Efficient Local-to-Global Collaborative Perception via Joint Communication and Computation Optimization

TL;DR

The paper addresses the high data and computation demands of collaborative perception in connected autonomous vehicles by proposing a local-to-global framework (LGCP). LGCP partitions the road of interest into non-overlapping areas, assigns area-specific CAV groups with leaders to perform local fusion, and uses an RSU for centralized task assignment, data aggregation, and global view broadcasting, thereby reducing redundant transmissions and distributing computation. The approach is formalized as an optimization problem balancing area confidence against latency under a real-time constraint, with a greedy group-selection algorithm and a priority-based transmission scheduler. Experimental results across datasets and co-simulation platforms show LGCP can achieve around 44x data transmission reduction and substantial latency improvements while maintaining or enhancing perception accuracy, demonstrating practical benefits for scalable, real-time cooperative perception in dense CAV scenarios.

Abstract

Autonomous driving relies on accurate perception to ensure safe driving. Collaborative perception improves accuracy by mitigating the sensing limitations of individual vehicles, such as limited perception range and occlusion-induced blind spots. However, collaborative perception often suffers from high communication overhead due to redundant data transmission, as well as increasing computation latency caused by excessive load with growing connected and autonomous vehicles (CAVs) participation. To address these challenges, we propose a novel local-to-global collaborative perception framework (LGCP) to achieve collaboration in a communication- and computation-efficient manner. The road of interest is partitioned into non-overlapping areas, each of which is assigned a dedicated CAV group to perform localized perception. A designated leader in each group collects and fuses perception data from its members, and uploads the perception result to the roadside unit (RSU), establishing a link between local perception and global awareness. The RSU aggregates perception results from all groups and broadcasts a global view to all CAVs. LGCP employs a centralized scheduling strategy via the RSU, which assigns CAV groups to each area, schedules their transmissions, aggregates area-level local perception results, and propagates the global view to all CAVs. Experimental results demonstrate that the proposed LGCP framework achieves an average 44 times reduction in the amount of data transmission, while maintaining or even improving the overall collaborative performance.

Figures (7)

• Figure 1: Different collaboration perspectives.
• Figure 2: Object detection architecture of LGCP.
• Figure 3: AP@0.7 performance among three collaboration models integrated with LGCP framework and the amount of data transmission for varying $\Delta_g$.
• Figure 4: Amount of data transmission for varying number of CAVs under the OPV2V dataset.
• Figure 5: End-to-end latency for varying number of CAVs under the OPV2V dataset.