Safe Reinforcement Learning with Mixture Density Network: A Case Study in Autonomous Highway Driving

TL;DR

This paper introduces a safe reinforcement learning framework for autonomous highway driving that combines heuristic and data-driven safety modules, utilizing mixture density RNNs for multimodal trajectory prediction to improve safety and performance.

Contribution

It proposes a novel safety system integrating heuristic and learning-based safety modules with MD-RNNs for multimodal predictions, enhancing autonomous driving safety.

Findings

Outperforms previous methods in average reward

Reduces the number of collisions

Accelerates learning progress in simulation

Abstract

This paper presents a safe reinforcement learning system for automated driving that benefits from multimodal future trajectory predictions. We propose a safety system that consists of two safety components: a heuristic safety and a learning-based safety. The heuristic safety module is based on common driving rules. On the other hand, the learning-based safety module is a data-driven safety rule that learns safety patterns from driving data. Specifically, it utilizes mixture density recurrent neural networks (MD-RNN) for multimodal future trajectory predictions to accelerate the learning progress. Our simulation results demonstrate that the proposed safety system outperforms previously reported results in terms of average reward and number of collisions.