FastDriveVLA: Efficient End-to-End Driving via Plug-and-Play Reconstruction-based Token Pruning

TL;DR

FastDriveVLA introduces a reconstruction-based token pruning method for autonomous driving that efficiently retains foreground information, significantly reducing computational costs while maintaining high performance in scene understanding and decision-making.

Contribution

The paper presents a novel plug-and-play reconstruction-based token pruner, ReconPruner, trained with a new adversarial foreground-background strategy and a large-scale dataset, nuScenes-FG, for efficient autonomous driving models.

Findings

Achieves state-of-the-art results on nuScenes planning benchmark.

Effectively retains foreground information with high pruning ratios.

Seamless application to different VLA models without retraining.

Abstract

Vision-Language-Action (VLA) models have demonstrated significant potential in complex scene understanding and action reasoning, leading to their increasing adoption in end-to-end autonomous driving systems. However, the long visual tokens of VLA models greatly increase computational costs. Current visual token pruning methods in Vision-Language Models (VLM) rely on either visual token similarity or visual-text attention, but both have shown poor performance in autonomous driving scenarios. Given that human drivers concentrate on relevant foreground areas while driving, we assert that retaining visual tokens containing this foreground information is essential for effective decision-making. Inspired by this, we propose FastDriveVLA, a novel reconstruction-based vision token pruning framework designed specifically for autonomous driving. FastDriveVLA includes a plug-and-play visual token pruner called ReconPruner, which prioritizes foreground information through MAE-style pixel reconstruction. A novel adversarial foreground-background reconstruction strategy is designed to train ReconPruner for the visual encoder of VLA models. Once trained, ReconPruner can be seamlessly applied to different VLA models with the same visual encoder without retraining. To train ReconPruner, we also introduce a large-scale dataset called nuScenes-FG, consisting of 241K image-mask pairs with annotated foreground regions. Our approach achieves state-of-the-art results on the nuScenes open-loop planning benchmark across different pruning ratios.