Generalizing Motion Planners with Mixture of Experts for Autonomous Driving

TL;DR

This paper introduces StateTransformer-2, a scalable motion planner using a mixture-of-experts Transformer architecture, which improves generalization in autonomous driving by addressing model complexity and training issues.

Contribution

The paper presents a novel scalable motion planner with a mixture-of-experts architecture that enhances generalization and addresses training challenges in autonomous driving.

Findings

STR2 outperforms previous methods on NuPlan dataset

Model scalability improves generalization and accuracy

MoE architecture addresses modality collapse and reward balancing

Abstract

Large real-world driving datasets have sparked significant research into various aspects of data-driven motion planners for autonomous driving. These include data augmentation, model architecture, reward design, training strategies, and planner pipelines. These planners promise better generalizations on complicated and few-shot cases than previous methods. However, experiment results show that many of these approaches produce limited generalization abilities in planning performance due to overly complex designs or training paradigms. In this paper, we review and benchmark previous methods focusing on generalizations. The experimental results indicate that as models are appropriately scaled, many design elements become redundant. We introduce StateTransformer-2 (STR2), a scalable, decoder-only motion planner that uses a Vision Transformer (ViT) encoder and a mixture-of-experts (MoE) causal Transformer architecture. The MoE backbone addresses modality collapse and reward balancing by expert routing during training. Extensive experiments on the NuPlan dataset show that our method generalizes better than previous approaches across different test sets and closed-loop simulations. Furthermore, we assess its scalability on billions of real-world urban driving scenarios, demonstrating consistent accuracy improvements as both data and model size grow.