Optimizing UAV Trajectory Tracking using Neural Network-Based PID Control in Dynamic Environments

The current study provides a comparison of Neural Network Proportional-Integral-Derivative (NN PID) control and conventional PID control of Unmanned Aerial Vehicles (UAVs) for the purpose of path-following control. This study evaluates both types of control in terms of trajectory tracking accuracy, errors, and dynamic performance during the flight. For this purpose, the UAV dynamics are modeled to capture essential UAV properties (pitch, roll, and yaw dynamics) and positional dynamics of the UAV in the X, Y, and Z axes. The methodology involves implementing both NN PID and conventional PID controllers on a simulated UAV platform, with the desired trajectory serving as the reference path. The NN PID controller leverages machine learning techniques to adaptively tune its parameters, aiming to enhance tracking precision and robustness. Simulation results are presented, comparing the two controllers across various metrics, including position error, angle error, and overall trajectory deviation. It is found that the NN PID controller performs better than the conventional PID controller in terms of reducing steady state errors and improved transient response, especially in complex flight scenarios. The findings suggest that the adaptive nature of the NN PID controller makes it a promising alternative for UAV applications requiring high precision and stability. This study contributes to the ongoing research in UAV control systems by providing empirical evidence on the advantages of integrating neural networks with traditional control methods. The results underscore the potential of NN PID controllers in advancing UAV autonomy and operational efficiency.