Constrained Neural Network Model Predictive Controller Based on Archimedes Optimization Algorithm with Application to Robot Manipulators

This paper introduces a novel neural network model predictive controller based on Archimedes optimization algorithm to control constrained nonlinear systems, namely robot manipulators. Neural network models of a simple structure are used to accurately predict the system’s future outputs. The new Archimedes optimization algorithm is utilized to solve the optimization problem of model predictive control and compute the optimal control action; this algorithm was widely used in different engineering fields due to its good accuracy and convergence speed. The performances of the suggested control algorithm are investigated by simulating a two-degrees-of-freedom robot manipulator. The obtained results are compared with those of various techniques, namely the PID controller, the computed torque controller, and the neural network-based model predictive control using the teaching–learning-based optimization and the particle swarm optimization. To complete the study, the developed controller is implemented on a DSP board and used to control a three-degrees-of-freedom robot manipulator; its results are compared to the neural network-based model predictive control based on the teaching–learning-based optimization and the particle swarm optimization. The simulation and the experimental results demonstrate that the proposed controller provides satisfactory robustness and accuracy, can handle constraints, and can be used to control systems with fast dynamics.