Advanced model predictive control for trajectory tracking of mobile robots with complex dynamics

Abstract

This thesis presents the development of advanced control algorithms for tail-sitter unmanned aerial vehicles (UAVs) and unmanned ground vehicles (UGVs) in challenging operating scenarios. For the tail-sitter UAV, a hybrid model predictive control (HMPC) framework was developed based on a neural network (NN)-augmented aerodynamic model. This model offers a balance between accuracy and computational efficiency, enabling real-time position tracking during hovering flight. The HMPC demonstrated a 57% reduction in tracking error compared to traditional nonlinear model predictive control (NMPC) and showed robust performance under wind disturbances up to 3 m/s. For the UGV, the research focused on autonomous drifting control using a MPC based approach. A novel control framework was designed, incorporating an MPC-based trajectory generator and a cascaded PID-based controller to handle the vehicle's complex dynamics and actuator coupling during drifting maneuvers. The proposed control strategy was validated in real-world experiment on a 1/10 scale RC car, achieving precise path tracking and stability in aggressive drifting scenarios. The results demonstrate that the developed control frameworks significantly enhance the performance and robustness of both UAVs and UGVs in their respective challenging conditions. This work contributes to the advancement of autonomous vehicle control technologies, with potential applications in urban environments, autonomous racing, and other dynamic operational contexts.