Coordinate Descent Optimization for Winged-UAV Design

Abstract

In this paper, a powerful optimization framework is proposed to design highly efficient winged unmanned aerial vehicle (UAV) that is powered by electric motors. In the proposed approach, the design of key UAV parameters including both aerodynamic configurations, (e.g. wing span, sweep angle, chord, taper ratio, cruise speed and angle of attack) and the propulsion systems (e.g. propeller, motor and battery) are cast into an unified optimization problem, where the optimization objective is the design goal (e.g. flight range, endurance). Moreover, practical constraints are naturally incorporated into the design procedures as constraints of the optimization problem. These constraints may arise from the preliminary UAV shape and layout determined by industrial design, weight constraints, etc. The backend of the optimization based UAV design framework are highly accurate aerodynamic models and propulsion system models proposed in this paper and verified by actual experiment data. The optimization framework is inherently non-convex and involves both continuous variables (e.g. the aerodynamic configuration parameters) and discrete variables (e.g. propulsion system combinations). To solve this problem, a novel coordinate descent method is proposed. Trial designs show that the proposed method works rather efficiently, converging in a few iterations. And the returned solution is rather stable with different initial conditions. Finally, the entire approach is applied to design a quadrotor tail-sitter VTOL UAV. The designed UAV is validated by both CFD simulations and intensive real-world flight tests.