SRA-CP: Risk-Aware Car-to-Car Perception That Shares Only What Matters
Cars that share only what matters
Today’s cooperative perception lets connected cars see around corners by sharing sensor data—but it often floods the airwaves with non-critical info and assumes fixed partners. SRA-CP fixes both.
- Risk-aware: Each car detects occlusion-driven blind zones that threaten its current driving task.
- Spontaneous and selective: Vehicles broadcast tiny coverage summaries, then trigger targeted sharing only when needed.
- Smart teaming: The ego car picks peers that best cover the blind zone.
- Bandwidth-savvy fusion: Exchanges prioritize safety-critical objects and adapt to bandwidth limits.
Results: On a public dataset, SRA-CP keeps safety-critical detection within <1% of full generic CP while using only ~20% of the bandwidth, and improves performance by 15% over selective CP methods without risk awareness.
Why it matters: scalable V2X perception that boosts safety in dynamic traffic without overwhelming networks.
Paper: https://arxiv.org/abs/2511.17461v1 • Authors: Jiaxi Liu, Chengyuan Ma, Hang Zhou, Weizhe Tang, Shixiao Liang, Haoyang Ding, Xiaopeng Li, Bin Ran
Paper: https://arxiv.org/abs/2511.17461v1
Register: https://www.AiFeta.com
#AutonomousDriving #V2X #CooperativePerception #ConnectedVehicles #AI #Safety #Robotics #EdgeAI
