Vehicle Dynamics Embedded World Models for Autonomous Driving

2025-12-02

IEEE transactions on intelligent transportation systems (Print)

AI Summary

Current driving world models jointly model ego-vehicle dynamics and environmental dynamics, resulting in poor cross-vehicle generalization and low robustness to dynamical variations. To address this, we propose VDD, a decoupled driving world model that explicitly embeds and isolates ego-dynamics modeling within the world model—achieving principled decoupling from environmental dynamics for the first time. VDD introduces a dynamics-aware state decomposition mechanism and a two-stage optimization framework: Policy Adaptation at Deployment (PAD) and Policy Augmentation during Training (PAT). The method integrates latent-variable modeling, sequential decision learning, and dynamics-constrained policy gradient optimization. Experiments demonstrate that VDD significantly improves driving performance and robustness to vehicle parameter variations in simulation, outperforming existing state-of-the-art methods in cross-vehicle generalization.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Planning, Routing, and Scheduling: Planning with Markov Models (MDPs, POMDPs)
• Search and Optimization: Learning to Search

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

World models have gained significant attention as a promising approach for autonomous driving. By emulating human-like perception and decision-making processes, these models can predict and adapt to dynamic environments. Existing methods typically map high-dimensional observations into compact latent spaces and learn optimal policies within these latent representations. However, prior work usually jointly learns ego-vehicle dynamics and environmental transition dynamics from the image input, leading to inefficiencies and a lack of robustness to variations in vehicle dynamics. To address these issues, we propose the Vehicle Dynamics embedded Dreamer (VDD) method, which decouples the modeling of ego-vehicle dynamics from environmental transition dynamics. This separation allows the world model to generalize effectively across vehicles with diverse parameters. Additionally, we introduce two strategies to further enhance the robustness of the learned policy: Policy Adjustment during Deployment (PAD) and Policy Augmentation during Training (PAT). Comprehensive experiments in simulated environments demonstrate that the proposed model significantly improves both driving performance and robustness to variations in vehicle dynamics, outperforming existing approaches.

Problem

• Decouples ego-vehicle dynamics from environmental transitions
• Enhances generalization across vehicles with diverse parameters
• Improves driving performance and robustness to dynamics variations

Innovation

• Decouples ego-vehicle dynamics from environmental dynamics
• Introduces Policy Adjustment during Deployment strategy
• Uses Policy Augmentation during Training for robustness