Reproduce “V2X-DSC: Multi-Agent Collaborative Perception with Distributed Source Coding Guided Communication” 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

Yuankun Zeng, Shaohui Li, Zhi Li, Shulan Ruan, Yu Liu, You He

V2X-DSC introduces a multi-agent collaborative perception framework that leverages Distributed Source Coding to enable efficient intermediate feature fusion, achieving high 3D object detection accuracy with kilobyte-level communication payloads on real-world V2X datasets.

Key insights

  • Intermediate BEV features from collaborating agents exhibit strong statistical correlation due to observing overlapping parts of the same physical world.
  • Distributed Source Coding principles, particularly Wyner-Ziv coding, provide a theoretically sound method to exploit this correlation by using the receiver's local perception as side information to achieve higher compression rates.
  • By transmitting only the 'innovation' or complementary information not inferable from the receiver's local context, communication efficiency can be drastically improved without significant loss in perception accuracy.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.