d21.4 v1.0 final - cordis | european commission · the research activities of the “radio...

PROPRIETARY RIGHTS STATEMENT This document contains information, which is proprietary to the NEWCOM# Consortium.

Network of Excellence

NEWCOM#

Network of Excellence in Wireless Communications#

FP7 Contract Number: 318306

WP21 – Radio interfaces for next generation wireless systems

D21.4

Consolidated results at EuWIn@CTTC

Contractual Delivery Date: October 31, 2015 Actual Delivery Date: November 20, 2015 Responsible Beneficiary: CTTC Contributing Beneficiaries: CNIT/PoliTO, AAU, Bilkent, CNRS, CTTC/UPC, PUT, UCL Estimated Person Months: 29 Dissemination Level: Public Nature: Report Version: 1.0

Ref. Ares(2015)5384546 - 26/11/2015

PROPRIETARY RIGHTS STATEMENT This document contains information, which is proprietary to the NEWCOM# Consortium.

This page is left blank intentionally

FP7 Contract Number: 318306 Deliverable ID: WP21 / D21.4

Distribution Level: Public

Document Information

Document ID: D21.4 Version Date: November 19, 2015 Total Number of Pages: 66 Abstract: The nature of this final Deliverable of WP2.1 (“Radio interfaces

for next-generation wireless systems”) is mainly descriptive and its purpose is to provide the final report on the status and the outcomes of the different Joint Research Activities (JRAs) that compose this work package.

Keywords: The main topics addressed are the need for more (efficient use of the) spectrum, channel modelling and performance evaluation of wireless systems, high data rates, and indoor location.

Authors

Full Name Beneficiary /

Organisation e-mail Role

Miquel Payaró CTTC [email protected] Editor, Contributor Guido Masera CNIT/Polito [email protected] Contributor Andrea Biroli CNIT/Polito [email protected] Contributor Jean Pierre-Barbot CNRS/ENS de

Cachan [email protected] Contributor

Sinan Gezici Bilkent [email protected] Contributor Anna Umbert CTTC/UPC [email protected] Contributor Jordi Pérez Romero CTTC/UPC [email protected] Contributor Adrian Kliks PUT [email protected] Contributor Paweł Kryszkiewicz PUT [email protected] Contributor Ferran Casadevall CTTC/UPC [email protected] Contributor Pau Closas CTTC [email protected] Contributor Ana Moragrega CTTC [email protected] Contributor Javier Arribas CTTC [email protected] Contributor Nikolaos Bartzoudis CTTC [email protected] Contributor Oriol Font-Bach CTTC [email protected] Contributor Carles Fernández-Prades CTTC [email protected] Contributor Evgenii Vinogradov UCL [email protected] Contributor Claude Oestges UCL [email protected] Contributor Bernard Fleury AAU [email protected] Contributor Troels Pedersen AAU [email protected] Contributor



Reviewers

Full Name Beneficiary / Organisation

e-mail Date

Roberto Verdone CNIT/UniBO [email protected] October 8, 2015 Marco Luise CNIT/UniPI [email protected] November 19, 2015

Version history

Issue Date of Issue Comments 0.1 July 20, 2015 Draft 0.2 October 8, 2015 Internal review 1.0 November 19, 2015 Final Version



Executive Summary

The research activities of the “Radio interfaces for next-generation wireless systems” (WP21) work packages have addressed the challenges that were identified as key in the development radio interfaces for the Future Internet and that, today more than ever, continue to be of utmost importance. In particular, the studied topics revolve around low-emission and low-energy radio interfaces, high spectral efficiency techniques, positioning features in wireless terminals and, lastly, channel measurements, channel modelling, and exploitation of radio environment databases.

The nature of the material included in this deliverable is mainly descriptive and its purpose is to provide a final report on the status and outcome of the different Joint Research Activities (JRAs) related to this WP. As it is shown along the document, there have been an important number of achievements that emerged from those activities, materialized in a number of joint papers and knowledge transfer to the industry.



Table of Contents

1. Introduction ......................................................................................................... 9 1.1 Glossary .................................................................................................................. 9 1.2 List of Joint Research Activities (JRAs) .............................................................. 9 1.3 Description of the Main WP Achievements in the Reporting Period ................. 9

2. Detailed Activity and Achieved Results ......................................................... 12 2.1 JRA #A: Enhanced NC-OFDM transmission with reduced spurious emission level 12

2.1.1 Description of activity .......................................................................................... 12 2.1.2 Relevance with the identified fundamental open issues ..................................... 12 2.1.3 Results and achievements ................................................................................. 12 2.1.4 Publications ........................................................................................................ 12

2.2 JRA #B: Practical Implementation of Polar Codes ............................................ 13 2.2.1 Description of activity .......................................................................................... 13 2.2.2 Relevance with the identified fundamental open issues ..................................... 13 2.2.3 Results and achievements ................................................................................. 13 2.2.4 Publications ........................................................................................................ 13

2.3 JRA #C: Assessment and development of multi-link channel models ........... 14 2.3.1 Description of activity .......................................................................................... 14 2.3.2 Relevance with the identified fundamental open issues ..................................... 14 2.3.3 Results and achievements ................................................................................. 14 2.3.4 Publications ........................................................................................................ 14

2.4 JRA #D: Channel models for cooperative positioning ..................................... 15 Troels Pedersen, Gerhard Steinböck, Morten L. Jakobsen, Bernard H. Fleury (AAU) .. 15 Ronald Raulefs, Wei Wang (DLR) .................................................................................. 15 Bernard Uguen (Uni. Rennes I) ...................................................................................... 15 2.4.1 Description of activity .......................................................................................... 15 2.4.2 Relevance with the identified fundamental open issues ..................................... 15 2.4.3 Results and achievements ................................................................................. 15 2.4.4 Publications ........................................................................................................ 16

2.5 JRA #E: Compressive sensing for sparse propagation channel estimation .. 16 2.5.1 Description of activity .......................................................................................... 16 2.5.2 Relevance with the identified fundamental open issues ..................................... 16 2.5.3 Results and achievements ................................................................................. 16 2.5.4 Publications ........................................................................................................ 17

2.6 JRA #F: Design and experimental validation of algorithms for active and passive indoor positioning .............................................................................................. 17

2.6.1 Description of activity .......................................................................................... 17 2.6.2 Relevance with the identified fundamental open issues ..................................... 17 2.6.3 Results and achievements ................................................................................. 18 2.6.4 Publications ........................................................................................................ 18

2.7 JRA #G: Spectrum occupation measurements and database exploitation .... 18 2.7.1 Description of activity .......................................................................................... 19 2.7.2 Relevance with the identified fundamental open issues ..................................... 19 2.7.3 Results and achievements ................................................................................. 19 2.7.4 Publications ........................................................................................................ 19

2.8 JRA #H: Impact of channel model in the performance evaluation of wireless systems ............................................................................................................................. 20

2.8.1 Description of Activity ......................................................................................... 21 2.8.2 Relevance with the identified fundamental open issues ..................................... 21 2.8.3 Results and achievements ................................................................................. 21



2.8.4 Publications ........................................................................................................ 21 2.9 JRA #I: Hybrid radio channel modeling using graphs and rays ...................... 22

Troels Pedersen, Gerhard Steinböck (AAU) .................................................................. 22 Mingming Gan, Thomas Zemen (FTW) .......................................................................... 22 Paul Meissner (TU Graz), Erik Leitinger (TU Graz), Klaus Witrisal (TU Graz) ............... 22 2.9.1 Description of Activity ......................................................................................... 22 2.9.2 Relevance with the identified fundamental open issues ..................................... 22 2.9.3 Main Results Achieved in the Reproting Period and planned activities .............. 22 2.9.4 Publications ........................................................................................................ 23

3. General Conclusions and Prospects .............................................................. 24

4. Annex I: Detailed Description of Main technical WP Achievements ........... 25 4.1 JRA #A: Enhanced NC-OFDM transmission with reduced spurious emission level 25

4.1.1 Predistortion of NC-OFDM transmitter utilizing USRP frontend ......................... 25 4.2 JRA #B: Practical Implementation of Polar Codes ............................................ 31

4.2.1 Study of algorithms and architectures for the decoding of Polar Codes ............. 31 4.3 JRA #C: Assessment and development of multi-link channel models ........... 36

4.3.1 Development of indoor multi-link channel models (UCL, UGent) ....................... 36 4.3.2 Dense multipath depolarization modeling (UCL, UNIBO) ................................... 37 4.3.3 Extended ray-tracing modeling for UWB transmissions (UCL, FTW) ................. 38

4.4 JRA #D: Channel models for cooperative positioning ..................................... 39 4.4.1 Estimation of the Human Absorption Cross Section Via Reverberation Models 39 4.4.2 Experimental Validation of the Reverberation Effect in Room Electromagnetics40 4.4.3 Bayesian Ranging for Radio Localization with and without Line-of-Sight Detection ........................................................................................................................ 40

4.5 JRA #E: Compressive sensing for sparse propagation channel estimation .. 41 4.6 JRA #F: Design and experimental validation of algorithms for active and passive indoor positioning .............................................................................................. 43

4.6.1 Testbed description ............................................................................................ 43 4.6.2 Autonomous mobile robot ................................................................................... 43 4.6.3 Positioning Payload for Indoor Positioning ......................................................... 44 4.6.4 Experiment description ....................................................................................... 46

4.7 JRA #G: Spectrum occupation measurements and database exploitation .... 48 4.7.1 New measurements ............................................................................................ 48 4.7.2 Interpolation of measurements to build the REM ............................................... 50 4.7.3 REM Graphical Interface .................................................................................... 55

4.8 JRA #H: Impact of channel model in the performance evaluation of wireless systems ............................................................................................................................. 57

4.8.1 Motivation ........................................................................................................... 57 4.8.2 Target 5G scenario ............................................................................................. 57 4.8.3 Field-measured channels ................................................................................... 58 4.8.4 Overview of the DL FBMC scheme .................................................................... 58 4.8.5 Experimental results ........................................................................................... 59

4.9 JRA #I: Hybrid radio channel modeling using graphs and rays ...................... 61 4.9.1 A Sub-band Divided Ray Tracing Algorithm Using the DPS Subspace in UWB Indoor Scenarios ............................................................................................................ 61 4.9.2 Hybrid Model For Reverberant Indoor Radio Channels Using Rays and Graphs 62 4.9.3 Modeling of outdoor-to-indoor radio channels via propagation graphs .............. 62 4.9.4 A Sub-band Divided Ray Tracing Algorithm Using the DPS Subspace in UWB Indoor Scenarios ............................................................................................................ 63



4.10 References ............................................................................................................ 63


Security: Public

1. Introduction 1.1 Glossary

ADC Analog-‐to-‐digital converter BER Bit Error Rate BP Belief Propagation BS Base Station CDMA Code Division Multiple Access CIR Channel impulse response DVB-‐T Digital Video Broadcasting – Terrestrial FPGA Field Programmable Gate Array GSM Global System for Mobile communications JRA Joint Research Action LDPC Low-‐Density Parity-‐Check LTE-‐A Long Term Evolution – Advanced NC-‐OFDM Non-‐Contiguous Orthogonal Frequency Division Multiplexing PHY Physical layer REM Radio Environmental Map RSS Received Signal Strength RX Receiver SC Successive Cancellation SNIR Signal to Noise plus Interference Ratio SNR Signal Noise Ratio TVWS TeleVision White Spaces UE User Equipment UMTS Universal Mobile Telecommunications System USRP Universal Software Radio Peripheral VLSI Very Large Scale Integration

1.2 List of Joint Research Activities (JRAs) A.- Enhanced NC-OFDM transmission with reduced spurious emission level B.- Practical implementation of polar codes C.- Assessment and development of multi-link channel models D.- Channel models for cooperative positioning E.- Compressive sensing for sparse propagation channel estimation F.- Design and experimental validation of algorithms for active and passive indoor positioning G.- Spectrum occupation measurements and database exploitation H.- Impact of channel model in the performance evaluation of wireless systems I.- Hybrid radio channel modeling using graphs and rays 1.3 Description of the Main WP Achievements in the Reporting Period

The Joint Research Activities (JRAs) described in this document address those challenges identified as key issues in the development of next generation (namely, 5G) radio interfaces, such as the need for more (efficient use of the) spectrum, channel modelling and performance evaluation of wireless systems, high data rates, and indoor location. The list of main achievements after the second year of NEWCOM# is the following:


Security: Public

A.- Enhanced NC-OFDM transmission with reduced spurious emission level It was managed to measure nonlinearity characteristics of USRP frontend. I has been used by memoryless predistortion that significantly reduced nonlinear distortion of NC-OFDM transmitter. This predistortion has been combined with NC-OFDM spectrum shaping method (developed within WP 1.3) and reported in paper accepted for ISWCS 2015. B.- Practical implementation of polar codes Two main outcomes from this JRA: 1) From the cooperation with Bilkent University, a Belief Propagation based decoder has been designed, which offers higher throughput than previous decoders implementing the Successive Cancellation algorithms. The activity is complete and a journal paper has been submitted. 2) A novel predictive successive cancellation decoding scheme has been studied, which reduces latency at high signal to noise ratios. The activity is still running, towards a VHDL synthesizable model. C.- Assessment and development of multi-link channel models Regarding the development of indoor multi-link channel models (UCL, UGgent), the first model (from UCL) was combined with room-electromagnetic theory (from UGent) to include late components. A joint UCL-UGent experimental campaign was carried out in February 2015 to estimate the channel dynamics in various indoor environments. A joint paper was presented at EuCAP 2015, and a journal paper is in preparation. D.- Channel models for cooperative positioning A channel model for the purpose of link-level simulation of ranging in wide-band channels was developed. This model, which is based on Turin's channel model, has been applied in the derivation and test of approximate MAP and MMSE ranging with unknown LOS state. A paper has been published at the ICC-ANLN workshop, and a journal paper is in preparation. E.- Compressive sensing for sparse propagation channel estimation A sparse model of the Complex Impulse Responses (CIR) had been introduced. Then, a l0-optimization had been performed and two algorithm tested: the “Iterative Hard Thresholding” (IHT) algorithm and the recently proposed CELO algorithm. The resulting analysis will be presented in a contribution to the “CAMSAP” conference. F.- Design and experimental validation of algorithms for active and passive indoor positioning As a wrap-up of the research activities developed within this JRA, a testbed of indoor positioning techniques has been developed. It consists of a mobile robotic platform equipped with different air interfaces (WiFi, Bluetooth, Zigbee) and an onboard computer implementing the different algorithms for indoor positioning. This allows repeatable trajectories and fair comparison between different approaches. G.- Spectrum occupation measurements and database exploitation 1) Regarding the work on indoor REM creation for TVWS, we have implemented a Kriging algorithm to interpolate measurements. 2) The indoor measurement results have been used as input to an algorithm that optimizes the position and transmit power of a number of small cells to be deployed inside the building, acting as secondary transmitters, ensuring that no interference is generated to TV users. (link with WP1.3) H.- Impact of channel model in the performance evaluation of wireless systems Within the GEDOMIS testbed, an experimental performance evaluation of a communication scenario that features spectrum sharing capabilities and the use of post-OFDM modulations


Security: Public

has been carried out. The experimental lab set-up was assembled so as to operate in conditions as realistic as possible via the actual real-time implementation of the involved transceivers and also via the utilization of propagation channels which have been recorded in a field measurement campaign and which are loaded in a channel emulator that also operates in real-time (via collaboration with UCL). The results of this activity have been accepted for publication the PIMRC'15 conference proceedings. I.- Hybrid radio channel modeling using graphs and rays For in-room reverberant channels ray-tracing with propagation graphs were combined to obtain a simplified, less computationally intensive hybrid model. Preliminary results and final results were submitted as a technical document and presented at the meetings of COST IC1004 on May 26th to 28th 2014 and May 5th to 7th 2015 and the final results have been submitted to IEEE Transactions on Antennas and Propagation (joint publication AAU and FTW).


Security: Public

2. Detailed Activity and Achieved Results 2.1 JRA #A: Enhanced NC-OFDM transmission with reduced spurious emission level Paweł Kryszkiewicz (PUT), Adrian Kliks (PUT), Amor Nafkha (SUPELEC), Malek Naoues (SUPELEC).

2.1.1 Description of activity The aim of this JRA is to reduce Out-of-Band (OOB) Radiation of NC-OFDM transmission. It is caused by both subcarrier spectrum sidelobes (resulting from finite length of each NC-OFDM symbol) and nonlinear distortion of NC-OFDM symbols on practical radio front-end. Although typically both problems has been addressed separately, it is essential to do both process in parallel. The nonlinear distortion can be reduced by NC-OFDM signal power variation minimization (known as Peak-to-Average Power Ratio minimization), linearization of radio front-end (by means of predistortion) or decrease of mean transmitted power. An important issue is if PAPR reduction algorithms (utilizing high computational energy) combined with highly energy efficient radio front-ends are advantageous over utilization of higher power, more linear front-ends. Both PUT (USRP) and SUPELEC (USRP and Xilinx) have appropriate hardware platforms to address these issues. 2.1.2 Relevance with the identified fundamental open issues This JRA is relevant to tasks 2.1.2 Low-energy-consumption and low-emission radio interfaces and 2.1.4 High spectrally-efficient radio interfaces. 2.1.3 Results and achievements This work has been carried in cooperation with JRA on “Advanced filtering and adaptive signal processing” (Track 1). Based on the radio platforms specification algorithms for reduction of subcarrier spectrum sidelobes has been proposed in [JRAA_1, JRAA_3]. Moreover, both subcarrier spectrum sidelobes and nonlinearity has been reduced in a single, low computationally complex algorithm [JRAA_2]. The next step was accurate measurement of USRP frontend nonlinearity. It has been modeled and predistorted as presented in [JRAA_4]. It reports that combination of spectrum shaping algorithm proposed in [JRAA_2] with predistortion allows for significant OOB radiation reduction at the output of USRP TX. The other addressed problem, i.e. energy efficiency of OFDM/NC-OFDM radio front-end with/without implementation of PAPR reduction, has been carried by SUPELEC on Xilinx boards as USRP does not allow to measure exact power consumption in FPGA. High complexity of implementation in FPGA causes this task is still ongoing. 2.1.4 Publications [JRAA_1] P. Kryszkiewicz, H. Bogucka ,"Out-of-Band Power Reduction in NC-OFDM with Optimized Cancellation Carriers Selection," Communications Letters, IEEE , vol.17, no.10, pp.1901-1904, October 2013 [JRAA_2] P. Kryszkiewicz, A. Kliks, Y. Louet " Reduction of subcarriers spectrum sidelobes and intermodulation in NC-OFDM systems " WiMOB 2013. [JRAA_3] P. Kryszkiewicz, H. Bogucka "Advanced Interference Reduction in NC-OFDM Based Cognitive Radio with Cancellation Carriers" EUSIPCO 2014. [JRAA_4] P. Kryszkiewicz, A. Kliks, H. Bogucka “Obtaining Low Out-Of-Band Emission Level of an NC-OFDM Waveform in the SDR Platform” ISWCS 2015.


Security: Public

2.2 JRA #B: Practical Implementation of Polar Codes Andrea D. G. Biroli and Guido Masera, CNIT (Politecnico di Torino), Erdal Arıkan, Bilkent University, Luis Blanco, Centre Tecnològic Telecomunicacions Catalunya.

2.2.1 Description of activity Polar are a recent class of channel codes with low-complexity encoding, decoding, and code-construction algorithms. Different algorithms have been proposed for decoding polar codes with different tradeoffs between decoding complexity and performance. This task addresses the design of efficient hardware architectures for decoding polar codes. Among the algorithms that have been studied are belief propagation decoding and successive cancellation decoding. For the belief propagation decoding, the main objective of the activity is the design of implementation architectures able to fully exploit the intrinsic high degree of parallelism of this decoding method. As for the successive cancellation, the activity aims to improve throughput over existing approaches. 2.2.2 Relevance with the identified fundamental open issues Two highly relevant activities have been related to fundamental open issues:

• To achieve efficient hardware architectures for implementing decoders for polar codes and quantify the tradeoff between performance and complexity.

• To compare polar codes with turbo and LDPC codes from an applications perspective and for different transmission channels (AWGN, LTE, ...).

2.2.3 Results and achievements Two architectures for BP decoding of polar codes have been identified. The best achieved throughput is 2 Gb/s, and the lower energy consumption is 19 pJ per decoded bit. The proposed decoders are about 50% faster than comparable BP decoders reported in the literature. Several scheduling options have been analyzed for BP decoding in terms of implementation complexity and performance. A third, successive cancellation based decoder is currently designed, with a novel approach which promises to improve throughput with respect to fast decoding approaches like SSC and ML-SSC. A further activity in cooperation with CTTC deals with the performance evaluation of polar codes on real wireless channels. Initial steps have been completed in order to set up a common simulation scenario with CTTC. The activity is still ongoing. 2.2.4 Publications A. D. G. Biroli and G. Masera, “A VLSI implementation of the belief propagation algorithm applied to the decoding of polar codes,” in EuCNC2014 - European Conference on Networks and Communications, Bologna, 23-26 June 2014.

A.D.G. Biroli, G. Masera, E. Arıkan “High-throughput belief propagation decoder architectures for polar codes” submitted to IEEE Signal Processing Letters


Security: Public

2.3 JRA #C: Assessment and development of multi-link channel models Evgenii Vinogradov (UCL), Claude Oestges (UCL), Wout Joseph (UGent), Brecht Hanssens (UGent), Vittorio Degli-Esposti (CNIT-UNIBO), Enrico Vitucci (CNIT-UNIBO), Thomas Zemen (FTW), Mingming Gan (FTW).

2.3.1 Description of activity This JRA ambitions to develop and experimentally validate accurate and computationally effective multi-link channel models applicable to cooperative and interference-limited networks. The first aim is to assess the assets and deficiencies of existing cooperative and multi-link channel models. The second and main goal is to develop and experimentally validate representations of complex multi-dimensional radio channels for cooperative communications. These models shall then be included within EuWIN for channel emulations and performance simulations. 2.3.2 Relevance with the identified fundamental open issues The multi-dimensional radio channel remains central in interference-limited scenarios. Multiple antenna systems, interference recognition and management as well as cooperation among separate network nodes are inherently multi-dimensional techniques and should always be designed with a proper knowledge not only of the channel, but also of the interference. Hence, a multi-dimensional description of the radio channel characteristics is required in order to exploit all these dimensions when designing transmission protocols. This is why some of the developed channel models will be included in the performance evaluation of JRA-H. Also, channel models will be developed in parallel to JRA-D to guarantee as much consistency as possible. 2.3.3 Results and achievements Regarding the development of channel models, three activities were carried out.

1. Development of indoor multi-link channel models (UCL, UGent): a first model was designed and published. It is further being extended to include angular proprties, as well combined with room-electromagnetic theory to include late components. This model also serves as a basis for interaction with JRA-H. A joint UCL-UGent experimental campaign was carried out in February 2014 to estimate the channel dynamics in various indoor environments. The combined ùodel, mixing both UCL and UGent models will be published in international journal conferences.

2. Dense multipath depolarization modeling (UCL, UNIBO): the polarimetric properties of so-called Dense Multipath Components have been investigated in various outdoor and indoor environments, relying on broadband measurements and Ray-Tracing simulations.

3. Extended ray-tracing modeling for UWB transmissions (UCL, FTW): an existing ray-tracing tool have been extended with UWB capability.

2.3.4 Publications E. Vinogradov, W. Joseph, C. Oestges, Modeling and simulation of fast fading channels in indoor peer-to-peer scenarios, 8th European Conference on on Antennas and Propagation, EuCAP ’14 (Den Haag, The Netherlands), April 2014.

E. Vinogradov, W. Joseph, C. Oestges, Measurement-based modeling of time-variant fading statistics in indoor peer-to-peer scenarios, IEEE Trans. Ant. Propagat., vol. 63, no. 5, pp. 2252 – 2263, May 2015.


Security: Public

E. Vitucci, F. Mani, V. Degli-Esposti, C. Oestges, Dense multipath depolarization in outdoor and indoor radio transmissions, XXXIst General Assembly of International Union of Radio Science – URSI (Beijing, China), August 2014.

M. Gan, P. Meissner, F. Mani, E. Leitinger, M. Frohle, C. Oestges, K. Witrisal, T. Zemen, Calibration of indoor UWB sub-band divided ray tracing using multiobjective simulated annealing, 2014 IEEE International Conference on Communications (ICC), pp. 4844 - 4849, July 2014.

M. Gan, P. Meissner, F. Mani, E. Leitinger, M. Frohle, C. Oestges, K. Witrisal, T. Zemen, Low-complexity sub-band divided ray tracing for UWB indoor channels, 2014 IEEE Wireless Communications and Networking Conference (WCNC), pp. 305 – 310, April 2014.

2.4 JRA #D: Channel models for cooperative positioning Troels Pedersen, Gerhard Steinböck, Morten L. Jakobsen, Bernard H. Fleury (AAU)

Ronald Raulefs, Wei Wang (DLR)

Bernard Uguen (Uni. Rennes I)

2.4.1 Description of activity The aim of this JRA is to develop and experimentally validate channel models for the complex multi-link environment considered in indoor cooperative positioning. This JRA aims at exploring the single- and multi-link aspects of radio channels for positioning purposes, and at proposing models for design and simulation. Closely connected to this is the development of estimators for empirical calibration of the models parameters and for localization parameters such as range and direction. 2.4.2 Relevance with the identified fundamental open issues This JRA has addressed the challenge by focusing on three main activities, namely reverberation models indoor channels related to multilink models, stochastic modeling of antenna responses related to localization systems, and on parameter estimation for stochastic channel models for model calibration and for ranging. 2.4.3 Results and achievements Modeling of reverberant radio channels (single- and multi-link) The reveberation models by Sabine and Eyring from room acoustic have been studied in connection with models for in-room radio links. These models allow to characterize the reverberant component. We found that Eyring’s model provides a more accurate prediction of the parameters characterizing the decaying tail, like the reverberation time, than Sabine’s model. Furthermore, we found that the considered reverberation models in combination with a newly proposed model for the delay power spectrum enables us to forecast the path loss, mean delay and rms delay spread versus distance in good agreement with experimental results. Stochastic Antenna Modelling To analyze, simulate and emulate channel variability due to user interaction effects (near body effects, antenna orientation shift due to user movement), it is necessary to include these in stochastic radio channel models. In this work, measurements of antenna responses are analyzed and modeled statistically. A first paper was published at EuCAP 2014. In continuation hereto, AAU and UR1 are working a covariance model for the frequency dependent spherical harmonics coefficients in UWB antennas. This joint work is planned to continue after Newcom# ends.


Security: Public

Parameter Estimation for Stochastic Models AAU has initiated some activities aiming at designing parameter estimators for stochastic channel models based on standard estimation theory. It is worth mentioning that no such estimator for the parameters of commonly used stochastic models, like 3GPP, Winner, etc., exist. As a result, the parameters of these models are set in an ad-hoc fashion. To this end, we have so far developed parameter estimators for Turin’s and Suzuki's models. Currently, discussions are ongoing with UR1 on a measurement campaign to test the parameter estimator for Turin’s model. 2.4.4 Publications Jing, L, Pedersen, T & Fleury, BH 2015, 'Bayesian Ranging for Radio Localization with and without Line-of-Sight Detection'. IEEE ICC 2015 Workshop on Advances in Network Localization and Navigation (ANLN). IEEE Press. Steinböck, G, Pedersen, T, Fleury, BH, Wang, W & Raulefs, R 2015, 'Estimation of the Human Absorption Cross Section Via Reverberation Models' IEEE Antennas and Wireless Propagation Letters, in review. Steinböck, G, Pedersen, T, Fleury, BH, Wang, W & Raulefs, R 2015, 'Experimental Validation of the Reverberation Effect in Room Electromagnetics' IEEE Transactions on Antennas and Propagation, vol 63, nr. 5, s. 2041-2053., 10.1109/TAP.2015.2423636 2.5 JRA #E: Compressive sensing for sparse propagation channel estimation R. Boyer (CNRS/UPS) J.P. Barbot, P. Larzabal, M. N. El Korso, T. NGuyen (CNRS/SATIE) B. Fleury, T. Pedersen (Aalborg University) 2.5.1 Description of activity Mainly due to the relative mobility of the transmitting and receiving equipments, the wireless propagation channel is time varying. Under the assumptions of linearity and time invariance during the measurements, the propagation channel is completely characterized by its Complex Impulse Response (CIR). In order to overcome this time varying property, the wireless telecommunication systems periodically estimate the CIRs. In this framework, we propose to use compressive sensing toward the estimation of the CIR. The main objective is to obtain a sparse description of the CIRs, i.e. in that case, to exploit a minimum number of complex parameters while keeping estimation performances (bias, MSE). 2.5.2 Relevance with the identified fundamental open issues This topic is related to sparse estimation of wireless complex impulse response in various propagation environments, propagation modelisation and dynamic security of wireless transmissions. 2.5.3 Results and achievements In a first step, a sparse linear model of the Complex Impulse Responses (CIR) had been proposed. In wireless communications, due to multipath effect, the signal at the receiver is a composition of scaled and delayed versions of the transmitted signal. The channel can be modeled as a finite set of K scaled delays applied to the transmitted signal. Considering that the K delays can be distributed on a grid of N > K positions TG and that the signal can be represented in discrete time domain with a sample rate of 1/TG , the problem can be written in its matricial form as y = Sh + n , where y = [y1 y2 … yM]T are the observations, h =[h1 h2 … hN]T are the unknown channel taps to be estimated and n the noise vector assumed to be a circular additive white and Gaussian noise.


Security: Public

If some l2 and l1 regularisation techniques can be exploited to inverse this linear problem, we propose to perfom the sparse identification using a l0 norm. Consequently a l0-optimization had been performed and two algorithm tested: the “Iterative Hard Thresholding” (IHT) algorithm and the recently proposed CELO algorithm. It is shown that nonconvex penalties outperform classical l1 relaxation in the case of this identification. 2.5.4 Publications A. Chinatto, E. Soubies, C. Junqueira, J. M. T. Romano, P. Larzabal, J.-P. Barbot and Laure Blanc-Féraud, l0-Optimization for Channel and DOA Sparse Estimation, IEEE Internationnal workshop CAMSAP 2015, Cancun, Mexico, December 13-16, 2015 T. H. T. Nguyen, J.-P. Barbot, Secret key management and dynamic security coding system, in Proc. IEEE ICCE'2014, Da Nang, Vietnam, July 2014 T. H. T. Nguyen, J.-P. Barbot, Dynamic secret key generation for joint error control and security coding, in Proc. IEEE WCNC'2014, Istanbul, Turkey, April 2014 T. H. T Nguyen, J.-P. Barbot, Joint error control and dynamic security coding, in International IEEE Conference on Advanced Technologies for Communications 2013 (ATC’13), pages 285–290, Ho Chi Minh City, Vietnam, October 2013. 2.6 JRA #F: Design and experimental validation of algorithms for active and passive indoor positioning Sinan Gezici, Bilkent University Davide Dardari, Giacomo Calanchi CNIT Bologna Luc Vandendorpe, Université catholique de Louvain Carles Fernández-Prades, Pau Closas, Javier Arribas, Ana Moragrega, Juan Manuel Castro Arbizu, CTTC External researchers: Héctor Torres (Universidad Tecnológica Metropolitana – Chile) 2.6.1 Description of activity In this JRA, theoretical studies are performed for range and position estimation problems. In particular, the statistical characterization of the maximum likelihood estimator is performed and fundamental limits on positioning are investigated. Also, experiments are performed on target tracking for ultra-wideband (UWB) passive localization by using Gaussian mixture probability hypothesis density (GM-PHD) filters for multiple person tracking with UWB radar sensors in an indoor environment. In addition, this JRA has also addressed different aspects and technologies for indoor location. After a survey of the current available technologies for indoor positioning, researchers have continued the development of an open, reliable and modular system for testing several technologies and collect the data measurements, as well as the development of algorithms for indoor positioning exploiting data fusion between WiFi RSS and inertial measurements. This resulted in a testbed, named OpenInLocation Lab, in order to allow researchers to perform an experimental, hands-on approach to the problem. 2.6.2 Relevance with the identified fundamental open issues Fundamental analysis is performed on the statistical characterization of the maximum likelihood estimator (MLE). After specifying upper and lower bounds and studying optimal signal design, applications to TOA estimation are considered. Also, for the indoor positioning problem, both theoretical performance limits in the presence of clock skew are derived and practical optimization theory based algorithms are proposed.


