Deep Reinforcement-Learning-Guided Model Predictive Control for Preventing Overtakes in Autonomous Racing

2026-07-23

AI Summary

This study addresses the challenge of effective defensive maneuvering for autonomous race cars operating near the dynamic limits when confronted by faster opponents attempting overtakes. The authors propose a hierarchical reinforcement learning–guided model predictive control framework that uniquely integrates spatial occupancy regulation with geometry-aware trajectory generation. In this architecture, a high-level policy based on Soft Actor-Critic learns defensive strategies in the Frenet frame, while a low-level nonlinear model predictive controller embeds tire friction constraints to enforce spatial regularization. Evaluated on the Thunderhill West circuit, the approach increases the average time before being overtaken from 8.8 to 14.6 seconds, significantly reducing opponent progress. The system operates in real time with an average solve time of 33.3 milliseconds and utilizes 83.4% of available tire force, achieving a balanced trade-off between performance and feasibility.

Abstract

This paper addresses defensive blocking in autonomous racing, where a vehicle must prevent a faster opponent from overtaking while operating near its dynamic limits. Different from lap-time minimization, we formulate defense as a spatial occupancy regulation problem via a hierarchical reinforcement-learning guided model predictive control framework. A Soft Actor-Critic strategic layer operates in the Frenet domain to generate geometry-aware defensive references, which are embedded into the nonlinear model predictive control formulation as spatial regularization under friction constraints. Evaluated on the Thunderhill West circuit in simulation, the framework increases average overtake time from 8.8 s to 14.6 s while significantly reducing opponent progress. Meanwhile, it allows the vehicle to utilize 83.4% of available tire force. The framework achieves a 33.3 ms mean solve time (13.9 ms std), supporting real-time high-speed adversarial interaction.

Problem

• autonomous racing
• defensive blocking
• overtake prevention
• spatial occupancy regulation
• adversarial interaction

Innovation

• reinforcement learning
• model predictive control
• autonomous racing
• defensive blocking
• Frenet frame