Deep Reinforcement Learning-Based Lane-Changing Trajectory Planning for Connected and Automated Vehicles

A deep reinforcement learning-based method for planning the lane-changing trajectory of connected and automated vehicles (CAVs) is proposed to increase the traffic efficiency and reduce fuel consumption. The long-short-termmemory-based twin delayed deep deterministic policy gradient (LSTM-TD3) algorithm is implemented and trained to achieve the optimal longitudinal and lateral lane-changing trajectory. The instantaneous fuel consumption, the difference with the desired speed and acceleration are used as reward and penalty terms. The effectiveness of the algorithm was verified through two typical lane-changing simulation scenarios. The results indicate that the proposed LSTM-TD3-based lane-changing planning method reduced the average fuel consumption by
$13.71{{\% }}$
compared to TD3,
$16.54{{\% }}$
compared to LSTM-DDPG, and
$27.09{{\% }}$
compared to DDPG.