Security: Public

2.6.3 Results and achievements Theoretical results are obtained for both range estimation (via MLE estimation) and position estimation. In addition, an experimental setup consisting of a network of UWB radar sensors is designed, and a new detection algorithm is developed. The results of this experimental proof-of-concept study show the possibility of accurately tracking multiple targets using a UWB radar sensor network in indoor environments based on the proposed approach. The results are published in various journals and presented in conferences. Also a special issue is edited for EURASIP JASP named "Signal Processing Techniques for Anywhere, Anytime Positioning". The publications on "Statistics of the MLE and Approximate Upper and Lower Bounds" involve inter-track activities (collaboration with Prof. Dardari). Also these studies are awarded the NEWCOM# 2014 Best Paper Award. 2.6.4 Publications J. M. Castro-Arvizu, A. Moragrega, P. Closas, and J. A. Fernández-Rubio, “Assessment of RSS Model Calibration with Real WLAN Devices,” in Twelfth International Symposium on Wireless Communication Systems (ISWCS 2015), Brussels (Belgium), Aug. 2015. A. Mallat, S. Gezici, D. Dardari, and L. Vandendorpe, "Statistics of the MLE and Approximate Upper and Lower Bounds – Part 2: Threshold Computation and Optimal Signal Design," IEEE Transactions on Signal Processing, vol. 62, no. 21, pp. 5677-5689, Nov. 2014. A. Mallat, S. Gezici, D. Dardari, C. Craeye and L. Vandendorpe, "Statistics of the MLE and Approximate Upper and Lower Bounds – Part 1: Application to TOA Estimation," IEEE Transactions on Signal Processing, vol. 62, no. 21, pp. 5663-5676, Nov. 2014. B. Gulmezoglu, M. B. Guldogan, and S. Gezici, "Multi-person tracking with a network of ultra-wideband radar sensors based on Gaussian mixture PHD filters," IEEE Sensors Journal, vol. 15, no. 4, pp. 2227-2237, April 2015. M. Luise, C. Fernández-Prades, S. Gezici, and H. Wymeersch, "Signal processing techniques for anywhere, anytime positioning," EURASIP Journal on Advances in Signal Processing, 2014:93 doi:10.1186/1687-6180-2014-93, 2014 (editorial). M. R. Gholami, E. G. Ström, H. Wymeersch, and S. Gezici, "Upper bounds on position error of a single location estimate in wireless sensor networks," EURASIP Journal on Advances in Signal Processing, 2014:4, DOI:10.1186/1687-6180-2014-4, 2014. H. Soganci, S. Gezici, and M. B Guldogan, "Enhancements to threshold based range estimation for ultra-wideband systems," IEEE International Conference on Ultra-Wideband (ICUWB 2014), Paris, France, Sep. 1-3, 2014. M. R. Gholami, S. Gezici, H. Wymeersch, E. G. Ström, and M. Jansson, "Characterizing the worst-case position error in bearing-only target localization," 12th Workshop on Positioning, Navigation and Communication (WPNC 2015), Dresden, Germany, March 11-12, 2015 (invited). M. R. Gholami, S. Gezici, and E. G. Ström, "TDOA based positioning in the presence of unknown clock skew," IEEE Transactions on Communications, vol. 61, no. 6, pp. 2522-2534, June 2013. M. R. Gholami, S. Gezici, and E. G. Ström, "A concave-convex procedure for TDOA based positioning," IEEE Communications Letters, vol. 17, no. 4, pp. 765-768, April 2013. M. R. Gholami, H. Wymeersch, E. G. Ström, and S. Gezici, "Distributed bounding of feasible sets in cooperative wireless network positioning," IEEE Communications Letters, vol. 17, no.8, pp. 1596-1599, Aug. 2013. 2.7 JRA #G: Spectrum occupation measurements and database exploitation Adrian Kliks, Paweł Kryszkiewicz, Poznan University of Technology Jordi Pérez-Romero, Anna Umbert, Ferran Casadevall, Universitat Politècnica de Catalunya


Security: Public

2.7.1 Description of activity The purpose of this JRA is to make spectrum measurements to identify the spectrum availability for cognitive radio. Several frequency bands from 200 MHz up to 3 GHz have been measured up to date in the two sites, both Poznan and Barcelona. Then, the focus has been more on the TeleVision White Spaces (TVWS) band, identified as a candidate band to extend the capacity of cellular systems when considering small cell deployments. In this respect, measurements both in indoor and outdoor locations are being done with the objective of building a Radio Environmental Map (REM) database that can be used as a support for small cell deployment strategies in connection with JRA 1.3.3.A of WP1.3. 2.7.2 Relevance with the identified fundamental open issues As discussed in Deliverable D13.1 [1] the use of shared spectrum such as TVWS to extend the capacity in LTE and LTE-A networks has been found particularly relevant in different works [2]-[11] for small cell scenarios. However, there are actually still very few works that have addressed the problem of how to allocate TVWS spectrum in an optimized way. In this context, this activity intends to support the theoretical activities by analyzing in particular the characteristics of the TVWS band (e.g. stability, etc.) so that a REM database can be built. Moreover, the activity also addresses the design of the REM structure including the relevant parameters to be stored. 2.7.3 Results and achievements During this third year more measurements were carried out in the building in Barcelona to build a REM, characterizing the received power in multiple points for the different floors. We started also to update previous measurements due to the changes caused by the digital dividend. We developed a graphical interface to represent the REM data and to observe the spectrum occupancy and the received power at different points of the building using the Kriging algorithm for power estimation. In addition, during this third year, based on all the measurements carried out up to date, we implemented and validated a Kriging algorithm to interpolate measurements and create an indoor REM to support the deployment of small cells operating in the TVWS. The number of nodal points as well as the radius used by the Kriging algorithm have been analyzed to reduce the error. The indoor measurement results have been used as input to an algorithm that optimizes the position and transmit power of a number of small cells to be deployed inside the building, acting as secondary transmitters, ensuring that no interference is generated to TV users. Details of these results will be given in the corresponding deliverable D13.3. 2.7.4 Publications A. Kliks, P. Kryszkiewicz, J. Pérez-Romero, A. Umbert, F. Casadevall, “Spectrum occupancy in big cities–comparative study”, ISWCS 2013 conference – Workshop on Cognitive Radio Advances, Applications and Future Emerging Technologies. August, 2013, Joint paper PUT/UPC.

A. Kliks, P. Kryszkiewicz, A. Umbert, J.Pérez-Romero, F. Casadevall, “TVWS Indoor Measurements for HetNets”, IEEE WCNC 2014 - Workshop on Interference and Design Issues for Future Heterogeneous Networks, Istambul, 6th April, 2014. Joint paper PUT/UPC.

A. Kliks, P. Kryszkiewicz, K. Cichon, A. Umbert, J.Pérez-Romero, F. Casadevall, "DVB-T channels power measurements in indoor/outdoor cases", ICTF conference, May, 2014, Joint paper PUT/UPC.


Security: Public

A. Umbert, J.Pérez-Romero, F. Casadevall, A. Kliks, P. Kryszkiewicz, "On the use of Indoor Radio Environment Maps for HetNets Deployment", CROWNCOM conference, June, 2014, Joint paper PUT/UPC.

P. Kryszkiewicz, A. Kliks, J.Pérez-Romero, A. Umbert, “DVB-T signal detection for indoor environments in low-SNR regime”, TELECOMMUNICATION REVIEW AND TELECOMMUNICATION NEWS, 1-6, 2014, ISSN: 1230-3496, Joint paper PUT/UPC.

A. Kliks, P. Kryszkiewicz, K. Cichon, A. Umbert, J. Perez-Romero, F. Casadevall, “DVB-T channels measurements for the deployment of outdoor REM databases”, Journal of telecommunications and information technology(JTiT), 42-52, 2014, ISSN: 1509-4553, Joint paper PUT/UPC.

REFERENCES [1] A. Zalonis (Editor), “D13.1: Fundamental issues on energy- and bandwidth-efficient

communications and networking”, Deliverable of the Newcom #, March 2014.

[2] J. Xiao, R. Q. Hu, Y. Qian, L. Gong, B. Wang ”Expanding LTE Network Spectrum with Cognitive Radios: From Concept to Implementation”, IEEE Wireless Communications, April, 2013, pp. 13-19.

[3] J. Zander, L.K. Rasmussen, K.W. Song, P. Mähönen, M. Petrova, R. Jäntti, J. Kronander, ”On the scalability of Cognitive Radio: Assessing the Commercial Viability of Secondary Spectrum Access”, , IEEE Wireless Communications, April, 2013, pp.28-35.

[4] T. Dudda, T. Irnich, ”Capacity of cellular networks deployed in TV White Space”, IEEE International Symposium on Dynamic Spectrum Access Networks, 2012.

[5] A. Achtzehn, M. Petrova, P. Mähönen, ”On the Performance of Cellular Network Deployments in TV Whitespaces”, IEEE ICC 2012.

[6] C.F. Silva, H. Alves, A. Gomes, ”Extension of LTE Operational Mode over TV White Spaces”, Future Network and Mobile Summit, 2011.

[7] J. W. Mwangoka, P. Marques, J. Rodriguez, ”Broker Based Secondary Spectrum Trading”, 6th International ICST Conference on Cognitive Radio Oriented Wireless Networks and Communications (CROWNCOM), 2011.

[8] M. Parzy, H. Bogucka, ”Policies and technology constraints for auctions in TV White Spaces - a practical approach for LTE-A”, International Symposium on Wireless Communication Systems (ISWCS), 2012.

[9] F. Peng, Y. Gao, Y. Chen, K.K. Chai, L Cuthbert, ”Using TV White Space for Interference Mitigation in LTE Femtocell Networks”, IET International Conference on Communication Technology and Application (ICCTA), 2011.

[10] D3.3 “Control loops drive models & resource management assessment”, Deliverable D3.3 of the SACRA project, July, 2012.

[11] D5.2 ”Evaluation of Selected Measurement-based Techniques”, Deliverable of the FARAMIR project, February 2012.

2.8 JRA #H: Impact of channel model in the performance evaluation of wireless systems Nikolaos Bartzoudis, Oriol Font-Bach, Miquel Payaró – CTTC Claude Oestges, Evgenii Vinogradov – UCL


Security: Public

2.8.1 Description of Activity This JRA has experimentally analysed the impact that different channel models have on the performance of modern wireless communication systems. Towards this end, a complex laboratory setup comprising the following entities has been developed:

• A real-time FPGA-based PHY-layer prototype that enables the evaluation of 4G (LTE/LTE-A) and 5G (FBMC-based) wireless systems.

• A real-time multi-channel emulator with capacity to provide both static and mobile channel conditions (i.e., for standard and custom channel models), as well as allowing to reproduce the desired interference scenarios in a controlled lab environement.

Based on this hardware setup, the performance of a real-time FBMC-based spectrum coexistence scheme has been evaluated under realistic operating conditions. The present JRA has a strong link with JRA#C, which provided different channel models to be loaded onto the channel emulator. Finally, a series of performance metrics have been extracted for different operating conditions, providing an assessment of the impact that the considered channel models have on the performance of the system under test. 2.8.2 Relevance with the identified fundamental open issues The validation and performance assessment of novel PHY-layer algorithms and technologies is typically limited to computer models and simulations. However the experimental validation which is not that common in academic research due to lack of resources, expertise and the high complexity involved, adds a top-up value to the developed PHY-layer technologies since it brings a research idea closer to reality (account for implementation losses, signal impairments, non idealities, realistic mobility-interference and respective propagation channels, …). The mentioned factors pose important limitations when the target scenario is modelled and validated using a computer-based simulation environment. Thus, an effective way to overcome these limitations is to develop a PHY-layer prototype capable of generating, receiving and processing the required signals in real-time as it has been done in this JRA. 2.8.3 Results and achievements An experimental performance evaluation has been carried out within a communication scenario that features two of the key enablers of 5G: (i) the use of post-OFDM modulations and (ii) an efficient use of the spectrum via spectrum sharing. The experimental lab set-up was assembled so as to operate in conditions as realistic as possible via the actual real-time implementation of the involved transceivers (work carried out by CTTC) and also via the utilization of propagation channels which have been recorded in a field measurement campaign (work carried out by UCL) and which are loaded in a channel emulator that also operates in real-time. The experimental results show that co-existence in a shared spectrum scenario is possible and that the performance degradation is kept at a low level as long as one of the two users is making use of spectrally agile post-OFDM modulations such as filterbank multicarrier (FBMC). 2.8.4 Publications O. Font-Bach, N. Bartzoudis, D. López, E. Vinogradov, M. Payaró, C. Oestges, T. A. Myrvoll, V. Ringset, “Experimental performance evaluation of a 5G spectrum sharing scenario based


Security: Public

on field-measured channels,” in Proc. of IEEE Intl. Symposium on Personal, Indoor, and Mobile Radio Communications (PIMRC), Aug. 30 – Sept. 2, 2015, Hong Kong (China). O. Font-Bach, N. Bartzoudis, A. Pascual Iserte, M. Payaró, L. Blanco, D. López Bueno and M. Molina, Interference Management in LTE-based HetNets: a Practical Approach, Transactions on Emerging Telecommunications Technologies, (available on-line, DOI: 10.1002/ett.2833; April 2014). O. Font-Bach, N. Bartzoudis, M. Payaró and A. Pascual-Iserte, Measuring the performance of a distributed interference management scheme in a LTE-based HetNet deployment, (to appear) in Proceedings of the IEEE Sensor Array and Multichannel Signal Processing Workshop (SAM 2014), 22-25 June 2014, A Coruña (Spain). N. Bartzoudis, O. Font-Bach, M. Payaró, A. Pascual-Iserte, J. Rubio, J. J. García Fernández and A. García Armada, Energy Profiling of FPGA-based PHY-layer Building Blocks Encountered in Modern Wireless Communication Systems, (to appear) in Proceedings of the IEEE Sensor Array and Multichannel Signal Processing Workshop (SAM 2014), 22-25 June 2014, A Coruña (Spain). O. Font-Bach, N. Bartzoudis, M. Payaró and A. Pascual-Iserte, Hardware-efficient implementation of a Femtocell/Macrocell interference-mitigation technique for high-performance LTE-based systems, in Proceedings of the International Conference on Field Programmable Logic and Applications (FPL 2013), 2-4 September 2013, Porto (Portugal). 2.9 JRA #I: Hybrid radio channel modeling using graphs and rays Troels Pedersen, Gerhard Steinböck (AAU)

Mingming Gan, Thomas Zemen (FTW)

Paul Meissner (TU Graz), Erik Leitinger (TU Graz), Klaus Witrisal (TU Graz)

2.9.1 Description of Activity Ray-tracing tools, which consider only specular reflection and diffraction, are not sufficient to characterize experimentally obtained ultra wideband channel responses. The experimental channel responses include diffuse components, which are created possibly by a multitude of higher order diffuse reflections. Capturing the diffuse tail by using a higher order diffuse reflections is computationally prohibited in a ray-tracing tool. In contrast hereto, the propagation graph model allows an efficient calculation of the channel response which includes a transition from specular components to a diffuse tail. A combination of the two modeling approaches will lead to a new and computationally efficient hybrid model. This hybrid model includes deterministic and diffuse components. The objective is to develop a new hybrid model that is able to reflect the deterministic and diffuse components observed in measurements. 2.9.2 Relevance with the identified fundamental open issues In this JRA we investigate how to combine ray-tracing with propagation graphs to obtain a simplified, less computationally intensive model. To tackle the challenge of developing hybrid models with wide applicability, by successively generalizing the model to deal with a number of scenarios covering diverse application areas, such as indoor communications, vehicular communications and indoor-to-outdoor communications. 2.9.3 Main Results Achieved in the Reproting Period and planned activities The hybrid model in the context of reverberant in-room channels: We proposed a hybrid model for simulation of in-room channels. The hybrid model allow for allows for computationally efficient modeling of the diffuse tail. Complexity Reduction for sub-band divided ray-tracing for UWB indoor scenarios:


Security: Public

We reduced the computational complexity by more than one order of magnitude of sub-band-divided ray-tracing by applying two-dimensional discrete prolate-spheroidal sequences (DPS) in the spatial and frequency domain. Outdoor-to-indoor radio channel modeling using propagation graphs: The propagation graph was used to study the outdoor-to-indoor radio channel (in particular, the clustering phenomenon caused by multiple excitation of the in-room reverberation phenomenon). The hybrid model in the context of vehicular to vehicular channels in tunnels: The hybrid model was extended for vehicular-to-vehicular channels in tunnels. The extension focuses in particular on the time-variant behaviour of such a radio channel. 2.9.4 Publications Gan, M, Xu, Z, Hofer, M, Steinböck, G & Zemen, T 2015, 'A Sub-band Divided Ray Tracing Algorithm Using the DPS Subspace in UWB Indoor Scenarios'. in IEEE 81th Vehicular Technology Conference (VTC Spring), 2015. Proceedings of the IEEE Vehicular Technology Conference. Steinböck, G, Gan, M, Meissner, P, Leitinger, E, Witrisal, K, Zemen, T & Pedersen, T 2015, 'Hybrid Model For Reverberant Indoor Radio Channels Using Rays and Graphs' IEEE Transactions on Antennas and Propagation. Pedersen, T, Steinböck, G & Fleury, BH 2014, 'Modeling of outdoor-to-indoor radio channels via propagation graphs'. General Assembly and Scientific Symposium (URSI GASS), 2014 XXXIth URSI., 2953, IEEE, s. 1-4. General Assembly and Scientific Symposium (URSI GASS), 10.1109/URSIGASS.2014.6929300 Gan, M, Xu, Z, Hofer, M, Steinböck, G & Zemen, T 2015, 'A Sub-band Divided Ray Tracing Algorithm Using the DPS Subspace in UWB Indoor Scenarios'. in IEEE 81th Vehicular Technology Conference (VTC Spring), 2015. Proceedings of the IEEE Vehicular Technology Conference.


