5-10cm
28
Sensor Deployment
Installed 9 sensors based on computation result
Seminar room at NAIST
S1S2S3S4
S5
S6
S7 S8 S9
29
Result of Mobile 3-Coverage
UP1 UP2 UP3 UP4-70.0
-60.0
-50.0
-40.0
rssi0
At least 3 sensors received beacon with RSSI more than -60dbM for any point P1—P4 and any user position UP1— UP4
mobile 3-coverage is achieved
P1
-70.0
-60.0
-50.0
-40.0
rssi0
P2
S1
S2
S3
S4
S5
S6
S7
S8
S9
UP1 UP2 UP3 UP4
30
Conclusion
Cost-efficient sensor deployment method for indoor Defined problem taking into account position-
dependent installing cost and obstacle influence Devised algorithm which places one sensor in
each 1/2(k+1) spherical wedge for mobile k-coverage Evaluated mobile 3-coverage on ZigBee testbed
Future work Integrating more accurate model of radio signal
diffraction and fading effect