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

This project develops and tests a constrained neural‑network model predictive control (NNMPC) method that uses the Archimedes Optimization Algorithm (AOA) for robot arms. Standard MPC methods often become slow when they have to handle very non‑linear, multi‑input multi‑output (MIMO) systems that must obey tight physical limits. To overcome this, we use a neural network to approximate the robot arm’s complex dynamics, and we use that network as the predictor inside the MPC. The constrained NNMPC leads to a non‑convex optimization problem. We solve it online with the Archimedes Optimization Algorithm, a metaheuristic that imitates buoyancy. The goal is to improve trajectory tracking, reduce control effort, and keep joint torque and

speed within their limits. We run extensive simulations to test the AOA‑based NNMPC with different payloads and external disturbances. The work also gives step‑by‑step guidance on modeling non‑linear robots, building metaheuristic‑based predictive controllers, and comparing their performance to classic methods such as Particle Swarm Optimization (PSO).

Message: Hi FE, I need help with “Constrained Neural Network Model Predictive Controller Based on Archimedes Optimization Algorithm with Application to Robot Manipulators” in “Electrical Engineering”