Security: Public

3. General Conclusions and Prospects In this deliverable, a report on the technical activities conducted in the different JRAs in Newcom#’s WP21 during the last year of the project has been presented. The JRAs described in this document have addressed the challenges that were identified as key issues in the development radio interfaces for the Future Internet and that, today more than ever, continue to be of utmost importance. In particular: • Nine JRAs have performed joint research on experimental activities in cutting-edge

topics that are relevant for next generation wireless systems in the topics of:

o Low-emission and low-energy

o High spectral efficiency

o Positioning (both indoors and outdoors)

o Channel measurements and channel modelling

• The facilities at EuWIn@CTTC (whose main assets are GEDOMIS®, GNSS-SDR, and OpenInLocation) have been successfully utilized by all Newcom# partners participating in WP2.1 and also by the Newcom# associate institutions and external, third parties. These facilities have been complemented with the lab facilities that other partners (Bilkent, CTTC/UPC, UCL, and PUT) have utilized for part of their experimental research activities.

• The experimental data generated as a result of the research activities is available for all the research community via the project website repository.

• The WP has constantly contributed to the Newcom# industry dissemination events.

Regarding future prospects for radio interfaces, in the wireless communication research arena, the focus is now set on 5G, around which a huge amount of questions remain to be solved. With respect to previous mobile network generations, the 5G network will need a much higher degree of flexibility and adaptability, to serve applications with extremely diversified requirements, whose viability is yet to be proven. Moreover, from the radio access network side, the envisaged massive use of small cells, with distributed or centralised coordination and control, the novel backhaul paradigms based on multi-hop links, the use of fragmented spectrum, together with spectrum efficiency requirements pose many challenges that will need to be addressed before 5G can become a reality. While theoretical research will be (of course) required to deal with all these challenges, it won’t be until the proposed approaches and solutions are evaluated and validated experimentally that the true potential of 5G will be actually demonstrated and unleashed.

Consequently, in order to play a key role in the development of 5G, Europe requires the availability and use of experimental platforms and facilities such as the ones provided by EuWIn (arguably, one of the main assets for the Newcom community survivability).


Security: Public

4. Annex I: Detailed Description of Main technical WP Achievements 4.1 JRA #A: Enhanced NC-OFDM transmission with reduced spurious emission level 4.1.1 Predistortion of NC-OFDM transmitter utilizing USRP frontend 4.1.1.1 Introduction and model presentation An efficient utilization of the available frequency band requires the out-of-band (OOB) emission to be reduced. This is of particular importance in the case of cognitive radio, where licensed users, called primary users (PU), cannot by harmfully interfered by the unlicensed devices, called secondary users (SU). Effective spectrum utilization is possible thanks to multicarrier transmission, e.g., commonly used orthogonal frequency division multiplexing (OFDM). The non-contiguous OFDM (NC-OFDM) allows to aggregate distanced in frequency unused spectrum resources. It is done thanks to modulation of only a subset of all IFFT subcarriers. However, NC-OFDM has two main sources of OOB radiation, i.e., the sidelobes of the waveform used for carrying data- the sinc-shaped subcarrier spectrum in OFDM system and the intermodulation products (IMD) that originate from signal processing in nonlinear front-end modules. The subcarrier spectrum sidelobes are typically reduced by some specially designed algorithms, e.g., Optimized Cancellation Carriers [1]. The characteristics of the observed IMD products depend on various factors, however, it is the power amplifier (PA) characteristic and transmitted signal amplitude distribution that mostly influences the OOB emission level. The immediate solutions to IMD reduction are: 1) to reduce the variation of the amplitudes of the time-domain signal by peak-to-average power ratio (PAPR) reduction techniques, such as tone reservation (TR), 2) to linearize the PA characteristic by predistortion, 3) to use a more advanced, and thus more expensive and energy consuming power amplifier. The third solution can be excluded as the most cost and energy inefficient. Previously, the signal amplitude variations minimization has been combined with subcarrier spectrum sidelobes reduction using Extra Carriers (ECs) [2]. In this work the NC-OFDM signal utilizing ECs, designed within Track 1, is transmitted via practical Software Defined Radio (SDR), i.e., USRP. However, in order to reduce OOB emission level even further, the predistortion of USRP frontend has been implemented. It has been presented in [3].

The considered NC-OFDM transmitter is depicted in Fig. 1. NC-OFDM symbols samples xn are processed in the predistorter, which generates modified (predistorted) signal samples yn according to the function yn = G(xn). Such samples are converted to their analog form and transmitted via USRP. The characteristic of the PA is nonlinear, so the transmitted signal can be described as , where ρ is the linear gain.


Security: Public

NC-‐OFDM transmitter

PREDISTORTER RF FRONTEND

INPUT BITS

xn yn

zϱ

)(~ nyPnz~

)~( nzG-‐

+ en

ny

POSTDISTORTER

)( nxG

Figure 1 Scheme of the predistorter design.

There is a number of different methods to model . It can be described by amplitude-amplitude (AM-AM) and amplitude-phase (AM-PM) characteristics. It describe how the amplitude and phase of the output signal change as a function of the amplitude of the signal observed at the PA input. Ideally, the AM-AM characteristics of the PA should be linear. The nonlinear behavior has two main components: static nonlinearity (i.e., where a given output sample depends only on one appropriate input sample) and memory effect (also known as dynamic nonlinearity, where the value of a certain output sample depends on the set of preceding samples). Contemporary solid-state PAs are usually modeled by means of the so-called Rapp model, however, it does not include the memory effect. Thus, various formulations can be found in the literature that consider specific features of the PA [4]. Let us mention, e.g., the Look-up-table model, Volterra model, Polynomial model, Wiener model or Hammerstein model. In this work, the memoryless polynomial model, since it provides a relatively high degree of freedom in model design, is easy for practical implementation and considers a mild dynamic nonlinearity. It is defined as:

(1)

where ⌈⌉ denotes the ceil function. For the polynomial of the Ω degree, only ⌈Ω/2⌉ (complex) coefficients pi are needed to define it. The polynomial of a higher degree allows for more precise modeling at the cost of increased computational complexity.

The predistorter function G(x) can be obtained by indirect learning architecture shown in Fig. 1. G(x) is estimated in postdistorter block and used to calculate predistorter coefficients. The PA output can be distorted by the nonlinear function G(x), so that the difference between the input of the PA and the output of the postdistorter is minimized. The predistorter characteristics can be modeled usinq quasi-memoryless polynomial model of Ξ degree (similarly as PA model). It can be defined as

(2)

Equation can vectorized for N consecutive samples n=1,…,N as

(3)

where Z = {Zn,i} for Zn,i = , and . The predistorter coefficeints can be obtained by means of pseudoinverse (denoted as +)

(4)


Security: Public

4.1.1.2 USRP nonlinearity extraction and predistortion The PA linearization shown above has been used for an USRP board. The experiment setup is shown in Figure 2.The NC-OFDM waveform is transmitted using a computer with a GNU Radio installed (UHD 003.007.001, GNU Radio v. 4.6.3). The baseband complex samples with rate of 10 MSps are transferred to USRP N210 with a WBX daughterboard installed and configured to a maximum output power of 20 dBm. In order to remove the local oscillator leakage and IQ imbalance effects occurring in WBX board, a local oscillator offset of 12 MHz is used (generated in USRP’s FPGA). After passing a TX amplifier (GVA-84+) and RF switch (HMC174MS8), the signal at the carrier frequency 500 MHz is

Figure 2 Connections setup

transmitted via a TX/RX port using a H155 cable to a Rohde&Schwarz FSL6 spectrum analyzer. It can provide raw IQ samples to a computer connected via an Ethernet cable. Matlab with Instrument Control Toolbox is used for signal reception and processing. In order to obtain reliable nonlinearity characteristic a number of conditions has to be meet. The test signal should have similar characteristic to the one used in a normal operation. A standard NC-OFDM symbol utilizing CCs has been used with a maximal amplitude equal 1 (not to be clipped in USRP’s FPGA). The observed bandwidth should be much wider than the test signal. It is obtained by modulating only 161 subcarriers out of available 1024. Additionally, the receiver should be linear as the high-end R&S device.


Security: Public

0 0.2 0.4 0.6 0.8 10

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

input amplitude

outp

ut a

mpl

itude

0 0.2 0.4 0.6 0.8 1-50

0

50

100

imput amplitudeph

ase

diffe

renc

e (d

egre

es)

before CFO anddelay correctionafter CFO andbefore delay correctionafter CFO anddelay correctionpolynomial model

Figure 3 AM-AM and AM-PM characteristics at different stages of measurement

In the received samples the test signal was sought for using crosscorelation. The resultant characteristics are shown in Figure 3 with black points (covered by blue points in AM-AM characteristic). As local oscillators in both transmitter and receiver are different, Carrier Frequency Offset (CFO) has to be estimated and corrected. The resultant plots (blue points) show significant improvement of AM-PM characteristic. Additionally, the sampling times in transmitter and receiver can be different. Lagrange interpolation has been used to estimate fractional time delay and compensate it. The resultant characteristics (red points) show nearly no memory effects of the WBX board. For a given input amplitude only one output value is visible. Only mild memory effect is visible as AM-PM characteristic is not flat. However, It can be modeled and predistorted by our polynomial model (black line) that equals

(5)

Interesting phenomenon has been observed for input amplitude exceeding 1. It is shown with grey points in Figure 4. Probably in the USRP’s FPGA most significant bit is zeroed. It causes high variations of AM-AM and AM-PM characteristic for amplitudes exceeding 1. As these amplitudes can be obtained at the output of predistorter an extra clipper (soft limiter) has been added in Gnu Radio. The resultant characteristics (black points) are improved.


Security: Public

0 0.5 1 1.50

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

input amplitude

outp

ut a

mpl

itude

0 0.5 1 1.5-200

-150

-100

-50

0

50

100

150

input amplitudeph

ase

diffe

renc

e (d

egre

es)

with extra clipperwithout extra clipper

Figure 4 Characteristic with and without extra clipper

0 0.2 0.4 0.6 0.8 10

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

input amplitude

outp

ut a

mpl

itude

0 0.2 0.4 0.6 0.8 1-30

-20

-10

0

10

20

30

input amplitude

phas

e di

ffere

nce

(deg

rees

)

RX with USRP(0 dB gain)RX with USRP(38 dB gain)RX with FSL6 -polynomial model

Figure 5 Characteristic for measurement with internal USRP RX path.

As shown in Figure 2, the RX path is connected to TX path by two switches (26 dB isolation each). The characteristics has been measured with the highest and the lowest possible RX gains, i.e., 38 dB and 0 dB, respectively. Unfortunately, it is visible that for high RX gain strong nonlinearity occurs in the RX path. Decreasing gain improves results (it is closer to the reference characteristics) but still important differences occur. In order to be able to estimate USRP TX nonlinearity by means of its RX path, nonlinearity of the RX path would need to be removed. It can be done by means of RX predistortion or redesigning RX path (attenuator before AMP1).

The predistortion has been implemented in GNU Radio by a speciall block utilizing Horner method for fast polynomial evaluation. The utilized predistorter is

(6)


Security: Public

The resultant characteristics after predistortion are shown in Figure 6. It is visible that the predistortion linearizes USRP’s characteristics. However, signals exceeding amplitude of about 0.8 will be clipped.

0 0.2 0.4 0.6 0.8 10

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

input amplitude

outp

ut a

mpl

itude

0 0.2 0.4 0.6 0.8 1-10

-8

-6

-4

-2

0

2

4

6

8

10

imput amplitude

phas

e di

ffere

nce

(deg

rees

)before predistortionafter predistortion

Figure 6 Characteristic of the USRP+XBW before and after proposed predistortion.

4.1.1.3 Improvement of NC-OFDM OOB radiation level thanks to predistortion The usefulness of proposed predistortion has been tested by transmitting NC-OFDM waveform via USRP board with/without predistortion applied and observing resultant PSDs on spectrum analyzer. The NC-OFDM utilizes N=1024 point IFFT with the occupied subcarriers indexed {−80;−1} � {1; 16} � {33; 80}. The cyclic prefix utilizes N/4 samples. 30 subcarriers out of those occupied are used for PAPR reduction (TR method), sidelobes suppression (CC method [1]) or both (ECs proposed in [2]). The other active subcarriers are modulated by QPSK data symbols. The maximal amplitude should be obtained by a system utilizing CC. It was normalized to be equal 1. All waveforms were normalized to have the same mean power. The resultant PSDs are shown in Figure 7. It is visible that application of only ECs is not enough to obtain low OOB radiation level. Usage of predistortion only does not result in low OOB radiation level as well. The lowest OOB radiation level is observed only for the combination of ECs with predistortion. It signigicanly (by at least 10 dB) outperforms the other presented methods.


Security: Public

-1000 0 1000 2000 3000 4000 5000

-60

-50

-40

-30

-20

-10

0

frequency (kHz)

Nor

mal

ized

PSD

(dB)

ECs&predist.(SA)ECs(SA)ECs(Soft.)CCs&predist.(SA)CCs (SA)CCs (Soft.)TR&predist.(SA)TR (SA)TR (Soft.)

Figure 7 PSDs of different NC-OFDM waveforms with and without predistortion applied.

4.2 JRA #B: Practical Implementation of Polar Codes 4.2.1 Study of algorithms and architectures for the decoding of Polar Codes POLAR codes are a recent class of channel codes with low complexity encoding, decoding, and code-construction algorithms. It was recognized from the beginning that the finite-length performance of polar codes was not competitive with the state-of-the-art codes such as Turbo and Low-Density Parity- Check (LDPC) codes. This was in part due to the suboptimal nature of the SC (Successive Cancellation) decoding algorithm, and in part due to the relatively weak minimum distance properties of polar codes. A second undesirable aspect was the fact that the SC decoder makes its decisions sequentially, which means that the decoder latency would grow at least linearly with the code length, creating a throughput bottleneck. The belief propagation (BP) algorithm is the most popular alternative proposed for the decoding of Polar Codes. The BP decoder did not improve performance significantly over SC decoding, but delivered higher throughput at high signal-to-noise ratio (SNR) thanks to the inherently parallel decision-making capability of such decoders. This Section describes the most important activities completed on the decoding of Polar Codes and particularly on implementation architectures aiming at achieving highest possible throughput with polar codes under BP decoding while also minimizing the implementation complexity. The presented results come from the joint cooperation between two research groups in Politecnico di Torino (Italy) and Bilkent University (Turkey).

As part of this cooperation, a PhD student from CNIT/Polito (Dr. Andrea Biroli) spent a 6 month period of collaborative work at Bilkent University. The final objective of the joint effort between the two partners is to design an innovative hardware decoder for Polar Codes.

Initially, the key elements to be pursued in the development of advanced channel coding decoder architectures have been deeply investigated. Starting from known results in the scientific literature and from current market requests in terms of throughput and energy dissipation, feasible solutions have been pointed out.

The main motivations behind the choice of belief propagation is the capability of the BP algorithm to output a complete codeword after a small number of iterations and its high degree of potential parallelism: because of these known characteristics, the BP approach


Security: Public

seems able to overcome throughput limits of successive cancellation (SC) decoders and to reduce decoding latency. The most important target performance for the planned decoder was identified as a throughput as high as 1Gb/s, which is a relevant improvement with respect existing architectures, able to reach 100Mb/s.

As a second step, the polar code BP algorithm has been deeply studied. Starting from previous software models available at Bilkent University and Polito, an improved C model has been developed for a system including BP polar encoder, transmission channel and decoder. Simulations allowed to validate the model and to generate test vectors to be used for the assessment of the hardware model.

As soon as the functional C model was ready, tested and debugged, a refined, hardware oriented C model has been derived, which closely follows the planned decoder architecture. The new model includes quantization, storage and arithmetic details that allow for direct comparison of software and hardware simulation results.

