UAMDynCon-DT: A Data-driven Dynamics and Robust Control Framework for UAM Vehicle Digitalization Using Deep Learning

This study presents a data-driven dynamics and robust control framework, referred to as UAMDynCon-DT, for the accurate cloning of ground-truth dynamics and the robust control of attitude in a personal aerial vehicle (PAV) known as the KP-1 eVTOL aircraft. The proposed framework utilizes exponentially stabilizing control Lyapunov functions to ensure stability and robustness, and utilizes a nonlinear flight dynamics model of the KP-1 to construct the controller. To evaluate its performance, real-world flight tests were conducted on a scaled model of the KP-1, and results were compared to those obtained from traditional control methods. The tests demonstrated improved resilience to disturbances and uncertainties when utilizing the proposed framework. This research suggests that UAMDynConDT could be a valuable tool for the future development of large-scale urban air mobility vehicles.