Predictive motion planning and control for autonomous racing under regulations: from compliance to strategic awareness

Autonomous driving in complex and dynamic environments demands decision-making systems that can operate safely, efficiently, and strategically while accounting for the behavior of other agents. As autonomy moves toward real-world deployment, vehicles must not only ensure safety and performance but also adhere to the regulations that govern interactions among traffic participants. In this context, autonomous racing serves as a powerful research platform for studying autonomy at its physical and computational limits, where high speeds, large accelerations, and competitive interactions expose the fundamental challenges of motion planning and control. This thesis addresses these challenges by proposing a comprehensive motion planning and control stack for head-to-head autonomous racing with full-scale vehicles. The proposed framework bridges predictive optimization–based planning with the requirements of competitive multi-vehicle racing, enabling real-time execution of dynamically feasible and strategically competitive maneuvers in full-scale autonomous racing competitions. While real-world interactions are typically governed by explicit rules naturally understood and exploit by human drivers, most existing predictive planning approaches neglect such logical and asymmetric regulations. To address this gap, the thesis develops methods for regulation-compliant planning, in which racing rules are explicitly encoded through mixed-integer optimization, and for regulation-aware planning, in which agents strategically reason about regulations constraining their opponents within a game-theoretic framework. Simulation and experimental results demonstrate that the proposed formulations improve behavioral realism, producing competitive yet rule-compliant interactions. The contributions advance the state of the art in predictive motion planning and control toward real-time, regulation-aware autonomy applicable to highly dynamic and interactive driving scenarios.