In parallel, a VHDL model has been produced describing a fully parallel decoder architecture, targeting ASIC implementation; the model incorporates several parameters, which enable the hardware tuning according to the results of simulations from C code.

In this phase of the work, the possible use of automatic tools for the generation of VHDL models from high level C language descriptions has been considered and evaluated; however, this approach has been abandoned because of the strict performance constraints.

Testbenches were written in order to prove the correctness of the VHDL code in compliance with expected results from the C simulations.

The synthesis of the fully parallel architecture has been performed by means of Synopsys Design Compiler (version Z-2007.03-SP1), and the obtained circuit has been characterized in terms of throughput, occupied Silicon area, working frequency and power consumption.

A lot of effort has been put in the best VHDL representation for synthesis of the processing element (PE) and the decoder overall. PE is the basic structure that compose the decoder. The basic PE for the proposed BP decoding architecture is a component that has four inputs and four outputs, implementing two sets of belief updates: a left-wave of updates and a right-wave updates.

It has been proven that 58% in terms of latency performance can be gained avoiding the use of common VHDL macro instantiations.

The results achieved were promising (overcoming prefixed target) so this architecture had been chosen as basic structure for the following developments.

Subsequently, the work followed two parallel branches. The first one aims at mapping a properly refined and optimized version of the obtained VHDL design onto an FPGA board. The second branch has to do with the design of a reduced complexity architecture, with some further architectural refinements and able to save implementation complexity with limited effects on communication performance.

To fulfill this task all VHDL writing and testing has been done and also new simulative C code for the novel architecture has been produced.

The new model has been synthesized and validated using the same approach and tools described for the initial implementation.

Both fully parallel and reduced complexity architecture designs are reported in Figure .

The reduced complexity (RC) architecture is composed of a single stage of PEs that receive the log likelihood ratio (LLR) values either from a register memory block or from a ROM memory R0 at the start of each iteration. R0 stores the a-priori information of frozen bits, the memory block stores (n + 1) x (N) LLR messages, where N=2n is the codeword length.


Security: Public

Figure 8: Parallel and folded architectures

Since the performance of the BP depend on the schedule, an extensive analysis on achievable results, with constrained small number of iterations has been performed on all following scheduling.

A. Linear LR, bidirectional: Each iteration starts at the leftmost stage and ends at the right-most stage. Updates are bidirectional.

B. Linear RL, bidirectional: Each iteration starts at the rightmost stage and ends at the left-most stage. Updates are bidirectional.

C. Circular LR, bidirectional: Stages enabled consecutively from left to right, then from right to left, then from left to right, and so on, as if the end stages “bounce” the propagating wave of beliefs.

D. Circular RL, bidirectional: Same as schedule C but starting from the rightmost stage. E. RL + A: Start at the rightmost stage and continue towards left, upon reaching the

leftmost stage continue as in Schedule A. F. LR + B: Start at the leftmost stage and continue towards right, upon reaching the

rightmost stage continue as in Schedule B. G. Circular RL, unidirectional: Same as Schedule B, except unidirectional updates in the

direction of schedule motion. H. Circular LR, unidirectional: Same as Schedule A, except unidirectional updates in the

direction of schedule motion. I. Simulatenous LR and RL, bidirectional: Schedule A and B superposed. L. All ON: All stages updated simulatenously. M. Odd-Even: Odd-numbered and even-numbered stages are updated in alternate clock

cycles.

Figure depicts the above schedules for the case of a decoder with 3 stages (block length N = 8).

A schedule is called bidirectional if the PE carries out both a left and a right update when it is activated. A unidirectional update is one in which either a left or a right update is carried out. The simulative C model has been modified in order to compare all possible scheduling and select best performance.


Security: Public

Figure 9: Scheduling alternatives for a 3-stage decoder. Underlined numbers indicate the start of an iteration, numbers in parenthesis are executed only at the first iteration and numbers in square brackets are executed in parallel. The stage s performs both right (sR) and left (sL)

updates ( ).

Fully parallel hardware implementation or reduced complexity architecture can be selected according to the considered scheduling, because some scheduling can only be carried out by fully parallel hardware without latency degradation.

Achieved results have been compared against published implementations available in the literature.

An implementation of BP fully parallel decoding for polar codes achieves a throughput of 2 Gb/s, with an energy consumption of 19 pJ per decoded bit.

The BP decoder architecture developed has about twice as high a throughput as best solutions reported in the literature while having a comparable area efficiency. Early stopping rules can increase the throughput of BP decoders significantly, but they were not implemented in order to distinguish the effect of architecture from them. It is also worth pointing out that the energy efficiency of the proposed reduced-complexity architecture is about five times better than that of the previous architectures in the literature.

The energy estimation has been evaluated by propagating the switching activity from input ports through all internal nets of the architecture. The Synopsys tool reports the power of the design as a sum of estimated clock tree power and netlist power.

On the negative side, the area efficiency of the proposed BP decoder is about four times worse than the fastest considered SC decoder.

The designed decoder and the collected comparisons will form the core of a new joint paper that will be submitted to an international journal for review.

In order to increase probability of acceptance of the paper, synthetized results will be substituted with post-layout simulation given by Cadence Encounter and detailed explanations of hardware architecture will be given.


Security: Public

As further prosecution on the improvement of the hardware architecture a deep analysis on all possible permutation of the previously developed architecture have been performed.

Polar code algorithm does not implicit give an order on partial computations of probability (computed by subsequent stages), but performance are strongly affected by this choices as also pointed out in literature. Nevertheless apart from the classical stage ordering, no other ordering study have been published.

In order to evaluate best architecture staging and compare it with the one used, the C code developed to evaluate performance of a single architecture, has been modified.

The high number of permutations, given by the number of stages used by the architecture, have a strong impact to the computation time needed.

To reduce the computation time Circular RL unidirectional scheduling (G.) has been chosen because according to obtained previous results it is the fastest way to evaluate polar code decoding in software. This approximation is based on the hypothesis that different schedulings affect convergence trend of BER/FER performance linearly.

A cluster of processors have been exploited to carry out all necessary computation.

Computational resources were provided by hpc@polito, which is a project of Academic Computing within the Department of Control and Computer Engineering at the Politecnico di Torino (http://www.hpc.polito.it).

Obtained results show that the commonly used ordering is among first 15 best results evaluated, however performance gain for the best solution is in the order of 0.1dB, so it is almost negligible to consider a more refined staging sequence to increase performance.

A new branch of study for polar codes decoders has also been followed, successive cancellation (SC) decoding with fast simplified SC (FSSC) have been studied. This particular simplification of the SC decoding have been recently published and it shows to largely increase the throughput decoding limitations of previous SC implementations.

A C code for SC polar encoding, transmission channel and decoding has been used as starting point for the FSSC decoding structure; its development is ongoing.

A collaboration with Centre Tecnològic Telecomunicacions Catalunya (CTTC), has run in parallel with the previous task.

The final objective of this collaboration is to verify polar code performance with more refined simulations and emulations.

It was agreed that Polito furnished binary coded sequence of sample data, via Newcom# repository. CTTC took these data and perform channel emulation, including physical problems and channel properties (synchronization, carrier frequency offset, rotation), and upload log likelihood probabilities (LLRs) of the received data to the repository so the decoding procedure and performance evaluation could be carried out by Polito.

Chosen channel to be tested were:

• simple AWGN channel to verify simulated and emulated convergence; • pedestrian LTE channel, with BPSK sample coding • pedestrian LTE channel, with QPSK sample coding to be compliant to the standard

As conclusion of the joint collaboration a report on the activity and results achieved will be prepared.


Security: Public

4.3 JRA #C: Assessment and development of multi-link channel models 4.3.1 Development of indoor multi-link channel models (UCL, UGent)

The channel can be represented as the addition (when expressed in logarithmic scale, i.e. in decibels) of the primary and reverberant components, shadowing, and small-scale fading:

(7)

In order to extract the components and to parameterize the model, measurements were carried out in a typical environment that was located on the second floor of an office building and consisted of two adjacent typical office rooms separated by a wall (see Fig. 10). Measurements were performed with UCL/ULB Elektrobit PROPSound™ Channel Sounder at a carrier frequency of 3.8 GHz using a transmission bandwidth of 200 MHz. The nodes were connected with the channel sounder using long low-loss RF cables, which had equal length. The RF cables had excellent RF stability, even when they were slightly bent or moved during the measurements. At the nodes, custom-made dipole antennas with a gain of 1.75 dB and an omnidirectional radiation pattern were used. The channel sounder used long (20.47 ) pseudo-noise (PN) sequences to estimate the impulse response of the radio channels between Tx and Rx nodes.

Figure 80. Floorplan of measurement rooms

The primary component is usually modeled as a Dirac delta function, i.e., only one specular component, however, for shorter distances the components is rather a set of Dirac functions:

(8)

where P0, M, are the strongest component at the distance d0, the orders of reflections and the characteristic time-delay that is required before a given ray makes one reflection in the room. It is shown in [5] that depends only on the room dimensions.

The modeling of the reverberant component is straightforward. The methodology is based on the one-slope method [6]. The reverberation time of the considered room and the power corresponding to the beginning of the reverberant components are the required parameters to model the reverberant component. An empirical formula to determine the reverberation time as a function of the frequency is derived in [6] and can be used to assess the reverberation time without resorting to measurements. However, note that in this study we consider a room-to-room scenario as illustrated in Fig. 10. The expressions of the reverberation time for in-room and room-to-room communications are derived in [7].

We observe in measured data that the shadowing for a fixed delay is described by a normal distribution with zero-mean by definition and . The model parameters are therefore the


Security: Public

delay-variant standard deviation , and the slope of the temporal autocorrelation. It turns out that decays exponentially over delay. Using exponential fit to measured data the standard deviation can be modeled by . The dynamic shadowing autocorrelation over time is modeled as a decreasing exponential, whose decay time

Hence, we use an autoregressive process to generate autocorrelated dynamic shadowing values:

(9) where x is a time series of length T, whose values are drawn independently from the normal distribution (zero mean with unit variance). No correlation between dynamic shadowing and the small-scale fading states has been found, so that we can combine independently simulated dynamic shadowing and small-scale fading realizations.

To model time-delay fading a concept similar to [8] has been used. For a fixed delay, fading is found to be Rician/Rayleigh distributed with the parameters varying over time. Temporal transitions between different distributions can be modeled by a Hidden Markov Model parameterized from the measurements. It turns out that the first order statistics are not correlated over delay bins, consequently, time-variant fading within a bin can be modeled independently from neighboring bins. After simulation of the temporal changes of the first order statistics of fading, we use the appropriate second order statistics [8] to generate fading realizations.

By addition of the four described channel components, we obtain a complete model combining physically based deterministic approach (room electromagnetic) and a measurement based stochastic approach using on the double ring geometry [9].

The method has been validated against the measurements and excellent agreements are obtained for 8 nodes out of 10. In the two cases of relatively bad agreement, observed errors of modeling can be explained by a significant influence of metallic cupboards (see Fig.1). Note, than the model can be easily extended to a double mobile cases by using a five-states HMM described in [8]. 4.3.2 Dense multipath depolarization modeling (UCL, UNIBO) The polarimetric characterization of the dense multipath components (DMC) builds upon a diffuse scattering model based on the “Effective Roughness” approach [10-11]. For each wall-surface element, it is assumed that the impinging power is partially reflected specularly and partially scattered according to a given diffuse scattering coefficient S, which is the amount of power scattered in all the directions at the expense of the specular reflection. Several shapes of the scattering patterns are possible, but prior studies have revealed that a directive lobe centered on the direction of specular reflection provides the best fit to measured data,

( ) 1 cos2

R

RRA

αψ

ψ+⎛ ⎞= ⎜ ⎟

⎝ ⎠ (10)

where ψR is the angle between the specular reflection direction and the scattering direction and αR is an integer that sets the width of the scattering lobe. An additional parameter Kxpol is defined such that 1 ≤ Kxpol < 1 to account for DMC depolarization. When Kxpol is equal to 0, the diffuse scattered field follows the polarization behavior of the specular field; when Kxpol increases, the cross-polarization coupling increases, and the scattered field reads as,

1 ˆ ˆ1 χ= − +js xpol S co xpol S xpK e i K iE E E (11)


Security: Public

where unit vectors coi describes the DMC polarization in the absence of depolarization, and

xpi is orthonormal to coi .

In order to validate the model and its parameters, measurements were performed in different types of outdoor and indoor environments with the UCL Elektrobit MIMO Channel Sounder, using a transmission bandwidth of 200 MHz [12-13]. The carrier frequency was 3.8 GHz for the outdoor measurements, and 3.6 GHz for the indoor measurements. The transmit power was always 23 dBm. In the outdoor measurements, uniform linear arrays of four dual-polarized patch antennas (with +45°/-45° slanted polarization) were used at both transmit (Tx) and receive (Rx) sides. In the indoor measurements, the same linear array as in the outdoor measurements was used for the fixed terminal, while at the mobile terminal a tri-polarized antenna was used, which was moved with an automatic positioning device to create a virtual 5 × 5 × 5 uniform cubic array. Three different categories of environments have been investigated to validate the model by means of measurements and RT simulations: two simple outdoor environments (isolated rural and office buildings), two complex urban environments (campus and street canyon) and two typical indoor environments (laboratory and office rooms). In each case, the following experimental setup has been used:

• isolated buildings: static measurements with Tx and Rx units facing the largest wall of the building under test; the Rx position is kept fixed, while the Tx unit is moved at different distances from the Rx unit;

• complex urban scenarios: dynamic measurements with a mobile pedestrian Tx and a fixed Rx unit (located on top of a building for the campus scenario and at pedestrian height in the canyon scenario);

• indoor environments: static measurements with the Rx unit being kept fixed in a room and the Tx unit being located at various positions along a route crossing different rooms

Based on all reported evidence, we can draw the following conclusions regarding the impact of diffuse scattering on the amount of channel depolarization in a variety of environments:

• the depolarization degree of the DMC, modeled into the ER model through the parameter Kxpol appears to increase with the complexity of the environment,

• the DMC seems to play an important role in the determination of the actual degree of depolarization coupling in all cases, since the XPD of the DMC is always lower than the overall XPD,

• the overall depolarization level can be low – i.e. the XPD relatively high – even in complex environments such as indoors, mainly due to the low link distance and to the ubiquitous presence of a strong dominant path (LOS or obstructed-LOS) which overshadows the DMC contribution.

4.3.3 Extended ray-tracing modeling for UWB transmissions (UCL, FTW)

For an ultra-wideband system, the channel characteristics vary significantly over the entire bandwidth. To cope with this, a sub-band divided ray-tracing (RT) has been proposed by dividing the frequency of interest into multiple sub-bands and superposing the RT results at the individual center frequency of each sub-band [14]. Thus, the computational complexity is directly proportional to the number of sub-bands.

In this research, we proposed a mathematical method to significantly reduce the computational complexity of the sub-band divided RT, making it almost independent of the number of sub-bands. This is illustrated in Fig. 11, where several methods are compared with experimental results. It is important to note that, based on this approach, not only the determination of the rays reaching a give location is made only once, but also the electromagnetic calculation of the received signal is not needed to perform repeatedly. The


Security: Public

accuracy of low-complexity sub-band divided RT algorithm is verified through a measurement campaign.

Figure 11. Normalized PDP comparison based on measurements, conventional sub-band divided RT, low-complexity

sub-band divided RT with 15 subbands, low-complexity sub-band divided RT with 50 subbands.

4.4 JRA #D: Channel models for cooperative positioning 4.4.1 Estimation of the Human Absorption Cross Section Via Reverberation Models It is well known that the characteristics of the radio channel may be altered by the presence of humans in the propagation environment. It is particularly relevant for communication systems to consider such effects in modeling densely populated propagation environments such as public transport systems, conferences, meeting rooms, concert halls, sports arenas, etc. For in-room scenarios, where the transmitter and receiver are located within the same room, it has been shown in [15] that important characteristics such as path loss and rms delay spread, can be predicted from volume, surface area, and average absorption coefficient of the room, by the theory of room electromagnetics [15]–[18]. Specifically, it has been proposed to model the so-called reverberation time, defined as the decay rate of the delay power spectrum, by room acoustical reverberation models by Sabine and Eyring. These models predict that the reverberation time changes if humans enter the room. The amount of change can be computed provided knowledge of the average absorption coefficient and the surface area of the human body or at least the product of these parameters, denoted as human absorption cross section. There exist multiple models for the surface area of a human body, e.g. [19]–[22]. These models have been used to determine the specific absorption rate of humans [22]. These investigations, however, focus on the exposure of the body to simplistic fields (plane waves etc.) and the power absorbed in the body due to that. An alternative approach was taken in [23], [24] by applying the electromagnetic version of Sabine’s model to experimentally estimate the human absorption cross section with the goal of investigating the SAR under realistic conditions.

