Artificial Neural Network Design for Stability Control of an Unmanned Aerial Vehicle (UAV)
Abstract
The project is aimed at modelling an intelligent flight control system based on a neural network which compensates for unexpected flight dynamics during UAV operations. This implies maintaining UAV balance during the flight dynamics. The
control system is to address the challenge of poor performance exhibited by conventional flight control systems when faced with these dynamics.
Poor performance during unexpected flight dynamics can result in loss of control over the UAV potentially causing damage to the aircraft, its payload and the people or property in the vicinity. It is therefore necessary to improve flight control system
performance in order to promote safe flight missions.
The neural network based control system is modelled and trained using reinforcement learning while data is collected from a simulation environment. Results from this training process are obtained and analyzed to observe system performance. This performance is further related to the performance of the conventional PID control system. The neural network based system is seen to exhibit greater performance than the PID system in regards to stability control of a UAV.