arXiv Robotics — research abstracts· Mahmoud Selim, Cristina Cipriani, Karl Henrik Johansson·· 2 days agoEditorial score55
Beyond Policy Support: Interaction Constrained Offline Reinforcement Learning for Autonomous Driving
Beyond Policy Support: Interaction Constrained Offline Reinforcement Learning for Autonomous Driving
Summary
This research introduces Interaction-Constrained Drive Policy (ICDP), an offline reinforcement learning framework designed to address interaction distribution shift in autonomous driving. By decomposing joint-support degradation into ego-support and interaction-support components, ICDP suppresses high-value yet interaction-unsupported trajectory selections, improving performance in interaction-critical driving scenarios through evaluations on nuPlan, Interplan, and real-world truck experiments.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
Source：arXiv Robotics — research abstracts · arxiv.org