PerlAD: Towards Enhanced Closed-loop End-to-end Autonomous Driving with Pseudo-simulation-based Reinforcement Learning

TL;DR

PerlAD introduces a novel pseudo-simulation-based reinforcement learning approach for end-to-end autonomous driving, enabling efficient, rendering-free training and improved performance in closed-loop scenarios using offline data.

Contribution

The paper proposes PerlAD, a pseudo-simulation framework with a world model and hierarchical planner, to enhance closed-loop autonomous driving training without online interactions.

Findings

Achieves 10.29% higher Driving Score on Bench2Drive benchmark.

Outperforms previous E2E RL methods in efficiency and safety.

Demonstrates robustness in occlusion scenarios on DOS benchmark.

Abstract

End-to-end autonomous driving policies based on Imitation Learning (IL) often struggle in closed-loop execution due to the misalignment between inadequate open-loop training objectives and real driving requirements. While Reinforcement Learning (RL) offers a solution by directly optimizing driving goals via reward signals, the rendering-based training environments introduce the rendering gap and are inefficient due to high computational costs. To overcome these challenges, we present a novel Pseudo-simulation-based RL method for closed-loop end-to-end autonomous driving, PerlAD. Based on offline datasets, PerlAD constructs a pseudo-simulation that operates in vector space, enabling efficient, rendering-free trial-and-error training. To bridge the gap between static datasets and dynamic closed-loop environments, PerlAD introduces a prediction world model that generates reactive agent trajectories conditioned on the ego vehicle's plan. Furthermore, to facilitate efficient planning, PerlAD utilizes a hierarchical decoupled planner that combines IL for lateral path generation and RL for longitudinal speed optimization. Comprehensive experimental results demonstrate that PerlAD achieves state-of-the-art performance on the Bench2Drive benchmark, surpassing the previous E2E RL method by 10.29% in Driving Score without requiring expensive online interactions. Additional evaluations on the DOS benchmark further confirm its reliability in handling safety-critical occlusion scenarios.