Here, our focus is different: we are particularly interested in how persons impact the radio propagation conditions which may be used in e.g. cooperative localization for indoor applications. Motivated by our experiments in [18], which revealed that Eyring’s model is more accurate than Sabine’s for typical in-room environments, we consider how the choice of reverberation models affects the estimation of the human absorption cross section. To this end, we propose a new method for estimating the human absorption cross section relying on Eyring’s model. We present estimates of the human absorption cross section for both


Security: Public

Sabine’s and Eyring’s models and compare these to values found in the literature. The obtained values can be used to predict channel characteristics for various population densities via models such as [18], [25].

We conclude that the absorption due to humans is significant enough to be measured with a wideband channel sounder. Furthermore, the estimates of the human absorption cross section obtained with Sabine’s and Eyring’s models differ. In the latter the absorption cross section is obtained as a function of the human surface area. The obtained values of the human absorption cross section are comparable to reported values in the literature. 4.4.2 Experimental Validation of the Reverberation Effect in Room Electromagnetics The delay power spectrum is widely used in both communication and localization communities for characterizing the temporal dispersion of the radio channel. Experimental investigations of in-room radio environments indicate that the delay power spectrum exhibits an exponentially decaying tail. This tail can be characterized with Sabine's or Eyring's reverberation models, which were initially developed in acoustics. So far, these models were only fitted to data collected from radio measurements, but no thorough validation of their prediction ability in electromagnetics has been performed yet. This paper provides a contribution to fill this gap.

We follow Sabine's original experimental approach, which consists in comparing model predictions to experimental observations in a room, while varying its mean absorption coefficient and total room surface. We find that Eyring's model provides a more accurate prediction of the parameters characterizing the decaying tail, like the reverberation time, than Sabine's model.

We further use the reverberation models to predict the parameters of a recently proposed model of a distance-dependent delay power spectrum. This model enables us to predict the path loss, mean delay and rms delay spread versus transmitter-receiver distance. We observe good agreement between predictions and experimental results.

4.4.3 Bayesian Ranging for Radio Localization with and without Line-of-Sight Detection Having accurate localization capability is increasingly important for wireless communication systems [26] [27]. One approach to increase localization performance is to rely on high precision ranging techniques [28]. State-of-the-art ranging techniques based on, for example, the received signal strength, angle-of-arrival, time-of-arrival, time-difference-of-arrival, etc, may be sensitive to line-of-sight (LOS) conditions [27] [29]. Therefore, accounting for the unknown LOS or non-LOS (NLOS) conditions is an issue considered in many ranging and localization techniques [30] [31].

To tackle this issue, one approach is to rely on LOS identification techniques. Such an approach is reliable provided that the signal bandwidth and signal-to-noise ratio (SNR) are sufficiently large [29] [32] [33] [34] [35] [36]. Existing LOS identification techniques include methods based on machine learning [35] [37] and hypothesis-testing [27] [30] [34]. The LOS identification step labels range estimates as “LOS” or “NLOS” to facilitate the localization algorithms [37] [38]. The rationale is that if the LOS condition can be correctly identified, this information can be used to improve the ranging and localization accuracy. However, in communication systems with limited bandwidth and SNR, the LOS detector may be unreliable [29].

Instead of identifying and mitigating NLOS range estimates, direct ranging, which infers the range parameter directly from the received signal, can be potentially applied. Employing the first- and second-order moments of the received signal, a direct ranging method has been


Security: Public

proposed in [39] for bypassing a related problem, i.e. first-path detection. This method relies on a channel model formulated via a point process to compute the required moments of the received signal. The proposed approximate maximum-likelihood ranging method exhibits significant gain in ranging accuracy compared to the non-coherent correlator-based estimator. In contrast to the methods in [40] [41], direct ranging operates without knowledge of the number of multi-path components in the channel response and separability condition on these components. Although the method in [39] still relies on the LOS state information, the principle of ranging without estimating intermediate parameters seems promising.

In the present contribution, we propose Bayesian ranging methods with and without LOS detection for multi-path channels. Inspired by the direct ranging principle, we make use of a channel model to approximate pdfs of the received signal. In addition, we incorporate prior information on the range and the LOS condition. For this setup, we propose and evaluate approximate MAP and MMSE estimators. In addition, we derive variants of these estimators with approximate MAP and Bayes

decision rules for LOS detection. We test the performance of the proposed methods by means of Monte Carlo simulations of an OFDM system with modest bandwidth. The approximate MMSE estimators outperform the approximate MAP estimators in terms of RMSE. Using the proposed pdf approximations, we observe that including LOS detection in the estimators, while adding complexity, has no major impact on the ranging performance. Our simulation study indicates that there is a potential for improving the ranging performance by relying on better pdf approximations.

4.5 JRA #E: Compressive sensing for sparse propagation channel estimation In order to estimate the CIRs of the wireless propagation channel, and under the assumption of linearity and time invariance, a linear model was proposed

In the UHF band it is currently admitted that the propagation is multipath, i.e. the signal from the transmitter to the receiver is carried by the line of sight path (LOS) if it exists but also by other paths also known under the acronym NLOS for Non Line of Sight. In wireless communications, due to multipath propagation physically observed effect, the signal at the receiver is then a composition of scaled and delayed versions of the transmitted signal. The channel can be modeled as a finite set of K scaled delays applied to the transmitted signal, as usually admitted from the introduction of the GSM propagation models as FIR (Finite Impulse Response) filter structure.

Considering that the K delays can be distributed on a grid of N > K positions TG and that the signal can be represented in discrete time domain with a sample rate of 1/TG , the problem can be written in its matricial form as:

y = Sh + n , (12)

where y = [y1 y2 … yM]T are the observations, S a matrix containing the known signal transmitted by the source, h =[h1 h2 … hN]T are the unknown channel taps to be estimated and n the noise vector assumed to be a circular additive white and Gaussian noise.

Based on this linear model and under the assumption of a Gaussianly distributed noise, a maximum of likelihood estimation of the complex impulse response vector is achievable using the pseudo-inverse solution to this linear problem:

HML= (SHS)-1SHy, (13)

that is also the least squares solution.

If some l2 regularisation techniques can be exploited to inverse this linear problem, an identification may be obtained solving the constrained problem:


Security: Public

(14)

with the least squares constraint and a constraint on the solution properties: for example smoothness of the solution in the Philips-Tikhonov–Twomey (PTT) regularized solution.

Also based on a l2 norm minimization is the Wiener filtering, where a very deep a priori knowledge on the solution is required: the power spectral density functions of the solution and the noise.

All the methods previously presented require to estimate all the components of the unknown vector h =[h1 h2 … hN]T even if many of hi terms are nulls or neglectables.

In this JRA-E our goal was to find a way to directly obtain a sparse solution to the identification of the propagation channel impulse response.

Consequently, as it is now relatively well addressed in the literature, we propose to introduce this new constraint using a l0 and/or a l1 norm. In this JRA-E, we propose to perform the sparse identification using a l0 norm.

Then, in order to recover a sparse solution vector h, one can search for a solution to the following least-squares penalized minimization problem:

(15)

whose estimation performance depends on the penalization parameter l that allowing the research of a compromise between data fidelity and sparsity. This problem belongs to the so called class of l0-optimization problems known to be NP-hard.

Also the substitution of l0- by l1-norm enables the use of convex optimization techniques such as the Forward Backward Splitting algorithm,

Consequently a l0-optimization had been performed and two algorithm tested: the “Iterative Hard Thresholding” (IHT) algorithm and the recently proposed CELO algorithm.

The obtained identification results are presented Figure 12.

Figure 12: MSE vs. SNR for channel estimation using l1 minimization (Forward-Backward Splitting (FBS) with soft

threshold), l0 minimization (IHT or CELO with FBS) and Oracle Cramer-Rao bound.


Security: Public

It appears clearly on Figure 12 that nonconvex penalties outperform classical l1 relaxation in the case of this identification. This result will be presented during the international workshop CAMSAP’2015 and published in the proceedings of this conference.

Nevertheless, these results are achieved by simulation and confirmations with really measured signals have to be performed. One of the main difficulties will be due to the channel sounder complex impulse response that degrades the time delay resolution of the identified channel CIRs. Different solutions are possible and have to be tested.

4.6 JRA #F: Design and experimental validation of algorithms for active and passive indoor positioning 4.6.1 Testbed description The mission of OpenInLocation Lab is to accelerate the development of cutting-‐edge indoor location solutions, featuring a mobile autonomous platform with the required hardware and software equipment to simulate indoor/outdoor real-‐life dynamic scenarios and allow for precise benchmarking trajectories and repeatability of experiments. The autonomous mobile robot (named after Gauss) consists of a mobile platform with motors, encoders, motor controller, IMU (with 3-‐axis accelerometer and gyroscopes), distance sensors and a navigation controller board with an Arduino Duemilanove microcontroller. This microcontroller controls the IMU, sensors, velocity and number of counts of the motors to perform the programmed trajectory. New trajectories can be programmed easily with straight paths and rotations. The position of the mobile robot during the trajectory is obtained through the number of counts of the motors and gyroscope measurements. 4.6.2 Autonomous mobile robot

Figure 13.- Robotic platform for the experimentation of indoor positioning algorithms. The board has an Arduino Duemilanove compatible microcontroller with 6 ADC converters, 6 pins PWM output and UART/USB port. Four IR

distance sensors are connected to four ADC converters.

The functionality of the code is shown in next diagram.


Security: Public

Figure 14.- Robot’s control flowgraph.

Main features:

• Rotations with gyroscopes. • Movement of the robot to next waypoint and avoiding of obstacles. • Coordinates of the robot and position estimation from origin. • All the information managed by the Raspberry Pi.

4.6.3 Positioning Payload for Indoor Positioning The robotic platform includes a positioning payload where indoor positioning and mapping algorithms could be easily implemented and tested with the platform. The positioning payload (see Figure 15) is a development board with multiple connections where positioning algorithms can be easily implemented. It consists of a RaspberryPI board model B, an IMU (3-‐axis gyroscope, 3-‐axis accelerometer, 3-‐axis magnetometer, digital barometer) and a USB WiFi card (IEEE 802.11n, 802.11g, 802.11b). The RaspberryPI board has a CPU ARM11 @700 MHz featuring floating point ALU with Ethernet, USB 2.0, I2C, serial port and GPIOs, Linux O.S (Debian-‐based distribution) and dedicated HD camera


Security: Public

connector. The positioning payload is a cheap and easy-‐to-‐use system that allows WiFi RSS reading, IMU (inertial data) reading, Bluetooth 4.0/LE (small range location) with a CPU power (similar to an entry level smartphone). The overall system consists of the positioning payload and the data base. The Raspberry Pi microcontroller sends the RSS and inertial measurements to the data base of the server. These data are used by the positioning, tracking and data fusion methods to estimate the coordinates of the mobile robot. In Figure 15, schematic of the overall system is seen where the positioning payload reads RSS WiFi measurements and inertial data. Data fusion algorithms can be implemented and tested in this platform or logged into a database for off-‐line processing purposes. The OpenInLocation Lab allows researchers to test and validate their algorithms (implemented in the positioning payload) in the Gauss robot (see Figure 16). The mobile robot is used with the positioning payload to perform positioning in dynamic scenarios. It sends to the Raspberry Pi microcontroller its coordinates during the trajectory, which can be compared with the position estimate delivered by the algorithms to have an accurate benchmark of its accuracy. In the figure below, it can be observed the Gauss robot incorporating a WiFi dongle (elevated to emulate a smartphone carried by a human being), an IMU, and the positioning payload with a fusion algorithm.

Figure 15.- Schematic view of the positioning payload for a WiFi/IMU case.


Security: Public

Figure 16.- Picture of the overall test of a WiFI/IMU IPS algorithm mounted in the testing and validation Gauss platform. 4.6.4 Experiment description The indoor positioning testbed was used to assess the performance of Received Signal Strength (RSS)-based positioning algorithms based on WiFi technology. Particularly, real data was recorded using the demonstrator and processed by various algorithms.

One of the works that benefited from the use of the testbed investigated an enhanced path loss model for RSS measurements, the so-called two-slope path loss model is considered. It is a mathematical model of RSS that accounts for the dependence of path loss exponent and measurement variance with distance. It is composed of two models, each characterizing the RSS measurements when the mobile node and the corresponding AP are close or far, with respect to a breakpoint distance value. That is, for distances below the breakpoint the RSS measurement obeys a first model, and above such breakpoint distance a second one. This overcomes some practical limitations of the standard path loss formulation.

Considering this channel model, this contribution proposes a robust indoor localization solution based on a parallel architecture using a set of Interacting Multiple Models (IMM), each one involving two Extended Kalman filters (EKF) and dealing with the estimation of the distance to a given AP. Within each IMM, the two-slope path loss model parameters are sequentially estimated to provide a robust solution. Finally, the set of distance estimates are fused into a standard EKF-based solution to mobile target tracking. Notice that some very preliminary results related to this contribution have been presented in [26]. The contribution of this paper is a complete online two-slope channel calibration besides with a mobile target tracking using an IMM algorithm.


Security: Public

The use of real data (extracted from the testbed) is considered in this work as well to validate the computational simulations. The overall setup can be seen in Figure 17, where it can be observed that the robot (able to maneuver in a predetermined way) is equipped with a WiFi card that senses RSS from the APs in the building; this RSS measurements are sent to a remote server and stored, as well as the real trajectory that is accurately computed by the robot itself using wheel encoders and INS technology. The latter is used as a benchmark for the positioning algorithms that operate solely with the WiFi RSS data; as a final step, these RSS recordings are processed off-line to compute an estimated trajectory of the robot. An experiment was carried where the robot made a 15 meters long trajectory consisting on a straight line. Figure 18 shows the floor map of CTTC premises, where the experiment was performed. The robot moved left to right. Along with the true trajectory (black dotted line), the estimation results of two algorithms can be observed in the plot. The comparison is made between a tracking method considering the classical one-slope path loss model (black dashed line) and the proposed algorithm incorporating the two-slope path loss model (green line), the latter showing enhanced results.

Figure 17.- System diagram showing the connection of the various components.


Security: Public

Figure 18.- CTTC’s floor map. Anchor point locations and mobile path are marked. The trajectory of the robot consisted

in 15 meters in straight line. 4.7 JRA #G: Spectrum occupation measurements and database exploitation During the last year more measurements were carried out in the building of Barcelona, characterizing the received power in multiple points for the different floors. In addition, an interpolation strategy has been used to build the indoor REM based on the available measurements.

4.7.1 New measurements In Spain the process of spectrum release due to the digital dividend started in June 2014 and ended on 31st March 2015. Some TV stations have changed their channel and others have modified some parameters. For this reason, a new measurement campaign has been carried out in Barcelona to complement the measurements available in previous periods. Figure 19 shows the power received at the rooftop of D4 Building (located at UPC North Campus in Barcelona; latitude: 41º 23’ 20” N; longitude: 2º 6’ 43” E; altitude: 175 m), in January 2014 and in April 2015. Details of the measurement position and system configuration can be found in previous deliverable D21.2 “Preliminary tests over the lab infrastructures” [42]. Changes in the received power in the different channels can be appreciated in the figure. Channels with power density below -80.5 dBm/8 MHz have been assumed as free channels according to DVB-T reception values in [43] (TV-White Space).


Security: Public

-‐90

-‐80

-‐70

-‐60

-‐50

-‐40

-‐3021 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69

Power (d

Bm/8MHz

)

Channel

January 2014

April 2015

Figure 19. Changes in channel 26 received power after digital dividend changes in Spain Table 1 summarizes the channel occupancy in Barcelona, before, during and after that release. Table 1. TV Channels occupancy in Barcelona (Orange box indicates change and grey box indicates that the multiplex

does not exist).

TV Channels occupancy in Barcelona

Until 26/10/2014

Between 26/10/2014 and 31/03/2015

From 01/04/2015

STATE MULTIPLEXES MPE1 47 47 47 MPE2 27 27 27 MPE3 34 34 34 MPE4 29 29 MPE5 -‐ -‐ RGE1 64 31 and 64* 31 RGE2 31 41 41 SFN67 67 67 SFN68 68 68 SFN69 69 69 REGIONAL MULTIPLEXES MAUT1 61 44 and 61* 44 MAUT2 44 MAUTP 33 33 33 LOCAL MULTIPLEXES LOC1 (Barcelona) 26 26 26 LOC2 (Barcelona) 48 48 48 LOC3 (Cornellà) 53 53 53 LOC4 (Cornellà) 46 46 46

*Broadcast in Simulcast Mode


Security: Public

Some multiplexes have not changed the frequency, but its transmission parameters, like power. This is the case for channel 26 as shown in Figure 20. The figure shows 50 measurements of the power received in channel 26 (Central frequency 514 MHz) in December 2013 and in February 2015 at a known point on the first floor of the considered building. As can be seen, the power received from channel 26 in December 2013 was in average 4 dB higher than the power received from the same channel in February 2015. This change is explained by a reduction of the power transmitted in this channel from the tower of Collserola (our main TV transmitter).

