From modeling to race strategy: multi-level and multi-agent optimization of Formula 1 power units

Modern Formula 1 Power Units represent one of the most complex optimization challenges in engineering, requiring a balance between performance, efficiency, and reliability under strict regulatory constraints. Their design involves tightly coupled subsystems, mechanical, thermal, and electrical, whose interactions are nonlinear and dynamic. This thesis introduces a suite of fast, physics-informed models and optimization frameworks that connect component-level thermodynamics, vehicle-level co-design, and multi-agent race strategies. The first contribution is a zero-dimensional heat-rejection model for high-performance Internal Combustion Engines. It captures heat transfer between combustion gases, solid structures, and fluids using a lumped-parameter energy balance. Validated against high-fidelity models and telemetry data, it predicts fluid temperatures within 1-2°C and total heat rejection with less than 2% error, while enabling real-time simulation and optimization. The second contribution is a multi-domain co-design framework for hybrid F1 vehicles. It couples hardware choices, like cooling layout, battery size, and aerodynamics, with control strategies over lap and race horizons. Using monotone system theory, it quantifies trade-offs between drag, battery aging, and cooling efficiency. Case studies show, for example, that smaller batteries may outperform larger ones due to mass and wear effects, and that high-downforce setups can be optimal over race distances despite slower lap times. The third contribution is a game-theoretic multi-agent optimization framework for competitive racing. It integrates trajectory planning, energy management, aerodynamic wake effects, and strategic interactions using Nash and Stackelberg games. A local Stackelberg refinement algorithm exploits leader advantages in typical race scenarios like overtaking and slipstreaming, aligning well with real-world behavior. Together, these models offer a unified approach to optimizing F1 Power Units across physical, systemic, and strategic layers.