CRLLK: Constrained Reinforcement Learning for Lane Keeping in Autonomous Driving
Published in AAMAS 2025, Detroit, Michigan, USA, 2025
We modeled the problem of lane-keeping as a constrained optimization problem and presented a constrained RL based solution to the problem. The weight coefficients are adaptively learned without scenario-specific tuning and grid search. Empirically, we observe that our approach surpasses traditional RL-based approaches.
Recommended citation: Xinwei Gao, Arambam James Singh, Gangadhar Royyuru, Michael Yuhas, and Arvind Easwaran. 2025. CRLLK: Constrained Reinforcement Learning for Lane Keeping in Autonomous Driving. In Proc. of the 24th International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2025), Detroit, Michigan, USA, May 19 – 23, 2025, IFAAMAS, 5 pages.