Cooperative Autonomous Driving Research

The field of cooperative autonomous driving is moving towards more efficient and scalable solutions, with a focus on end-to-end learning and bandwidth-efficient communication. Researchers are exploring new frameworks and architectures that enable cooperative perception and decision-making among multiple agents, including vehicles and infrastructure. These advancements have the potential to enhance driving safety and overcome the limitations of single-vehicle autonomous driving systems. Noteworthy papers include: CoopTrack, which proposes a fully instance-level end-to-end framework for cooperative tracking, and EffiComm, which introduces an end-to-end framework for bandwidth-efficient multi-agent communication. CDA-SimBoost is also a significant contribution, providing a unified framework for bridging real data and simulation for infrastructure-based CDA systems. The End-to-End V2X Cooperative Autonomous Driving Competition has also facilitated research in this area, establishing a unified benchmark for evaluating cooperative driving systems and highlighting key research problems.

Sources

CoopTrack: Exploring End-to-End Learning for Efficient Cooperative Sequential Perception

EffiComm: Bandwidth Efficient Multi Agent Communication

CDA-SimBoost: A Unified Framework Bridging Real Data and Simulation for Infrastructure-Based CDA Systems

Research Challenges and Progress in the End-to-End V2X Cooperative Autonomous Driving Competition

Built with on top of