Efficient and Explainable End-to-End Autonomous Driving via Masked Vision-Language-Action Diffusion

TL;DR

This paper introduces MVLAD-AD, a novel diffusion-based framework for autonomous driving that enhances planning efficiency, precision, and explainability through a discrete action tokenization and geometry-aware embeddings.

Contribution

The paper proposes a new diffusion model with discrete action tokenization and geometry-aware embeddings for autonomous driving, improving efficiency and explainability over prior methods.

Findings

Outperforms state-of-the-art autoregressive and diffusion models in planning accuracy.

Achieves higher efficiency in inference compared to existing approaches.

Provides high-fidelity, explainable reasoning in autonomous driving scenarios.

Abstract

Large Language Models (LLMs) and Vision-Language Models (VLMs) have emerged as promising candidates for end-to-end autonomous driving. However, these models typically face challenges in inference latency, action precision, and explainability. Existing autoregressive approaches struggle with slow token-by-token generation, while prior diffusion-based planners often rely on verbose, general-purpose language tokens that lack explicit geometric structure. In this work, we propose Masked Vision-Language-Action Diffusion for Autonomous Driving (MVLAD-AD), a novel framework designed to bridge the gap between efficient planning and semantic explainability via a masked vision-language-action diffusion model. Unlike methods that force actions into the language space, we introduce a discrete action tokenization strategy that constructs a compact codebook of kinematically feasible waypoints from real-world driving distributions. Moreover, we propose geometry-aware embedding learning to ensure that embeddings in the latent space approximate physical geometric metrics. Finally, an action-priority decoding strategy is introduced to prioritize trajectory generation. Extensive experiments on nuScenes and derived benchmarks demonstrate that MVLAD-AD achieves superior efficiency and outperforms state-of-the-art autoregressive and diffusion baselines in planning precision, while providing high-fidelity and explainable reasoning.