Robustez del control hı́brido neuro-PD ante incertidumbres y perturbaciones

Abstract

This work presents the development and validation of a hybrid control system for a two-degree-of-freedom (2-DOF) robotic manipulator. The proposed scheme integrates a Multi-Layer Perceptron (MLP) Artificial Neural Network (ANN), designed to approximate the inverse dynamics for system linearization, with a Proportional-Derivative (PD) regulator to close the feedback loop. The robustness of the architecture is evaluated through stress tests involving parametric uncertainty in the distal link mass, Gaussian sensor noise, and external torque disturbances. Experimental simulation results demonstrate that the hybrid controller effectively mitigates the drift characteristic of purely machine learning-based models, achieving precise trajectory tracking even under 900% variations in the robot’s load.