Trace number

Channel 26 in point D4-115-C2

February 2015 December 2013

Rec

eive

d po

wer

(dBm

) / 8

MH

z

Figure 20. Changes in the received power of channel 26 after digital dividend changes in Spain

4.7.2 Interpolation of measurements to build the REM 4.7.2.1 Kriging Algorithm

The considered approach to build the REM of TVWS utilization uses methods based on spatial statistics. These methods take into account the measurements at specific locations and then the rest of REM coverage area at other positions is estimated as a function of that measured data. The principles of spatial statistics estimation methods are based on the basic principle that geographically near locations are more related to each other compared to the more distant locations. Several basic approaches exist in the literature for carrying out this estimation, such as Kriging, Inverse Distance Weighted and Nearest Neighbour Interpolation [44]. All three techniques estimate the values of interest at unexplored locations using weighted linear combination of measurements at known data points (called nodal points of the REM). The high precision of Kriging method if there are enough available nodal points makes this method one of the most commonly applied technique in the literature, so it has been selected in this work.

Kriging method [45], initially developed by Danie G. Krige, is an optimal interpolation based on regression against N observed values of surrounding data points. It estimates the


Security: Public

corresponding value of a certain point by an interpolation of weighted measured data points according to spatial covariance values. Near points in space tend to have closer values than farer points. In our case, the mean measured power of a certain TV channel in a certain point will be treated as an observed value and it will be used to compute the mean power of another non-measured point. The standard deviation of this new computed point will be also provided by the Kriging method, assuming a log-normal distribution very close to the statistical distribution of measured nodal points.

Several variants of the Kriging method exist depending on the data types to be estimated, namely simple Kriging, ordinary Kriging, Kriging with a trend, non-linear Kriging, co-kriging, etc. The variant considered here is the ordinary Kriging, in which the mean value of the data is unknown and the covariance function is known.

In particular, the ordinary Kriging method estimates the mean power of a TV channel at a point (x,y) by using the following expression:

(16)

Where: N is the total considered nodal points to be used in Kriging computation. This number is given by all the measured points that fall inside a circle of radius dmax (i.e. the so-called Kriging radius) around point (x,y)

P(xn,yn) corresponds to the measured received power of a TV channel in the nodal point (xn,yn).

λn is the weight of the nodal point n. This weight decreases with the distance between (xn,yn) and (x,y).

The algorithm must be unbiased. For that reason, all kriging weights (λn) have to sum 1, that is:

(17)

To find the optimal weight parameters λn it is necessary to get the covariance matrix (C) and the covariance vector (c) and use Lagrange method. The entire process is explained in [45]. The equation system to optimize is:

(18) where λ is the vector that includes the weights and the Lagrange parameter L of the optimization, that is:

(19)

The key of the Kriging method is the covariance function, which must be taken in correspondence with the nature of the problem to solve. The covariance function comes from the kind of semivariogram, the one that defines the dependency between the values of two nodal points. It is appropriate to take here a spherical covariance function, because


Security: Public

electromagnetic field propagation is also spherical. The covariance function has the following expression when the distance between two points is lower than dmax and 0 otherwise:

(20)

The value of dmax must be selected in accordance with the desired number of nodal points (N).

The covariance function helps to construct the covariance matrix C. This matrix will contain all the covariance values among all N nodal points to be considered. In addition, the matrix will contain and additional row and column with values equal to 1 except the lower right extreme with value equal to 0. Matrix C, thus, will be square of dimension N+1.

(21)

The same covariance function helps to make the covariance vector c. This vector will contain all covariance values between the point (x,y) to be interpolated and all N nodal points. In addition, it will contain an additional value equal to 1. Vector c has dimension N+1, too.

(22)

Solving equation system (18) by minimizing error variance using Lagrange method, all weight λ factors and the Lagrange parameter L are found. From this, the estimated mean received power of a TV channel at any point (x,y) can be found by computing expression (16).

4.7.2.2 Impact of the Kriging radius The Kriging radius dmax determines the nodal points that are considered in the interpolation process. In particular, the selected nodal points when estimating the received power at point (x,y) are those within distance dmax from this point. Obviously the larger the radius the more nodal points are considered. To illustrate this, Figure 21 depicts the nodal points considered to estimate position D4-115-A1 with Kriging radius ranging from 3.5m to 6.5m in steps of 0.5m.


Security: Public

Figure 21. Nodal points selected for kriging radius from 3.5m to 6.5m at position D4-115-A1

To assess the effect of the Kriging radius, let consider as a figure of merit the absolute error, which is the difference (in dB) between the value of received power estimated by the Kriging interpolation and the actual measured value at that point.

Figure 22 plots the absolute error for the point at position D4-115-A1 as a function of the Kriging radius. It can be observed that in general the error increases when increasing too much the Kriging radius, since we consider nodal points far away from the estimated point that disturb the estimation.

0123456789

10

3 3.5 4 4.5 5 5.5 6 6.5 7

Absolute error (d

B)

Radius (m)

Position D4-‐115-‐ A1

Channel 26

Channel 44

Figure 22. Absolute Error in terms of Kriging radius at point D4-115-A1 Figure 23 depicts the Root Mean Squared Error (RMSE) as another figure of merit at the same point D4-115-A1 and considering all the measurements at both channels 26 and 44. As expected the trend is the same as observed in the absolute error.


Security: Public

0

0.5

1

1.5

2

2.5

3

3.5

3 3.5 4 4.5 5 5.5 6 6.5 7

RMSE (d

B)

Radius (m)

RMSE in D4-‐115-‐A1

RMSE

Figure 23. RMSE in terms of Kriging radius at point D4-115-A1 After analyzing the behavior of the Kriging radius for different points inside the building we conclude that:

• The best results are obtained with a Kriging radius of 5 meters. It is found that this value is the minimum one that ensures that all the interpolated points have at least two nodal points.

• The use of distant nodal points worsens estimation. • The choice of the radius (dmax) for Kriging cannot be generalized, it depends on the

scenario being analyzed. 4.7.2.3 Resolution of Kriging The resolution used by the Kriging algorithm to estimate values in a given scenario is also an input of the interpolation procedure. Figure 24 shows the points estimated when a 1 meter resolution is selected, whereas Figure 25 shows the points needed for a 0.5 meter resolution.

Figure 24. Points estimated for 1 meter resolution in the first floor.


Security: Public

Figure 25. Points estimated for 0.5 meter resolution in the first floor.

Obviously, a low resolution requires longer processing time. To illustrate this, Table 2 collects some execution times on a computer with 8 GB of RAM, an Intel Core i5 processor and Windows 7 64-bit operating system, for 0.5m and 1m resolution and for 5m and 7m Kriging radius.

Table 2. Processing times depending on resolution and Kriging radius. Resolution (m) Kriging Radius(m) Processing time (s)

0.5 5 224 0.5 7 252 1 5 48 1 7 64

As it can be seen, the processing time increases slightly with the Kriging Radius, however it increases by four when reducing the resolution. 4.7.3 REM Graphical Interface We developed a graphical interface to represent the REM data and to observe the spectrum occupancy and the received power at different points of the building using the Kriging algorithm for power estimation.

In Figure 26 the received power at 1st floor of D4 building, located in the UPC Campus Nord in Barcelona (Spain) for TV channel 26 estimated using Kriging algorithm with the following values is shown:

• Grid resolution: 0.5m x 0.5m • Number of nodal points: 49 • Radius used in Kriging: 5m


Security: Public

Figure 26. Channel 26 received power at D4 first floor estimated with ordinary Kriging algorithm

In Figure 27 the TV white space information in channel 26 is shown at D4 first floor. As expected there is no chance to use that channel as it is busy.

Figure 27. Channel 26 TV White Space Information at D4 first floor

The graphical interface also calculates the absolute error and the RMSE. In Figure 28 the values for a Kriging radius of 5m are shown.

Figure 28. Example of error calculation for Kriging radius 5m.


Security: Public

4.8 JRA #H: Impact of channel model in the performance evaluation of wireless systems In this JRA, an experimental performance evaluation is carried out within a communication scenario that features two of the key enablers of 5G: (i) the use of post-OFDM modulations and (ii) an efficient use of the spectrum via spectrum sharing. The experimental lab set-up has been assembled so as to operate in conditions as realistic as possible via the actual real-time implementation of the involved transceivers and also via the utilization of propagation channels which have been recorded in a field measurement campaign and which are loaded in a channel emulator that also operates in real-time. The experimental results show that co-existence in a shared spectrum scenario is possible and that the performance degradation is kept at a low level as long as one of the two users is making use of spectrally agile post-OFDM modulations such as filterbank multicarrier (FBMC).

4.8.1 Motivation A joint effort involving the entire industry and academic community is making steady progress towards the definition and specification of 5G systems, providing as well early proof-of-concepts and prototypes of key technology enablers. This JRA focuses on two specific technology enablers of 5G wireless communication systems. The first is related with new spectrally efficient waveforms based on a filter bank multi-carrier (FBMC) scheme, which could complement or enhance the current 4G LTE technology. The second aims at exploiting existing underutilized licensed spectrum as a means to sustain the anticipated increased traffic demands of 5G systems.

These two 5G technology enablers were combined in a FBMC downlink (DL) communication system demonstrating promising results in terms of spectral efficiency and spectral contention. In this JRA we present an experimental analysis of the gains achieved when FBMC waveforms are used as secondary transmissions having as a goal to exploit the unused spectrum of other licensed narrowband primary transmissions. The assembled testbed allowed the performance evaluation of (a) a configurable DL FBMC system able to optimally utilize fragmented spectrum in a band where narrowband primary transmissions take place and (b) a point-to-point standalone DL FBMC system. Field-measured channels that were recorded employing a channel sounder in different indoor and outdoor configuration setups have also been used to increase the practical interest of the experimental analysis.

4.8.2 Target 5G scenario Due to the high complexity involved with the experimental validation, a simple, yet highly representative 5G operating scenario was considered; field measured channel data and a real-time DL communication system were used towards this end. The scenario features a primary narrowband professional mobile radio (PMR) transmission coexisting with an LTE-like opportunistic FBMC-based broadband transmission. Both systems are sharing the same frequency band and make use of a point-to-point DL communication link. Two different use cases were considered: i) the coexisting narrowband and broadband systems operate outdoors and ii) the primary system is located outdoors, with its receiver (Rx) located at its cell's edge and near an indoor FBMC transmitter (Tx; e.g., femtocell network). In both use cases, the FBMC signal had to be dynamically adapted so as not to interfere with the primary transmission. This was made feasible by deactivating subcarriers in the broadband transmission. The different channel conditions considered by the two scenarios required the acquisition of realistic outdoors-to-outdoors (O2O), indoors-to-indoors (I2I) and indoors-to-outdoors (I2O) field-measurements.


Security: Public

4.8.3 Field-measured channels The field-captures are based on three channel measurement campaigns carried out at the Université Catholique de Louvain (UCL), Louvain-la-Neuve, Belgium. The investigated outdoor environment was located in the campus area and consists of three-to-four storey office buildings, parking lots and relatively narrow streets. The area is characterized by small tree density and pedestrian traffic. The investigated indoor environment is a typical office building that consists of rooms along a corridor separated by brick walls.

Figure 29: Outdoor scenario satellite picture 4.8.4 Overview of the DL FBMC scheme

The FBMC system uses a fast-convolution (FC) multirate filter bank scheme, which yields an efficient and flexible FBMC implementation. The frame structure of the FBMC/OQAM DL signal is based on the equivalent LTE configuration for 1.4 MHz BW. 72 active subcarriers (i.e., 15 kHz spacing) are included in each FBMC symbol for a total 10 ms long radio frame comprising 150 FBMC symbols. Channel coding was not implemented and user-data symbols were mapped onto a 16-QAM constellation. A pilot pattern scheme similar to one used in LTE has been adopted. The precise values of the imaginary part of the latter were calculated using the surrounding data in order to compensate its coupling with the actual pilot symbol.


Security: Public

Figure 30: Lab set-up to demonstrate the co-existence of two systems in a shared spectrum scenario

4.8.5 Experimental results For each implemented testing scenario the BER was calculated by averaging the values obtained by 10.000 FBMC or PMR frames. It must be underlined that no channel coding is implemented in the FBMC system and consequently performance is measured in terms of uncoded (raw) BER.

Each testing scenario was defined by combining specific configuration values regarding: a) the relative subcarrier power levels between the PMR and FBMC signals, b) the utilized field-measured channel and c) the FBMC subcarrier allocation. With respect to the latter, three different FBMC signal configurations were considered in order to illustrate the achieved gains in a spectrum coexistence scenario: i) the FBMC system is quiet (i.e., no interference is caused to the standalone PMR system), ii) the FBMC transmission allocates all 72 active subcarriers (i.e., maximum possible interference) and iii) 70 subcarriers are allocated in the FBMC signal, thus providing a 30 kHz spectrum-hole to accommodate the existing PMR 12.5 kHz signal.

Figure 31: BER versus received signal power strength (dBm) in an outdoors scenario


Security: Public

Fig. 31 shows the performance at the PMR terminal when both coexisting wireless communication systems are deployed outdoors. Similarly, Fig. 32 considers an operating scenario where the FBMC system is deployed indoors. As it could be expected, minimal performance differences are observed in relation to the relative power levels of both coexisting transmission when a 30 kHz spectrum-hole is set in the FBMC signal.

Figure 32: BER versus received signal power strength (dBm) in an indoors scenario The performance of the PMR system is gradually degraded as the subcarrier power level of the PMR and FBMC signals is getting closer (without having counter-measures to tackle the DL interference). The figure provides a comparative insight of how the specific channel models that define a given operating scenario affect the performance of coexisting wireless communication systems. In both coexistence scenarios though, the high spectral contention of the FBMC scheme is experimentally verified, guarantying likewise a minimal interference to other inband signals. This is also clearly observed in the spectrum plot of the FBMC signal shown in Fig. 33, where a 35 dB drop in the spectrum-hole was created to accommodate the primary PMR signal.

Figure 33: Spectrum of the transmitted signal


Security: Public

Fig. 34 shows the performance that was observed at the FBMC receiver when considering the reverse situation from the one described before (i.e., interference from the DL PMR signal) for the all-outdoor operating scenario. As expected, a marginal BER gain is still attained when a spectrum-hole is used (i.e., a 12.5 kHz may only impair a small number of channels in the FBMC signal). An interesting comparison point is provided by the introduction of the BER curve of a standalone indoor FBMC system. As it is expected, in the deployed scenarios, the measured channels have a greater impact on the performance of the multicarrier FBMC transmission, compared to that provoked by the narrowband interference.

Figure 34: BER versus received signal power strength (dBm) for the FBMC secondary system 4.9 JRA #I: Hybrid radio channel modeling using graphs and rays 4.9.1 A Sub-band Divided Ray Tracing Algorithm Using the DPS Subspace in UWB Indoor Scenarios Sub-band divided ray tracing (SDRT) is one technique that has been extensively used to obtain the channel characteristics for ultra-wideband (UWB) radio wave propagation in realistic indoor environments. However, the computational complexity of SDRT scales directly with the number of sub-bands. Although we have proposed a low-complexity SDRT algorithm for one terminal position, the computational complexity i s still extremely high when involving multiple mobile terminal positions. Moreover, some indoor positioning techniques require for high positioning accuracy data from measurements/simulations with a very fine spatial resolution. To cope with this, we propose an algorithm to reduce the computational complexity of SDRT for multiple mobile terminal positions. The algorithm uses a projection of all propagation paths on a subspace spanned by two-dimensional discrete prolate spheroidal (DPS) sequences at each sub-band. It is important to note that, since the geometrical information of the propagation paths is the same in all sub-bands, the subspace dimension and basis coefficients in frequency dimension do not need to be recalculated at different sub-bands. We justify the simplifications of the proposed method by numerical simulations. Furthermore, we evaluate the effect of antenna characteristics on the proposed algorithm.


Security: Public

Our proposed algorithm reduces the computational complexity by more than one order of magnitude for indoor scenarios.

