Robust Adaptive Control of Aerial Vehicles under Significant Model Uncertainty

Abstract

Control systems for flying vehicles must satisfy stringent demands for high performance while remaining robust against system uncertainties, such as aerodynamic variations and environmental disturbances. Conventional controllers often face a fundamental trade-off between nominal performance and robustness. This thesis investigates the robustness guarantees and performance of adaptive control methods to mitigate these challenges. The primary focus is the design and implementation of an L1 Composite Model Reference Adaptive Controller. This architecture utilizes a cascade configuration where an outer loop employs direct MRAC to ensure tracking, while an inner loop retains an L1 adaptive structure with a fast predictor to satisfy small-gain stability conditions. The vehicle is modeled as a Linear Parameter-Varying system with decoupled dynamics for roll, pitch, and yaw. The performance of the L1 CMRAC is systematically benchmarked against a conventional Linear Quadratic controller and a standard direct MRAC within a sixdegree- of-freedom simulation environment. Evaluation is conducted across multiple reference paths designed to excite various flight dynamics and cross-coupling effects. To ensure statistical robustness, Monte Carlo simulations are utilized to quantify success rates and tracking accuracy under a broad range of uncertainty conditions. Results indicate that while the baseline LQ controller may achieve a lower rootmean- square error in nominal scenarios, the L1 CMRAC provides a higher success rate and superior robustness under significant perturbations. The analysis further highlights an inherent trade-off within the L1 framework: conservative predictor