Autotuning of a Formula 1 Clutch Torque Control System

This thesis presents a hybrid model- and data-based framework for the automatic calibration of a clutch torque control system in Formula 1 powertrains. The approach integrates physics-based modeling of the nonlinear thermo-mechanical drivetrain, model identification from telemetry data, and model-free optimization of feedback and feedforward control parameters. The control architecture combines a gain-scheduled PI regulator, a band-pass compensator for torsional damping and a feedforward map derived from identified clutch characteristics. An Extremum Seeking (ES) algorithm performs autonomous tuning of controller gains from closed-loop performance, without analytical gradients or explicit linearization. The methodology has been validated through a multi-stage campaign including Model-in-the-Loop simulations and experimental evaluations in both test-bench and track environments. The complete workflow—spanning model identification, optimization, robustness assessment, and deployment—is implemented in MATLAB within a dedicated GUI, enabling reproducible, automated calibration and seamless integration with experimental control platforms.