Autonomous Aircraft Navigation Control System with Machine Learning-Based Landing Target Detection

Autonomous aircraft navigation has applications in support and surveillance in environments unreasonable or impossible for a pilot. SAE Aero Design is an organization that allows students to design, build, and compete with remote control aircraft with new design constraints and success parameters coming out each year. They recognize the need for autonomous aircraft and always incorporate it into one of their competition categories. Union College's Aero team would like to gain the knowledge and experience in designing this type of system, so the aim of this project is to build a prototype autonomous aircraft and document findings for future reference. The main objectives of the system are training a machine learning algorithm for target identification and distance measurement, as well as modeling a control system capable of maintaining flight stability and navigation. A 2ft diameter red circular landing target was identified from distances ranging from 5 to 35 meters with an accuracy of 97.4% during testing. The controllers monitor speed, roll, turning, and altitude and are tuned via an electromechanical model consisting of the plane and motors. The Simulink models yield settling times under 10 seconds and a maximum error of 2% for control surface actuation and 5% for speed control. A successful project will equip the Aero team with the necessary tools and procedures to create their own autonomous aircraft.