Reconfigurable autopilot design for a high performance aircraft using model predictive control

Abstract

The losses of military and civilian aircraft due to control surface failures have prompted research into controllers with a degree of reconfiguration. This thesis will describe a design approach incorporating Model Predictive Control (MPC) with a self updating model to achieve a level of reconfiguration in a generic high performance aircraft. MPC has the advantage of explicitly taking a model of the failed system and incorporating it into a receding horizon optimization problem. MPC also has the added benefits of allowing constraints on the inputs, outputs, and states of the system as well as tuning flexibility. This thesis describes the development of four types of MPC autopilots. A description of the controller implementation and failure implementation is also included. Each autopilot is subject to a surface failure during certain times in a sample maneuver and the resulting controller adaptation is analyzed. All MPC controllers are found to maintain good performance in the event of certain failures with an updated internal model. It is when the internal model is not updated that full performance is not recovered and in some cases, loss of the aircraft results.