Modeling and Prediction of Powered Parafoil Unmanned Aerial Vehicle Throttle and Servo Controls through Artificial Neural Networks

This study proposes a framework for developing a realistic model for throttle and servo control algorithms for a Powered Parafoil Unmanned Aerial Vehicle (PPUAV) using Artificial Neural Networks (ANN). Two servo motors on an L-shaped platform, controls and steers the PPUAV. Six degrees of freedom (DOF) mathematical model of a dynamic parafoil system is built to test the technique's efficacy using a simulation in which disturbances mimic actual flights. A guiding law is then established, including the cross-track error and the line of sight approach. Furthermore, a path-following controller is constructed using the proportional-integral-derivative (PID), and a simulation platform was created to evaluate numerical data illustrating the route's validity following the technique. PPUAV was developed, built, and instrumented to collect real-time flight data to test the controller. These dynamic characteristics were sent into the ANN for training. A diverging-converging design was identified to obtain the best consistency between predicted and observed Throttle and servo control values. For a comparable flight route, the control signal of the simulated model is compared to that of the actual and ANN predicted models. The comparative findings show that the ANN-predicted and actual control inputs were almost identical, with an 80-99 % match. However, the simulated response showed deviation from the actual control input, with an accuracy of 50-80%.