CooperDrive: Enhancing Driving Decisions through Cooperative Perception
Deyuan Qu, Qi Chen, Onur Altintas, Takayuki Shimizu
AI summary
Problem
Autonomous vehicles struggle with delayed hazard recognition and increased collision risks in occluded or non-line-of-sight scenarios, while existing cooperative perception research lacks real-world validation and direct integration with downstream planning.
Approach
CooperDrive shares lightweight object-level Bird’s-Eye View detections between vehicles, allowing standard planners to anticipate conflicts and adjust trajectories proactively without modifying existing architectures.
Key results
- First real-vehicle closed-loop validation of a bandwidth-efficient cooperative driving system
- Novel multi-task perception network unifying 3D detection and semantic localization via shared BEV features
- Increased reaction lead time, minimum time-to-collision, and stopping margin in occlusion-heavy intersections
- Operates at only 90 kbps bandwidth with 89 ms average end-to-end latency
Why it matters
It provides practical evidence that cooperative perception can directly and safely enhance autonomous driving decisions without overhauling existing vehicle architectures.
Abstract
Autonomous vehicles equipped with robust on- board perception, localization, and planning still face limita- tions in occlusion and non-line-of-sight (NLOS) scenarios, where delayed reactions can increase collision risk. We propose Coop- erDrive, a cooperative perception framework that augments sit- uational awareness and enables earlier, safer driving decisions. CooperDrive offers two key advantages: (i) each vehicle retains its native perception, localization, and planning stack, and (ii) a lightweight object-level sharing and fusion strategy bridges per- ception and planning. Specifically, CooperDrive reuses detector Bird’s-Eye View (BEV) features to estimate accurate vehicle poses without additional heavy encoders, thereby reconstructing BEV representations and feeding the planner with low latency. On the planning side, CooperDrive leverages the expanded object set to anticipate potential conflicts earlier and adjust speed and trajectory proactively, thereby transforming reactive behaviors into predictive and safer driving decisions. Real- world closed-loop tests at occlusion-heavy NLOS intersections demonstrate that CooperDrive increases reaction lead time, minimum time-to-collision (TTC), and stopping margin, while requiring only 90 kbps bandwidth and maintaining an average end-to-end latency of 89 ms.