Reproduce “InfoCom: Kilobyte-Scale Communication-Efficient Collaborative Perception with Information Bottleneck” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Quanmin Wei, Penglin Dai, Wei Li, Bingyi Liu, Xiao Wu
InfoCom, a framework for collaborative perception, achieves kilobyte-scale data transmission between agents, dramatically reducing communication bandwidth requirements. It maintains near-lossless perception performance in autonomous driving by employing an information purification paradigm grounded in extended Information Bottleneck principles.
Key insights
- Extreme communication efficiency (kilobyte-scale) in collaborative perception is achievable without significant performance degradation by purifying and transmitting only task-critical semantic and spatial information.
- An extended Information Bottleneck principle can provide a robust theoretical foundation for optimizing communication-performance trade-offs under severe bandwidth constraints, moving beyond heuristic designs.
- Decoupling low-dimensional information-aware features (E) from sparse, quantized spatial masks (M) allows for both radical compression and accurate feature reconstruction at the receiver with minimal overhead.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.