Control and Guidance of a Highly-Flexible Micro Air Vehicle Using Model Predictive Control

Abstract

Man-portable micro air vehicles (MAVs) pose unique challenges for autonomous controlled flight. They consistently operate in a low Reynolds number flight regime dominated by laminar separation and separation bubble e!ects. As a result, MAVs are sensitive to small changes in flight conditions and atmospheric disturbances, such as turbulence and wind. MAVs may also have a high degree of structural flexibility due to packaging or aerodynamic considerations. For this class of MAV, aero-structural interactions lead to changes in the vehicle’s mass distribution, aerodynamic coecients, and stability derivatives. All of these factors must be considered during autonomous MAV flight control development. This paper proposes model predictive control (MPC) as a useful tool for both MAV structure and control design. MPC is a form of control in which the current control action is obtained by solving on-line a finite-horizon optimal control problem based on a model of the plant dynamics. O!-line least squares system identification is applied to flight data acquired through high-fidelity first-principles simulation to derive a ten-state model of a MAV with simple flexible wings. A Kalman filter uses measurements at each time step to convert physical disturbances on the MAV, such as wind gusts, into equivalent disturbances acting on the states. The MPC algorithm then adjusts the target states to counteract these disturbances. Two MAV design variations are investigated. One utilizes passive wing morphing, such that the position of the flexible wings depends only on aerodynamic and connection constraint forces and moments. The other utilizes active wing morphing, in which the MPC algorithm controls certain wing properties in order to improve vehicle maneuverability. Results demonstrate fast and accurate control of both MAV variants using MPC. Topics addressed include the development of a linear dynamic model, disturbance modeling and estimation, cost-function selection, tuning, computational complexity, and structure-control system co-design.