Communication-Efficient Collaborative Perception
Overview
Vehicles, robots, and drones use collaborative perception to reveal objects or scene details that a single agent may miss, especially under occlusion. To limit bandwidth, systems selectively transmit salient regions, instances, or object-level messages, or compress features with sparse encoding, quantization, and learned reconstruction; agents then fuse the received evidence. Work is evaluated on cooperative driving benchmarks such as OPV2V and DAIR-V2X, with increasing attention to heterogeneous sensors, latency, packet loss, and real-world deployment. The central tension is how to preserve complementary, well-aligned evidence and robust performance when communication budgets and network conditions are tight.
Latest papers
26 papers4 papers in the last four weeks, down 33% on the four weeks before. 0.0% of all new papers.