A Model Predictive Control Based Autopilot for Aircraft

This paper examines the use of model predictive control as an autopilot (for aircraft) that can follow prescribed flight trajectories. Aircraft dynamics were researched and it was found that the multiple inputs of the aircraft (elevator, aileron, rudder); affected each of the multiple outputs of the system (latitude, altitude, longitude) and therefore could not be coupled separately. Given this, multiple single-input single-output controllers would not suffice as the controllers would hinder each other’s efforts to control their set variables. Therefore a multi-input multi-output (MIMO) controller which could produce actuations that considered all objectives was needed. Model predictive control (MPC) is a widely used MIMO controller in many engineering fields e.g. autonomous driving and industrial process. The controller was researched and was found to be a good fit for the application. A realistic flight trajectory was made using FlightGear flight simulator and was then exported. MPC was then implemented to control a model of the aircraft in a Simulink simulation to follow the trajectory from FlightGear. The conclusions found were that MPC was a viable autopilot strategy for aircraft if the trajectory was not too aggressive and small delays in timing were allowed.