Encoding Distributional Soft Actor-Critic for Autonomous Driving in Multi-lane Scenarios

TL;DR

This paper introduces E-DSAC, a novel reinforcement learning algorithm for autonomous driving that handles variable surrounding vehicles without manual rules, achieving higher performance and safety in multi-lane scenarios.

Contribution

The paper develops an encoding distributional policy iteration framework with permutation invariance, and proposes E-DSAC, a new RL algorithm that improves decision-making in autonomous driving.

Findings

E-DSAC policy is about three times more effective than DSAC.

E-DSAC achieves efficient, smooth, and safe driving in multi-lane scenarios.

The algorithm's effectiveness is validated in real vehicle experiments.

Abstract

In this paper, we propose a new reinforcement learning (RL) algorithm, called encoding distributional soft actor-critic (E-DSAC), for decision-making in autonomous driving. Unlike existing RL-based decision-making methods, E-DSAC is suitable for situations where the number of surrounding vehicles is variable and eliminates the requirement for manually pre-designed sorting rules, resulting in higher policy performance and generality. We first develop an encoding distributional policy iteration (DPI) framework by embedding a permutation invariant module, which employs a feature neural network (NN) to encode the indicators of each vehicle, in the distributional RL framework. The proposed DPI framework is proved to exhibit important properties in terms of convergence and global optimality. Next, based on the developed encoding DPI framework, we propose the E-DSAC algorithm by adding the gradient-based update rule of the feature NN to the policy evaluation process of the DSAC algorithm. Then, the multi-lane driving task and the corresponding reward function are designed to verify the effectiveness of the proposed algorithm. Results show that the policy learned by E-DSAC can realize efficient, smooth, and relatively safe autonomous driving in the designed scenario. And the final policy performance learned by E-DSAC is about three times that of DSAC. Furthermore, its effectiveness has also been verified in real vehicle experiments.