Nicolai Czink Laura Bernadó, Roland Tresch, Alexander Paier, Thomas Zemen Bringing Robustness into 5.9 GHz ITS WAVE Systems FTW Forschungszentrum Telekommunikation Wien E-Mail: czink@ftw.at 2nd ETSI TC ITS Workshop, 10-12 February 2010 ETSI, Sophia Antipolis, France
Outline Motivation IEEE 802.11p (draft) road tests - Setup - Performance results - Robustness issues Transceiver performance - Conventional receivers - High-complexity implementation - Standard improvements to reduce complexity Conclusions - 2 -
Motivation In vehicle-to-vehicle and vehicle-to-infrastructure scenarios, safety-related messages must be passed on - Robustly - In real-time Currently, these issues are only addressed on the MAC layer - Acknowledgements - Retransmissions However, the main shortcoming lies on the PHY layer - High bit/frame error rates lead to many retransmissions - Strongly depends on environment - Worst case: non-line-of-sight - 3 -
WAVE Measurements Vomp, Tyrol, Austria Freeway test track from ASFINAG was available: - Gantries, power supply, LAN connection - 4 -
Environments -5-
Measurement Equipment Radio equipment: - CVIS testbed (including WAVE and GPS) Antennas: - 5.9 GHz omni-directional monopoles at the RSUs - CVIS antenna at the OBU Test vehicle: - ASFINAG minivan Additionally: - Video and photo equipment GPS WLAN 2-6 GHz - 6 -
Performance Measurements Throughput measurements over distance from RSU Measure: - Frame Success Ratio - 7 -
Unexpected Throughput Drops Performance jitter Throughput drops Non-line-of-sight (NLOS) environment Obstruction by other cars Radio receiver cannot cope with the channel effects - 8 -
Wave Transceiver Architecture Transmitter Encoder Interleaver Mapper Receiver Decoder Deinterleaver Demapper Equalizer Estimator Pilot insert Modulator Pilot remove Demodulator Radio Channel + Noise - 9 -
Pilot Distribution vs. Channel Variation Bandwidth 10 MHz OFDM symbol duration 8 µs Number of 52 + subcarriers 4 pilot -5-10 -15-20 -25-30 -35 Channel is changing significantly during one frame Pilot symbols at the beginning are suboptimal! - 10 -
Performance of simple channel estimators PER = 100% PER ~ 70-90% PER < 10 % NLOS channel model (worst case), L = 200 bytes, 3 Mbit/s Least squares channel estimation, hard decision Higher velocities lead to increased BER floors BER = 10-2 only for speed 50 km/h - 11 -
Advanced Receiver Architectures 10 0 10-1 1 iteration BER 10-2 2 iterations 10-3 0 5 10 15 20 E b /N 0 [db] 4 iterations 8 iterations 12 iterations 14 iterations PER < 10 % NLOS channel model (worst case), L = 200 bytes, 3 Mbit/s, v = 150 km/h High-complexity iterative receiver Performance depends on number of iterations - 12 -
Improving Pilot Structure Distributing pilots across time and frequencies - Midamble - Optimized placement (e.g. LTE-like or WiMAX-like) - 13 -
Complexity Reduction by Improved Pilot Pattern 10 0 10-1 1 iteration, old pattern BER 10-2 10-3 0 5 10 15 20 E b /N 0 [db] 1 iteration, midamble 2 iterations, midamble 14 iterations, old pattern More reasonable pilot placement leads to tremendous complexity reduction on chip! - 14 -
Conclusions The PHY layer of the IEEE 802.11p (draft) standard is not designed for mobile terminals In non-line-of-sight environments: - Performance problems with conventional receivers (demonstrated by measurements and simulations) - To circumvent the problem: very high receiver complexity is needed - An improved pilot pattern leads to much lower receiver complexity To establish a robust link - A new chip design must be done anyway - Why not do it with low complexity and an improved pilot pattern? - 15 -