TrajMoE: Scene-Adaptive Trajectory Planning with Mixture of Experts and Reinforcement Learning

TL;DR

TrajMoE introduces a scene-adaptive trajectory planning framework combining Mixture of Experts and Reinforcement Learning, improving autonomous driving performance by scenario-specific priors and policy-driven refinement.

Contribution

The paper proposes a novel approach that uses MoE for scenario-specific trajectory priors and RL for policy refinement, enhancing end-to-end autonomous driving models.

Findings

Achieved 51.08 score on navsim ICCV benchmark

Secured third place in the competition

Enhanced perceptual features with multiple backbones

Abstract

Current autonomous driving systems often favor end-to-end frameworks, which take sensor inputs like images and learn to map them into trajectory space via neural networks. Previous work has demonstrated that models can achieve better planning performance when provided with a prior distribution of possible trajectories. However, these approaches often overlook two critical aspects: 1) The appropriate trajectory prior can vary significantly across different driving scenarios. 2) Their trajectory evaluation mechanism lacks policy-driven refinement, remaining constrained by the limitations of one-stage supervised training. To address these issues, we explore improvements in two key areas. For problem 1, we employ MoE to apply different trajectory priors tailored to different scenarios. For problem 2, we utilize Reinforcement Learning to fine-tune the trajectory scoring mechanism. Additionally, we integrate models with different perception backbones to enhance perceptual features. Our integrated model achieved a score of 51.08 on the navsim ICCV benchmark, securing third place.