Deep Reinforcement Learning enhances CAV platoon joining safety

A new research paper explores the use of deep reinforcement learning (DRL) for controlling connected and automated vehicle (CAV) platoon joining maneuvers in mixed traffic environments. The study proposes a simulation framework using SUMO to compare algorithms like DQN, DDQN, and PPO. Results indicate that PPO achieved a high success rate of approximately 98% with a collision rate below 1%, largely due to its reward function incorporating penalties for risky behaviors. However, this enhanced safety came at the cost of increased decision steps, highlighting a trade-off between safety, efficiency, and decision-making speed. AI

IMPACT This research could lead to safer and more efficient autonomous vehicle navigation in complex traffic scenarios.

RANK_REASON Research paper published on arXiv detailing a new method for controlling CAVs. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• Connected and automated vehicle (CAV)
• Deep Q-Network (DQN)
• Deep Reinforcement Learning (DRL)
• Double Deep Q-Network (DDQN)
• Proximal Policy Optimization (PPO)
• reinforcement learning
• Simulation of Urban MObility (SUMO)

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →