Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

VLM-RL: A Unified Vision Language Models and Reinforcement Learning Framework for Safe Autonomous Driving

About

In recent years, reinforcement learning (RL)-based methods for learning driving policies have gained increasing attention in the autonomous driving community and have achieved remarkable progress in various driving scenarios. However, traditional RL approaches rely on manually engineered rewards, which require extensive human effort and often lack generalizability. To address these limitations, we propose \textbf{VLM-RL}, a unified framework that integrates pre-trained Vision-Language Models (VLMs) with RL to generate reward signals using image observation and natural language goals. The core of VLM-RL is the contrasting language goal (CLG)-as-reward paradigm, which uses positive and negative language goals to generate semantic rewards. We further introduce a hierarchical reward synthesis approach that combines CLG-based semantic rewards with vehicle state information, improving reward stability and offering a more comprehensive reward signal. Additionally, a batch-processing technique is employed to optimize computational efficiency during training. Extensive experiments in the CARLA simulator demonstrate that VLM-RL outperforms state-of-the-art baselines, achieving a 10.5\% reduction in collision rate, a 104.6\% increase in route completion rate, and robust generalization to unseen driving scenarios. Furthermore, VLM-RL can seamlessly integrate almost any standard RL algorithms, potentially revolutionizing the existing RL paradigm that relies on manual reward engineering and enabling continuous performance improvements. The demo video and code can be accessed at: https://zilin-huang.github.io/VLM-RL-website.

Zilin Huang, Zihao Sheng, Yansong Qu, Junwei You, Sikai Chen• 2024

Related benchmarks

TaskDatasetResultRank
Autonomous DrivingCARLA Town 2 (test)
AS19.3
15
Autonomous DrivingCARLA Town 2 (train)
AS22.53
12
Autonomous DrivingCARLA Town 2 10 unseen (test)
AS Score14.38
12
Autonomous DrivingCARLA Town 3
AS (Average Score)16.64
5
Autonomous DrivingCARLA Town 4
Average Speed (AS)19.39
5
Autonomous DrivingCARLA Town 01 (Challenging)
Route Completion80
3
Autonomous DrivingCARLA Town 01 Normal
Route Completion88
3
Autonomous DrivingCARLA Town 02 Challenging
Route Completion68
3
Autonomous DrivingCARLA Town 03 Challenging
Route Completion62
3
Autonomous DrivingCARLA Towns 01-03 Total Challenging
Route Completion2.1
3
Showing 10 of 19 rows

Other info

Follow for update