Reproduce “CATS-V2V: A Real-World Vehicle-to-Vehicle Cooperative Perception Dataset with Complex Adverse Traffic Scenarios” 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

Hangyu Li, Bofeng Cao, Zhaohui Liang, Wuzhen Li, Juyoung Oh, Yuxuan Chen, Shixiao Liang, Hang Zhou and 11 more

CATS-V2V presents the first real-world vehicle-to-vehicle cooperative perception dataset, meticulously collected across various adverse traffic scenarios and featuring 1 ms hardware synchronization and target-based temporal alignment. This multi-modal dataset offers high-quality annotations to facilitate the creation of robust autonomous driving systems capable of effective operation in challenging real-world conditions.

Key insights

  • Real-world adverse traffic scenarios, though constituting a 'long-tail' distribution, are safety-critical and necessitate dedicated, high-quality data for robust autonomous driving system development.
  • Achieving precise hardware time synchronization (1 ms accuracy) and advanced temporal alignment (target-based) is crucial for accurate multi-modal and multi-agent data fusion in cooperative perception.
  • Cooperative perception, when supported by high-fidelity and well-synchronized data, offers a potent solution to overcome individual sensor limitations and enhance situational awareness in challenging environments.

Summary by alphaXiv.

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