aiqbench analysis
This research applies deep reinforcement learning to solve the complex dual-objective control problem required for autonomous rally drifting, where agents must balance stabilizing nonlinear dynamics with minimizing lap times. By utilizing a track-guided progressive training framework that incorporates optimal planning trajectories as prior data, the study demonstrates how AI can master high-risk maneuvers previously exclusive to human intuition. The work advances the frontier of robotic vehicle control by proving that hierarchical RL policies can effectively manage extreme sideslip angles for performance optimization.
Track-Guided Hierarchical Reinforcement Learning for Autonomous Vehicle Drifting with Minimum-Lap-Time Planning
arXiv:2608.00113v1 Announce Type: new Abstract: In Formula 1, drivers optimize racing lines within tire grip limits to minimize lap times; however, in rally racing, drivers intentionally break traction to drift on loose surfaces. This maneuver rapidly aligns the vehicle for corner exits, ultimately reducing lap time. Autonomously executing such maneuvers formulates a complex dual-objective control problem: stabilizing highly nonlinear drift dynamics while strictly minimizing lap time. Addressing this challenge motivates the development of advanced Minimum-Lap-Time (MLT) drift control architectures. This paper proposes a planning-control framework specifically designed for MLT drifting scenario. First, we formulate an optimal control problem to generate a MLT drift planning trajectory, which is used as prior data to train a deep reinforcement learning drift controller. Given that drifting involves extremely large sideslip angles and is therefore challenging to learn directly, a Track-guided Reinforcement Learning (TgRL) drift control method is proposed to enable progressive training in a step-by-step manner, from drift control policy, to drift corner policy, and finally to a comprehensive drift race policy. The reward function incorporates both an instant reward term and an end reward term derived from the Minimum-Lap-Time objective. Simulation results demonstrate that the proposed framework enables the agent to learn a drift racing policy that not only ensures vehicle motion control performance but also effectively reduces lap time.
Abstract:In Formula 1, drivers optimize racing lines within tire grip limits to minimize lap times; however, in rally racing, drivers intentionally break traction to drift on loose surfaces. This maneuver rapidly aligns the vehicle for corner exits, ultimately reducing lap time. Autonomously executing such maneuvers formulates a complex dual-objective control problem: stabilizing highly nonlinear drift dynamics while strictly minimizing lap time. Addressing this challenge motivates the development of advanced Minimum-Lap-Time (MLT) drift control architectures. This paper proposes a planning-control framework specifically designed for MLT drifting scenario. First, we formulate an optimal control problem to generate a MLT drift planning trajectory, which is used as prior data to train a deep reinforcement learning drift controller. Given that drifting involves extremely large sideslip angles and is therefore challenging to learn directly, a Track-guided Reinforcement Learning (TgRL) drift control method is proposed to enable progressive training in a step-by-step manner, from drift control policy, to drift corner policy, and finally to a comprehensive drift race policy. The reward function incorporates both an instant reward term and an end reward term derived from the Minimum-Lap-Time objective. Simulation results demonstrate that the proposed framework enables the agent to learn a drift racing policy that not only ensures vehicle motion control performance but also effectively reduces lap time.
Robotics (cs.RO)
arXiv:2608.00113 [cs.RO]
(or arXiv:2608.00113v1 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2608.00113 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 303 words
Related coverage
News
Quantum Computing Report · Jul 25, 2026
Deep Reinforcement-Learning-Guided Model Predictive Control for Preventing Overtakes in Autonomous Racing
arXiv · Jul 24, 2026
Bicycle Acrobatics with Reinforcement Learning
arXiv · Aug 4, 2026
DiDrive: A Risk-Aware Hierarchical Diffusion Framework for Safe Offline Reinforcement Learning in Autonomous Driving
arXiv · Sep 3, 2026