Constrained Neural Network-based Model Predictive Control for Quadrotors Using the Sine Cosine Algorithm
Abstract
In this paper, an efficient nonlinear control algorithm, called Constrained Neural Networks based Model Predictive control using Sine Cosine Algorithm (CNNMPC-SCA) is developed to control the dynamics of quadrotors. The main objective is to design an efficient controller for quadrotors that ensures satisfactory performance while minimizing the gap between the quadrotor positions and the reference trajectories. Indeed, a novel dynamic model architecture of the quadrotor is developed using several Nonlinear Autoregressive Exogenous (NARX) neural networks, this model aims to accurately predict the future behavior of the quadrotor within a short and acceptable time frame, making it suitable for implementation in the control process. The designed model was validated and then integrated into the CNNMPC-SCA algorithm. Furthermore, the metaheuristic algorithm known as the Sine Cosine Algorithm (SCA) was modified and employed to solve the non-convex, nonlinear optimization problem of the proposed predictive controller. To assess the efficiency of the proposed CNNMPC-SCA algorithm, a comparative study was conducted using the Adaptive Fuzzy PID controller and the hybrid Fuzzy PID controller. The obtained results demonstrate that the proposed control algorithm achieves better control performances compared to those obtained using the other considered controllers.