Constrained PSO/TLBO-Optimized Neural Network Predictive Control for Nonlinear Quadrotor Systems

Abstract

Translate Article

In this paper, we propose a nonlinear model predictive controller (NMPC) that employs a feedforward neural network as the prediction model, with the underlying optimization problem solved separately using either Particle Swarm Optimization (PSO) or Teaching-Learning-Based Optimization (TLBO). A quadrotor system is chosen as the application platform due to its underactuated, multi-input multi-output (MIMO) nature and highly nonlinear dynamics, conditions under which NMPC is particularly effective, especially in handling constraints. A Type-2 fuzzy logic controller and a PSO-optimized PID controller are used to compare the performance of the suggested neural network-based NMPC framework.