MindDrive: A Vision-Language-Action Model for Autonomous Driving Utilizing Language as Action in Online Reinforcement Learning

Haoyu Fu^1*, Diankun Zhang^2*, Zongchuang Zhao¹,
Jianfeng Cui², Hongwei Xie^2†, Bing Wang², Guang Chen², Dingkang Liang^1†, Xiang Bai¹

¹ Huazhong University of Science & Technology, ² Xiaomi EV

(*) Equal contribution. (†) Project leader.

Abstract

Current Vision-Language-Action (VLA) paradigms in autonomous driving primarily rely on Imitation Learning (IL), which introduces inherent challenges such as distribution shift and causal confusion. Online Reinforcement Learning offers a promising pathway to address these issues through trial-and-error learning. However, applying online reinforcement learning to VLA models in autonomous driving is hindered by inefficient exploration in continuous action spaces. To overcome this limitation, we propose MindDrive, a VLA framework comprising a large language model (LLM) with two distinct sets of LoRA parameters. The one LLM serves as a Decision Expert for scenario reasoning and driving decision-making, while the other acts as an Action Expert that dynamically maps linguistic decisions into feasible trajectories. By feeding trajectory-level rewards back into the reasoning space, MindDrive enables trial-and-error learning over a finite set of discrete linguistic driving decisions, instead of operating directly in a continuous action space. This approach effectively balances optimal decision-making in complex scenarios, human-like driving behavior, and efficient exploration in online reinforcement learning. MindDrive achieves strong closed-loop performance on the challenging Bench2Drive benchmark, with a Driving Score (DS) of 78.04 and a Success Rate (SR) of 55.09%. To the best of our knowledge, this is the first work to demonstrate the effectiveness of online reinforcement learning for the VLA model in autonomous driving.

Overview

News

[2025/12/16] ArXiv paper release.

Currently Supported Features

• MindDrive Inference Framework
• Close-loop Evaluation
• MindDrive Checkpoint
• MindDrive Training Framework

Results and Checkpoints

Orion and other baselines

Method	L2 (m) 2s	Driving Score	Success Rate(%)	Config	Download	Eval Json
UniAD-Tiny	0.80	40.73	13.18	config	Hugging Face/Baidu Cloud	Json
UniAD-Base	0.73	45.81	16.36	config	Hugging Face/Baidu Cloud	Json
VAD	0.91	42.35	15.00	config	Hugging Face/Baidu Cloud	Json
ORION-7B	0.68	77.74	54.62	config	Hugging Face	Json
MindDrive-0.5B	0.73	78.04	55.09	config	-	-

Citation

If this work is helpful for your research, please consider citing:

@article{fu2025minddrive,
  title={MindDrive: A Vision-Language-Action Model for Autonomous Driving via Online Reinforcement Learning},
  author={Haoyu Fu and Diankun Zhang and Zongchuang Zhao and Jianfeng Cui and Hongwei Xie and Bing Wang and Guang Chen and Dingkang Liang and Xiang Bai},
  journal={arXiv Preprint arXiv:2512.13636},  
  year={2025},
}

@inproceedings{fu2025orion,
  title={ORION: A Holistic End-to-End Autonomous Driving Framework by Vision-Language Instructed Action Generation},
  author={Haoyu Fu and Diankun Zhang and Zongchuang Zhao and Jianfeng Cui and Dingkang Liang and Chong Zhang and Dingyuan Zhang and Hongwei Xie and Bing Wang and Xiang Bai},
  booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
  year={2025}
}