Neural networks meet PID control: Revolutionizing manipulator regulation with gravitational compensation

Marco Moran-Armenta; Jorge Montoya-Cháirez; Francisco G. Rossomando; Emanuel Slawiñski; Vicente Mut; Chicaiza Claudio, Fernando; Javier Moreno-Valenzuela

doi:10.1016/j.ifacsc.2025.100306

Neural networks meet PID control: Revolutionizing manipulator regulation with gravitational compensation

Journal

IFAC Journal of Systems and Control

ISSN

2468-6018

Date Issued

2025

Author(s)

Marco Moran-Armenta

Jorge Montoya-Cháirez

Francisco G. Rossomando

Emanuel Slawiñski

Vicente Mut

Chicaiza Claudio, Fernando

Centro de investigación en Mecatrónica y Sistemas Interactivos

Javier Moreno-Valenzuela

Type

journal-article

DOI

10.1016/j.ifacsc.2025.100306

URL

https://cris.indoamerica.edu.ec/handle/123456789/9279

Abstract

This research proposes an innovative approach to improve the performance of regulation control systems in manipulators by combining PID control with gravitational compensation using neural networks. In this work, a modified PID control structure that incorporates a gravitational compensation term given by a neural network is introduced, thus allowing a more precise and adaptive response to gravitational and dynamic perturbations of the system. Furthermore, the controller's performance is evaluated through real-time experiments in two manipulators, comparing its performance with the same structure, one without integral action, another without neural compensation and the last one assuming that the gravity vector is known. The results show a significant improvement in system regulation accuracy, demonstrating the proposed controller's effectiveness.