Abstract
<title>Abstract</title> <p>Despite their widespread use, PID controllers have several issues, such as the difficulty of implementation in highly nonlinear dynamic systems and the potential for integrator wind-up. Integrator wind-up can cause performance loss under control saturation, which is why an anti-windup method is implemented in the PID loop. The standard PID control system will need significant modifications in order to offer optimal position and trajectory control in dynamically changing environments. This paper presents two new hybrid control methods that have been designed to improve the trajectory tracking performance of quadrotors. The proposed methods utilize real-time reference modulation, via an Artificial Neural Network (ANN) or a Deep Neural Network (DNN), combined with a traditional Proportional–Integral–Derivative (PID) controller. In the proposed method, the ANN or DNN receives input of the current position and angle error, and then dynamically adjusts the reference to mitigate the PID control effect. The simulation of a 3-D environment with wind disturbances demonstrated that the PID+DNN method performed better than the PID+ANN with a smoother convergence, lower root-mean-square (RMS) error of control input, and decreased control effort. This study successfully demonstrated that classical control combined with a neural network based reference adaptation is a powerful and adaptable method for enabling intelligent UAV navigation.</p>