Autonomous Driving with Reinforcement Learning and Rule-based Policies

Abstract
The design of high-level decision-making systems is a topical problem in the field of autonomous driving. In this paper, we combine rule-based strategies and reinforcement learning (RL) with the goal of achieving transparency and robustness of the learned controllers while maintaining the generality of a policy learned by environment interaction. We combine the best properties of these two worlds by designing parametric rule-based controllers, in which interpretable rules can be provided by domain experts and their parameters are learned via RL. After illustrating how to apply parameter-based RL methods (PGPE), we present numerical simulations in the highway and in two urban scenarios: intersection and roundabout.

[Link] [Slides] [Talk] [BibTeX]

 @article{likmeta2020autonomous,
    author = "Likmeta, Amarildo and Metelli, Alberto Maria and Tirinzoni, Andrea and Giol, Riccardo and Restelli, Marcello and Romano, Danilo and Alessandretti, Andrea",
    title = "Autonomous Driving with Reinforcement Learning and Rule-based Policies",
    journal = "Workshop on AI for Autonomous Driving (AIAD) @ICML 2020",
    year = "2020",
    url = "https://drive.google.com/file/d/1ASJa-pOgZ\_Z78KTVjrTV1kqqtQ5-RP\_w/view"
}