4.9.2 Hybrid Model For Reverberant Indoor Radio Channels Using Rays and Graphs Ray-tracing tools allow for deterministic simulation of the channel impulse response. Studies show that these tools work well when the impulse response consists only of a few distinct components. However, measurements of the channel impulse response in indoor environments show a diffuse tail. This diffuse tail is difficult to include in ray-tracing due to the computational complexity.

We proposed a hybrid model to include deterministic components and the diffuse tail by combining ray-tracing with a propagation graph. The recursive structure of the propagation graph allows for a computationally efficient calculation of the channel transfer function considering infinitely many components. We use ray-tracing and the theory of room electromagnetics to obtain the parameter settings for the propagation graph. Thus the proposed hybrid model does not require new or additional parameters in comparison to ray-tracing. Simulation results show good agreement with measurements with respect to the inclusion of the diffuse tail in both the delay power spectrum and the azimuth delay power spectrum.

4.9.3 Modeling of outdoor-to-indoor radio channels via propagation graphs Outdoor-to-indoor communication is frequently encountered in modern radio systems and it is therefore relevant to propose models to describe and simulate such scenarios. However, this poses the challenge to provide a model structure encompassing both the outdoor propagation as well as the reverberation phenomena occurring indoors while still maintaining a computational complexity low enough to make simulation feasible.

A framework for modeling reverberant radio channels has been developed in [46]–[48]. This framework relies on a socalled propagation graph in which vertices represent transmitters, scatterers, and receivers and edges represent the propagation conditions between vertices. The propagation graph formalism yields a closed-form expression for the infinite-bounce propagation enabling computer simulation of impulse responses of reverberant channels. Although originally proposed for static in-room channels where reverberation is particularly relevant, propagation graphs have been applied by several research groups to a diverse range of scenarios including works on body area channels [49], high-speed railway communications [50], distributed antenna systems deployed in indoors [51], and analysis of the MIMO channel’s rank properties and capacity [52]. With these diverse applications in mind, the propagation graph is an attractive approach for simulation of the heterogeneous outdoor-to-indoor channel.

In this paper we employ the propagation graph terminology to state a simulation model for the outdoor-to-indoor channel including the in-room reverberation phenomenon. We show how the particular outdoor-to-indoor scenario lead to a propagation graph and corresponding transfer function with a special structure. This structure can be used to reduce the computational complexity in simulation but also brings an important insight into how clusters may appear in the outdoor-to-indoor channel. We simulate the spatial channel correlation at the in-room receiver and compare to the in-room channel where transmitter and receiver are placed within the same room. The simulation study reveals that the envelope correlation for the channel may depend on the specific orientation of the considered array


Security: Public

4.9.4 A Sub-band Divided Ray Tracing Algorithm Using the DPS Subspace in UWB Indoor Scenarios Sub-band divided ray tracing (SDRT) is one technique that has been extensively used to obtain the channel characteristics for ultra-wideband (UWB) radio wave propagation in realistic indoor environments. However, the computational complexity of SDRT scales directly with the number of sub-bands. Although we have previously proposed a low-complexity SDRT algorithm for one terminal position, the computational complexity is still extremely high when involving multiple mobile terminal positions. Moreover, some indoor positioning techniques require for high positioning accuracy data from measurements/simulations with a very fine spatial resolution. To cope with this, we propose an algorithm to reduce the computational complexity of SDRT for multiple mobile terminal positions. The algorithm uses a projection of all propagation paths on a subspace spanned by two-dimensional discrete prolate spheroidal (DPS) sequences at each sub-band. It is important to note that, since the geometrical information of the propagation paths is the same in all sub-bands, the subspace dimension and basis coefficients in frequency dimension do not need to be recalculated at different sub-bands. We justify the simplifications of the proposed method by numerical simulations. Furthermore, we evaluate the effect of antenna characteristics on the proposed algorithm. Our proposed algorithm reduces the computational complexity by more than one order of magnitude for indoor scenarios.

4.10 References

[1] P. Kryszkiewicz, H. Bogucka ,"Out-of-Band Power Reduction in NC-OFDM with Optimized Cancellation Carriers Selection," Communications Letters, IEEE , vol.17, no.10, pp.1901-1904, October 2013

[2] P. Kryszkiewicz, A. Kliks, Y. Louet " Reduction of subcarriers spectrum sidelobes and intermodulation in NC-OFDM systems " WiMOB 2013.

[3] P. Kryszkiewicz, A. Kliks, H. Bogucka “Obtaining Low Out-Of-Band Emission Level of an NC-OFDM Waveform in the SDR Platform” ISWCS 2015

[4] M. Gharaibeh, Nonlinear Distortion in Wireless Systems: Modeling and Simulation with MATLAB. New York: John Wiley & Sons, 2011.

[5] C. L. Holloway, M. G. Cotton, and P. McKenna, “A model for predicting the popower delay profile characteristics inside a room,” IEEE Trans. Veh. Technol., vol. 48 (4), pp. pages 1110–1120, 1999.

[6] Bamba, M. T. Martinez-Ingles, D. P. Gaillot, E. Tanghe, B. Hanssens, J. M. Molina-Garcia-Pardo, M. Lienard, L. Martens, andW. Joseph, “Experimental investigation of electromagnetic reverberation characteristics as a function of uwb frequencies,” IEEE Antennas Wireless Propag. Lett., 2014, dOI: 10.1109/LAWP.2014.2382873.

[7] Bamba, W. Joseph, J. B. Andersen, E. Tanghe, G. Vermeeren, D. Plets, J. O. Nielsen, and L. Martens, “Experimental assessment of specific absorption rate using room ectromagnetics,” IEEE Trans. Electromagn. Compat., vol. 54 (4), pp. pages 747–757, 2012.

[8] E. Vinogradov, W. Joseph, and C. Oestges, “Measurement-based modeling of time-variant fading statistics in indoor peer-to-peer scenarios,” Antennas and Propagation, IEEE Transactions on, vol. 63, no. 5, pp. 2252–2263, May 2015.

[9] Patel, G. Stuber, and T. Pratt, “Simulation of Rayleigh-faded mobile-to-mobile communication channels,” IEEE Transactions on Communications, vol. 53, no. 11, pp. 1876–1884, Nov 2005.

[10] V. Degli-Esposti, “A diffuse scattering model for urban propagation prediction,” IEEE Transactions on Antennas and Propagation, Vol. 49, No. 7, 2001, pp. 1111-1113.


Security: Public

[11] V. Degli-Esposti, F. Fuschini, E. Vitucci, G. Falciasecca, “Measurement and modelling of scattering from buildings,” IEEE Transactions on Antennas and Propagation, Vol. 55 No 1, 2007, pp. 143-153.

[12] E. M. Vitucci, F. Mani, V. Degli-Esposti, and C. Oestges, “Study of a Polarimetric Model for Diffuse Scattering in Urban Environment,” 6th European Conference on Antennas and Propagation, Prague, Czech Republic, 2012.

[13] F. Mani, F. Quitin, and C. Oestges, “Directional spreads of dense multipath components in indoor environments: Experimental validation of a ray-tracing approach,” IEEE Transactions on Antennas and Propagation, Vol. 60, No. 7, 2012, pp. 3389-3396.

[14] M. Gan, P. Meissner, F. Mani, E. Leitinger, M. Frohle, C. Oestges, K. Witrisal, T. Zemen, Low-complexity sub-band divided ray tracing for UWB indoor channels, 2014 IEEE Wireless Communications and Networking Conference (WCNC), Istanbul, Turkey, 2014.

[15] G. Steinböck, T. Pedersen, B. Fleury, W. Wang, and R. Raulefs, “Experimental Validation of the Reverberation Effect in Room Electromagnetics,” IEEE Trans. Antennas Propag., (revised),2014.

[16] C. Holloway, M. Cotton, and P. McKenna, “A Model for Predicting the Power Delay Profile Characteristics Inside a Room,” IEEE Trans. Veh. Technol., vol. 48, no. 4, pp. 1110–1120, 1999.

[17] J. Andersen, J. Nielsen, G. Pedersen, G. Bauch, and J. Herdin, “Room electromagnetics,” IEEE Antennas Propag. Mag., vol. 49, no. 2, pp. 27– 33, 2007.

[18] D. A. Hill, Electromagnetic Fields in Cavities: Deterministic and Statistical Theories, ser. IEEE Press Series on Electromagnetic Wave Theory. Piscataway, NJ: Wiley/IEEE Press, 2009.

[19] S. Fujimoto and T. Watanabe, “Studies on the Body Surface Area of Japanese,” Acta medica Nagasakiensia, vol. 14, no. 1-2, pp. 1–13, 1969.

[20] D. Du Bois and E. F. Du Bois, “A formula to estimate the approximate surface area if height and weight be known,” Nutrition (Burbank, Los Angeles County, Calif.), vol. 5, no. 5, p. 303, 1989.

[21] R. Mosteller, “Simplified Calculation of Body-Surface Area,” New England Journal of Medicine, vol. 317, no. 17, pp. 1098–1098, 1987, pMID: 3657876.

[22] A. Hirata, Y. Nagaya, F. Osamu, A. Nagaoka, and S. Watanabe, “Correlation Between Absorption Cross Section and Body Surface Area of Human for Far-Field Exposure at GHz Bands,” in Electromagnetic Compatibility, 2007. EMC 2007. IEEE International Symposium on, Jul. 2007, pp. 1–4.

[23] J. Andersen, K. L. Chee, M. Jacob, G. Pedersen, and T. Kurner, “Reverberation and Absorption in an Aircraft Cabin With the Impact of Passengers,” IEEE Trans. Antennas Propag., vol. 60, no. 5, pp. 2472– 2480, May 2012.

[24] A. Bamba, W. Joseph, J. Andersen, E. Tanghe, G. Vermeeren, D. Plets, J. Nielsen, and L. Martens, “Experimental Assessment of Specific Absorption Rate Using Room Electromagnetics,” IEEE Trans. Electromagn. Compat., vol. 54, no. 4, pp. 747–757, Aug. 2012.

[25] G. Steinböck, T. Pedersen, B. Fleury, W. Wang, and R. Raulefs, “Distance Dependent Model for the Delay Power Spectrum of In-room Reverberant Channels,” IEEE Trans. Antennas Propag., vol. 61, no. 8, pp. 4327–4340, Aug. 2013.

[26] C. Mensing, “Location Determination in OFDM Based Mobile Radio Systems,” Ph.D. dissertation, Technische Universit¨at M¨unchen, 2013.

[27] J. Borras, P. Hatrack, and N. Mandayam, “Decision theoretic framework for NLOS identification,” in Proc. IEEE Int. Technol. Conf., vol. 2, May 1998, pp. 1583–1587 vol.2.

[28] D. Dardari, A. Conti, U. Ferner, A. Giorgetti, and M. Z. Win, “Ranging with ultrawide bandwidth signals in multi-path environments,” Proc. IEEE, vol. 97, no. 2, pp. 404–426, 2009.


Security: Public

[29] R. Zekavat and R. M. Buehrer, Handbook of Position Location: Theory, Practice and Advances, ser. IEEE Series on Digital & Mobile Communication. Wiley, 2011.

[30] J. Khodjaev, Y. Park, and A. Saeed Malik, “Survey of NLOS identification and error mitigation problems in UWB-based positioning algorithms for dense environments,” Annals of telecommunications, vol. 65, no. 5-6, pp. 301–311, 2010.

[31] J. Shen and A. F. Molisch, “Indirect path detection based on wireless propagation measurements,” IEEE Trans. Wireless Commun., vol. 11, no. 12, pp. 4482–4493, Dec. 2012.

[32] Z. Xiao, H. Wen, A. Markham, N. Trigoni, P. Blunsom, and J. Frolik, “Non-line-of-sight identification and mitigation using received signal strength,” IEEE Trans. Wireless Commun., no. 99, pp. 1–1, Nov. 2014.

[33] Junyang Shen and A. F. Molisch, “Indirect path detection based on wireless propagation measurements,” IEEE Trans. Wireless Commun., vol. 11, no. 12, pp. 4482–4493, Dec. 2012.

[34] I. Guvenc and Chia-Chin Chong, “A survey on TOA based wireless localization and NLOS mitigation techniques,” IEEE Communications Surveys Tutorials, vol. 11, no. 3, pp. 107–124, Aug. 2009.

[35] S. Marano, W. Gifford, H. Wymeersch, and M. Win, “NLOS identification and mitigation for localization based on UWB experimental data,” IEEE J. Sel. Areas Commun., vol. 28, no. 7, pp. 1026–1035, 2010.

[36] B. Denis, J. Keignart, and N. Daniele, “Impact of NLOS propagation upon ranging precision in UWB systems,” in IEEE Conference on Ultra Wideband Systems and Technologies, Nov. 2003, pp. 379–383.

[37] H. Wymeersch, S. Marano, W. Gifford, and M. Z. Win, “A machine learning approach to ranging error mitigation for UWB localization,” IEEE Trans. Commun., vol. 60, no. 6, pp. 1719–1728, 2012.

[38] Y. Qi, K. Hisashi, and H. Suda, “On time-of-arrival positioning in a multipath environment,” IEEE Trans. Veh. Technol., vol. 55, no. 5, pp. 1516–1526, 2006.

[39] L. Jing, T. Pedersen, and B. Fleury, “Direct ranging in multi-path channels using OFDM pilot signals,” in 15th IEEE Int. Symposium on Signal Process. Advances in Wireless Commun., Jun. 2014, pp. 150–154.

[40] D. Wang, M. Fattouche, and F. Ghannouchi, “Fundamental limit of OFDM range estimation in a separable multipath environment,” Circuits, Systems, and Signal Processing, vol. 31, no. 3, 2012.

[41] R. Adam and P. A. Hoeher, “Semi-blind channel estimation for joint communication and positioning,” 10th Workshop on Positioning Navigation and Commun., pp. 1–5, 2013.

[42] Miquel Payaró, et al. “Preliminary tests over the lab infrastructures”, Deliverable D21.2 of the NEWCOM# project, November, 2013.

[43] H. Aïache, et al. “Use-cases Analysis and TVWS Systems Requirements”, Deliverable D3.1 of the COGEU project, August, 2010.

[44] H. Birkan Yilmaz, Tuna Tugcu, Fatih Alagöz and Suzan Bayhan. “Radio Environment Map as Enabler for Practical Cognitive Radio Networks". IEEE Communications Magazine, Electrical and Electronics Engineers, December 2013.

[45] Geoff Bohling. “Kriging". Kansas Geological Survey, C&PE 940, October 2005. [46] T. Pedersen and B. Fleury, “Radio channel modelling using stochastic propagation

graphs,” in Proc. IEEE International Conf. on Communications ICC ’07, Jun. 2007, pp. 2733–2738.

[47] T. Pedersen, G. Steinb¨ock, and B. H. Fleury, “Modeling of reverberant radio channels using propagation graphs,” IEEE Trans. Antennas Propag., vol. 60, no. 12, pp. 5978–5988, Dec. 2012.

[48] G. Steinb¨ock, T. Pedersen, B. Fleury, W. Wang, and R. Raulefs, “Calibration of the Propagation Graph Model in Reverberant Rooms,” in URSI Commission F Triennial Open Symposium on Radiowave Propagation and Remote Sensing, May


Security: Public

2013. [49] R. Zhang, X. Lu, Z. Zhong, and L. Cai, “A study on spatial-temporal dynamics

properties of indoor wireless channels,” in Wireless Algorithms, Systems, and Applications, ser. Lecture Notes in Computer Science, Springer Berlin Heidelberg, 2011, vol. 6843, pp. 410–421.

[50] L. Tian, X. Yin, Q. Zuo, J. Zhou, Z. Zhong, and S. Lu, “Channel modeling based on random propagation graphs for high speed railway scenarios,” in Personal Indoor and Mobile Radio Communications (PIMRC), 2012 IEEE 23rd International Symposium on, Sep. 2012, pp. 1746–1750.

[51] L. Tian, X. Yin, X. Zhou, and Q. Zuo, “Spatial cross-correlation modeling for propagation channels in indoor distributed antenna systems,” EURASIP Journal on Wireless Communications and Networking, vol. 2013, no. 1, pp. 1–11, 2013.

[52] O. Souihli and T. Ohtsuki, “Benefits of rich scattering in MIMO channels: A graph-theoretical perspective,” Commun. Letters, IEEE, vol. 17, no. 1, pp. 23–26, Jan. 2013.


Security: Public

Comments and suggestions for the improvement of this document are most welcome and should be sent to:

[email protected]

________________________________________________________________________

http://www.newcom-project.eu

d21.4 v1.0 final - cordis | european commission · the research activities of the “radio...

Documents