DESIGN AND PERFORMANCE EVALUATION OF NEURAL NETWORK CONTROLLER APPLICATION FOR AUTOPILOT OF AN UNMANNED AERIAL VEHICLE (UAV)

Abstract

This work is aimed at designing Neural Autopilot controllers for Pitch and Roll channels using linear aircraft model, created with MATLAB / SIMULINK tool-box and investigation of the Model Response during various phases of flight like climb, cruise and landing using MATLAB. The behavior of aircraft can be described using a set of non-linear differential equations, assuming six degrees of freedom (3 linear motions and 3 angular motions) about x, y & z axes. The solutions to these set of equations are very complex and for investigation purpose they can be approximated to linear differential equations assuming 3 degrees of freedom for the subject aircraft model. The input to the system for pitch attitude are U, ,, _dot and h. The output to pitch control channel is e , the elevator deflection. The input to the system for roll attitude are  and r. The output to roll control channel is r , the aileron deflection. The simulation data in the above study are converted to discrete...