Learning to Drift with Individual Wheel Drive: Maneuvering Autonomous Vehicle at the Handling Limits

2025-07-31

AI Summary

To address the significant performance degradation of reinforcement learning (RL) policies when transferring from simulation to real-world autonomous vehicles, this paper proposes a GPU-accelerated parallel simulation training framework incorporating systematic domain randomization, specifically designed for drift control under extreme maneuvering conditions. Methodologically, it integrates deep RL with high-fidelity multibody dynamics modeling and cross-physical-parameter domain randomization to enhance policy robustness in simulation. The primary contributions are: (i) the first successful end-to-end sim-to-real transfer of stable drift control on a custom-built, open-source 1/10-scale RC platform featuring independent wheel drive; and (ii) experimental validation demonstrating precise tracking of fixed-radius and variable-curvature trajectories, seamless steering-mode switching, and stable regulation of sideslip angle—thereby substantially narrowing the sim-to-real performance gap.

Technology Category

• Machine Learning: Imitation Learning & Inverse Reinforcement Learning
• Search and Optimization: Sampling/Simulation-based Search
• Planning, Routing, and Scheduling: Replanning and Plan Repair

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Drifting, characterized by controlled vehicle motion at high sideslip angles, is crucial for safely handling emergency scenarios at the friction limits. While recent reinforcement learning approaches show promise for drifting control, they struggle with the significant simulation-to-reality gap, as policies that perform well in simulation often fail when transferred to physical systems. In this paper, we present a reinforcement learning framework with GPU-accelerated parallel simulation and systematic domain randomization that effectively bridges the gap. The proposed approach is validated on both simulation and a custom-designed and open-sourced 1/10 scale Individual Wheel Drive (IWD) RC car platform featuring independent wheel speed control. Experiments across various scenarios from steady-state circular drifting to direction transitions and variable-curvature path following demonstrate that our approach achieves precise trajectory tracking while maintaining controlled sideslip angles throughout complex maneuvers in both simulated and real-world environments.

Problem

• Bridging simulation-to-reality gap in autonomous vehicle drifting control
• Achieving precise trajectory tracking at high sideslip angles
• Validating control on Individual Wheel Drive RC car platform

Innovation

• GPU-accelerated parallel simulation for reinforcement learning
• Systematic domain randomization to bridge reality gap
• Individual Wheel Drive platform for precise drifting control