Constrained Reinforcement Learning for Vehicle Motion Planning with Topological Reachability Analysis

Rule-based traditional motion planning methods usually perform well with prior knowledge of the macro-scale environments but encounter challenges in unknown and uncertain environments. Deep reinforcement learning (DRL) is a solution that can effectively deal micro-scale Nevertheless, DRL unstable lacks interpretability. Therefore, it raises new challenge: how to combine effectiveness overcome drawbacks two while guaranteeing stability In this study, multi-constraint multi-scale method proposed for automated driving use constrained (RL), named RLTT, comprising RL, topological reachability analysis used vehicle path space (TPS), trajectory lane model (TLM). First, dynamic vehicles formulated; then, TLM developed on basis model, thus constraining RL action state space. Second, achieved through TPS, range, discrete routing points are via RLTT. Third, designed by combining sophisticated rules, theoretical provided guarantee efficiency our method. Finally, related experiments conducted evaluate method; reduce 19.9% distance cost as compared Experimental results indicate help mitigate gap between data-driven methods, provide better performance driving, facilitate more fields.