On Prototyping IEEE 802.11p Channel Estimators in Real-World Environments using GNURadio

On Prototyping IEEE802.11p ChannelEstimators in Real-World Environments

Using GNURadio

Razvan-Andrei Stoica, Stefano Severi, Giuseppe [email protected]

Focus Area Mobility - Jacobs University Bremen (GERMANY)

June 19, 2016

Vehicular Ad-hoc Networks in ITS

CCP-IV 16 IEEE802.11p Channel Estimators Prototyping with GNURadio June 19, 2016 2/1

Wireless Channel in VANETs: Relevance

ITS Applications require robust wireless communications.⇓

Knowledge of the wireless channel is fundamental!


Wireless Channel in VANETs: Relevance

ITS Applications require robust wireless communications.⇓

Knowledge of the wireless channel is fundamental!⇓

Ranging may be refined and improved using ChannelState Information (CSI).


Wireless Channel in VANETs: Characteristics

Vehicular WiFi [2]↔ IEEE 802.11p [3].

Key Challenges:

Non-stationary channel distributions,

Short coherence times,

Short / medium link availability,

Frequency selective channels,

Fast fading and shadowing.

[2] ETSI EN 302 665 V1.1.1, European Standard (Telecommunications series): Intelligent Transport Systems (ITS); Communications Architecture, September 2010.

[3] IEEE Std. 802.11p-2010: “IEEE Standard for Information Technology - Telecommunications and Information Exchange Between Systems - Local and MetropolitanArea Networks - Specific Requirements, Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications, Amendment 6: WirelessAccess in Vehicular Environments” , July 2010.


Proposal Overview

Objective

Design a new practical channel estimation scheme andprototype its real-world deployment with Software DefinedRadio (SDR).

Strategy

Implement the SOF algorithm[4] in GNURadio andcontribute to an open-source community of engineers andenthusiats.

[4] R. A. Stoica, S. Severi, and G. T. F. de Abreu, “Learning the vehicular channel through the self-organization of frequencies,” in Vehicular Networking Conference(VNC), 2015 IEEE, Dec 2015, pp. 68-75.


GNURadio in a Nutshell

What is it?

Open-source community driven framework for SDRs[5].“Linux” for Software Defined Radios and Digital SignalProcessing.

[5] GNU Radio Website, accessed June 2016. [Online]. Available at: http://www.gnuradio.org


http://www.gnuradio.org

GNURadio in a Nutshell

Why use GNURadio?

open-source and free

large tutorial pool and thorough documentation

flowchart based modeling

many signal processing blocks implementated by default

C++ / Python based block programming

rapid prototyping through software

modular packaging of custom solutions

easy blocks and code reuse

SDR hardware prices go down

hardware reuse for different projects[5] GNU Radio Website, accessed June 2016. [Online]. Available at: http://www.gnuradio.org


http://www.gnuradio.org

GNURadio IEEE 802.11p Rx Prototyping

How to use GNURadio in given context?

Start from existent module of IEEE 801.11p transceiver[6],[7], integrate and test researched channel estimationsolution on the Rx side.

[6] B. Bloessl, M. Segata, C. Sommer, and F. Dressler, “Towards an open source IEEE 802.11p stack: A full SDR-based transceiver in GNURadio,” in VehicularNetworking Conference (VNC), 2013 IEEE, Dec 2013, pp. 143-149.[7] B. Bloessl, gr-ieee802-11 GNURadio module, accessed June 2016. [Online]. Available at: https://github.com/bastibl/gr-ieee802-11.


https://github.com/bastibl/gr-ieee802-11

GNURadio IEEE 802.11p Rx Prototyping

How to use GNURadio in given context?

Start from existent module of IEEE 801.11p transceiver[6],[7], integrate and test researched channel estimationsolution on the Rx side.

[6] B. Bloessl, M. Segata, C. Sommer, and F. Dressler, “Towards an open source IEEE 802.11p stack: A full SDR-based transceiver in GNURadio,” in VehicularNetworking Conference (VNC), 2013 IEEE, Dec 2013, pp. 143-149.[7] B. Bloessl, gr-ieee802-11 GNURadio module, accessed June 2016. [Online]. Available at: https://github.com/bastibl/gr-ieee802-11.


https://github.com/bastibl/gr-ieee802-11

The SOF Algorithm - Overview

What is the SOF ?

Improved channel estimator compliant with IEEE 802.11pOFDM PHY specifications and packet format [3].

Why use SOF ?

Superior performance than typical LS estimation and SotAsequential schemes (e.g. STA [8], CDP [9]) and lowercomplexity and latency than iterative schemes (e.g.soft-input soft-output turbo decoder [10]) [4].

[8] J. A. Fernandez, K. Borries, L. Cheng, B. Vijaya Kumar, D. D. Stancil, and F. Bai, “Performance of the 802.11p physical layer in vehicle-to-vehicle environments,”Vehicular Technology, IEEE Transactions on, vol. 61, no. 1, pp. 3-14, 2012.

[9] Z. Zhao, X. Cheng, M. Wen, L. Yang, and B. Jiao, “Constructed Data Pilot-Assisted Channel Estimators for Mobile Environments,” IEEE Transactions on IntelligentTransportation Systems, vol. 16, no. 2, pp. 947–957, April 2015.

[10] P. Alexander, D. Haley, and A. Grant, “Cooperative intelligent transport systems: 5.9-ghz field trials,” Proceedings of the IEEE, vol. 99, no. 7, pp. 1213-1235, July2011.


The SOF Algorithm - How does it work ?


