AutoMoT: A Unified Vision-Language-Action Model with Asynchronous Mixture-of-Transformers for End-to-End Autonomous Driving

TL;DR

AutoMoT introduces a unified vision-language-action model with a mixture-of-transformers architecture for efficient, end-to-end autonomous driving, leveraging pre-trained VLMs and asynchronous inference to improve scene understanding and decision-making.

Contribution

It proposes a novel end-to-end AD framework that unifies reasoning and action generation using a mixture-of-transformers with asynchronous execution, enhancing efficiency and reasoning capabilities.

Findings

AutoMoT achieves competitive performance on multiple benchmarks.

Pre-trained VLMs can handle scene understanding with semantic prompting without fine-tuning.

Fine-tuning remains necessary for action-level tasks like decision-making.

Abstract

Integrating vision-language models (VLMs) into end-to-end (E2E) autonomous driving (AD) systems has shown promise in improving scene understanding. However, existing integration strategies suffer from several limitations: they either struggle to resolve distribution misalignment between reasoning and action spaces, underexploit the general reasoning capabilities of pretrained VLMs, or incur substantial inference latency during action policy generation, which degrades driving performance. To address these challenges, we propose AutoMoT in this work, an end-to-end AD framework that unifies reasoning and action generation within a single vision-language-action (VLA) model. Our approach leverages a mixture-of-transformer (MoT) architecture with joint attention sharing, which preserves the general reasoning capabilities of pre-trained VLMs while enabling efficient fast-slow inference through asynchronous execution at different task frequencies. Extensive experiments on multiple benchmarks, under both open- and closed-loop settings, demonstrate that AutoMoT achieves competitive performance compared to state-of-the-art methods. We further investigate the functional boundary of pre-trained VLMs in AD, examining when AD-tailored fine-tuning is necessary. Our results show that pre-trained VLMs can achieve competitive multi-task scene understanding performance through semantic prompting alone, while fine-tuning remains essential for action-level tasks such as decision-making and trajectory planning. We refer to https://automot-website.github.io/ for the demonstration videos and qualitative results.