Autonomous vehicles learn to cooperate in complex traffic

Researchers from the National University of Singapore, The Hong Kong Polytechnic University, and Tsinghua University developed COIN, a multi-agent reinforcement learning framework that enables autonomous vehicles to collaboratively navigate complex traffic scenarios such as unsignalized intersections, roundabouts, and bottlenecks. COIN uses a centralized training and decentralized execution method to improve both individual vehicle navigation and overall traffic system cooperation, validated in simulations and hybrid real-simulation robotic platforms.

This Wire brief sits within Fusion42's coverage of Autonomous Vehicles and AI Agents.