The SOF Algorithm - Practical Amendments

Prototyping with GNURadio revealed two “issues”:high SNR performance improvement possible by adaptiveMA filteringconstellation drifting due to sequential estimation and lowresolution comb pilots


The SOF Algorithm - Practical AmendmentsMA Adaptive Filtering Optimization

Problem

MA filtering refines LS estimates for low and mediumSNRs, but decays performance in high SNR regimes.

Solution

Linearly adapt the MA filtering window. Use a widewindow (max. 11 taps) for low SNRs and decrease thewidth up to just one tap (no filtering) for high SNRs.

Adapt filtering window by thresholding — up to 30 dB allowMA and then turn it off upon results in [4].

σ(ρ)=

2.79, ρ ≤ 0 dB−2.33

30 ρ+ 2.79, ρ ∈ (0, 30] dB0.46, ρ > 30 dB

. (1)


The SOF Algorithm - Practical AmendmentsConstellation Drifting Fix

Problem

IEEE 802.11p comb pilots cannot lock and correct thephase offset of the symbols constellation→ errors.

Solution

Correlate previous CFR estimate with current one toestimate and correct the phase offset.

φcorr(i) = ∠

( ∑Nsc−1k=0 HSOF,i(k)H

∗SOF,i−1(k)

Nsc · |∑Nsc−1

k=0 HSOF,i(k)H∗SOF,i−1(k)|

), (2)

HcorrSOF,i(k) , HSOF,i(k) · exp(−jφcorr(i)). (3)

Works due to initial training block→ fixed reference locked.CCP-IV 16 IEEE802.11p Channel Estimators Prototyping with GNURadio June 19, 2016 15/1

The SOF Algorithm - Pseudocode

1: Compute initial LS estimate from training symbols: H0

2: Estimate SNR: ρ3: Compute filter window: MA(·)4: Filter initial estimate: HSOF,0 = MA(H0)5: for symbol i = 1 to F do6: Get OFDM symbol estimate: Xi = Yi/HSOF,i−1

7: Equalize OFDM Tx symbol: XTX,i 7−→ Xi

8: Estimate current symbol SNR: ρ9: Compute filter window: MA(·)

10: Filter intermediate estimate: Hi = MA(Hi)11: Update SOF channel estimate:

HSOF,i = HSOF,i−1 + γ(Hi − HSOF,i−1)12: Correct phase rotation:

φcorr(i), HSOF,i ← HSOF,i · exp(−jφcorr(i))13: end forCCP-IV 16 IEEE802.11p Channel Estimators Prototyping with GNURadio June 19, 2016 16/1

Testing the Implementation

Strategy

Perform unit tests and system tests.

unit tests↔ loopback Tx-Rx chain targeting channelestimation algorithms.

system tests↔ real-world Tx-Rx communication focusingon packet delivery and general operation.


Unit Testing

loopback GNURadio interfaces (does not require RF HW);builds on top of IEEE 802.11p transceiver [7];similar to MATLAB simulations;computes FER at different SNRs on emulated channels;implements 2 vehicular channels [11].

Channel v[km/h] Doppler [Hz] Max. τs[µs]V2V

express-wayoncoming300-400 m

210 1000-1200 0.3

V2Iurban canyon

100 m32-48 300 0.5

[11] G. Acosta-Marum and M. A. Ingram, “Six Time-and Frequency-selective empirical channel models for vehicular wireless LANs,” Vehicular Technology Magazine,IEEE, vol. 2, no. 4, pp. 4-11, 2007.


Performance Evaluation - FER 100B PSDUV2V(top), V2I(bottom) @ QPSK 1/2

0 5 10 15 20 25 30 3510

−2

10−1

100

Es/N0 dB

FER

LSCDPSOF

0 5 10 15 20 25 30 3510

−2

10−1

100

Es/N0 dB

FER

LSCDPSOF


Performance Evaluation - FER 300B PSDUV2V(top), V2I(bottom) @ QPSK 1/2

0 5 10 15 20 25 30 3510

−2

10−1

100

Es/N0 dB

FER

LSCDPSOF

0 5 10 15 20 25 30 3510

−2

10−1

100

Es/N0 dB

FER

LSCDPSOF


System Testing

runs real-time communication (requires RF HW [12],[13]);uses implemented IEEE 802.11p Rx;computes FER based on short-term statistics;displays results in real-time GUI for evaluation and debug;tested only indoor (for the moment).

[11] Cohda Wireless, accessed June 2016. [Online]. Available at: http://cohdawireless.com/Portals/0/MK5_OBU_10122015.pdf

[12] Ettus Research, accessed June 2016. [Online]. Available at:https://www.ettus.com/content/files/07495_Ettus_N200-210_DS_Flyer_HR_1.pdf.


http://cohdawireless.com/Portals/0/MK5_OBU_10122015.pdf

https://www.ettus.com/content/files/07495_Ettus_N200-210_DS_Flyer_HR_1.pdf

Performance Evaluation - Indoor TestingReal-Time OFDM Equalized Symbol Snapshot


Conclusions and Future Work

Take Home Points

SDRs: future of modular multi-purpose radio design

GNURadio: easy-to-use “Linux” of SDRs

CSI Knowledge: key of robust communication and ranging

SOF: IEEE 802.11 p(/a/g)-compliant, performant andlow-complexity channel estimation algorithm

Next Steps

SOF Optimization: antenna diversity anddeparameterization by channel correlations

Field Trials: setup outdoor V2X test framework of SDR vs.reference devices (e.g. Cohda Mk5)


Thank you!

Engineering

On Prototyping IEEE 802.11p Channel Estimators in Real-World Environments using GNURadio