The field of autonomous vehicle research is witnessing significant developments in decision-making capabilities, with a focus on improving safety, efficiency, and comfort. Recent studies have explored the integration of Bayesian inference and reinforcement learning to enhance agent decision-making, allowing for more effective and efficient navigation of complex scenarios. Notably, the development of novel risk-aware objectives and the formulation of hierarchical driving objectives have shown promise in promoting safer driving behaviors. Furthermore, the application of deep reinforcement learning to longitudinal control strategies at signalized intersections has demonstrated potential in improving traffic safety and efficiency. Noteworthy papers include:
- Balancing Progress and Safety: A Novel Risk-Aware Objective for RL in Autonomous Driving, which introduces a novel risk-aware objective for reinforcement learning in autonomous driving.
- Deep reinforcement learning-based longitudinal control strategy for automated vehicles at signalised intersections, which proposes a deep reinforcement learning-based control strategy for automated vehicles at signalized intersections.