Real-time Trajectory Tracking Controller of an Aircraft using Imitation NMPC

Autonomous flying vehicles are the future of aerospace industry. They require a robust trajectory tracking algorithm capable of finding an optimal solution in real-time. Optimal control, such as Nonlinear Model Predictive Control (NMPC) is a powerful technique which can handle the nonlinearities of a system by including the restrictions and constraints of its states and controls. However, a major drawback of NMPC is that it requires the optimization problem to be solved online involving a huge amount of real-time calculations.
For increasing the level of autonomy, a data-driven approach such as imitation learning (IL) can be implemented. The Neural Network can be trained by mimicking the optimal trajectory data generated by NMPC. This project aims to develop an Imitation NMPC that combines the optimal control and machine learning to reduce the computational cost in real-time trajectory tracking of fixed-wing aircraft.

Faculty Supervisor:

Luis Rodrigues

Student:

Partner:

Universität der Bundeswehr München

Discipline:

Engineering

Sector:

Aerospace; Artificial Intelligence; Technology

University:

Concordia University

Program: