Energy-Efficient UAV Trajectory Generation Based on System-Level Modeling of Multi-Physical Dynamics

Nicolas Michel, Zhaodan Kong, Xinfan Lin

Abstract: Electric multirotor aircraft is a major emerging technology, but currently limited by energy performance, which affects flight time and range. In this paper, energy-efficient trajectory generation is performed based on a system-level model of the multi-physical dynamics of an octorotor vehicle. An energy-optimal control problem is formulated using the model, and solutions are calculated and discussed for a range of hover-to-hover flight operations, which include both vertical and horizontal components. To allow for rapid, flexible, and computationally efficient generation of target state trajectories for real-time control applications, polynomial equations are used to approximate the optimized trajectories, which are segmented and normalized based on features of the optimized behaviors. The energy performance of the polynomial trajectories is compared to that of the original optimal trajectories in simulation with feedback control, demonstrating very close energy consumption. Additionally, a baseline PID controller is defined for comparison, and is shown to result in an increased energy use of 11.2% over feedback control using the optimized trajectories. The performances of each controller for a sample operation is compared, and behaviors which affect energy performance are discussed to demonstrate the value of the model-based trajectory generation approach.