ROSA-RL: Uncertainty-Aware Roundabout Optimized Speed Advisory with Reinforcement Learning
Researchers have developed ROSA-RL, a novel system that uses reinforcement learning and a Transformer-based model to improve automated driving at roundabouts. The system addresses the uncertainty inherent in mixed traffic by probabilistically forecasting conflict zone occupancy over a five-second horizon. By encoding uncertainty in future motion and intent, ROSA-RL enables safer and more efficient vehicle entry into roundabouts, outperforming existing model-based approaches in simulations. AI
IMPACT This research could enhance the safety and efficiency of autonomous vehicles in complex urban environments like roundabouts.