Trajectory Control of Flexible Manipulators Using Forward and Inverse Models with Neural Networks

This study investigates trajectory control in flexible manipulators through neural network-based forward and inverse modeling. Traditional improvements in manipulator precision often involve increasing rigidity, which results in greater weight and energy demands—factors that hinder usage in aerospace and other sensitive applications. Flexible manipulators, while lightweight, pose control challenges due to elastic deformations. This research proposes neural network-based models to enhance trajectory control for a two-link, three-degree-of-freedom (3-DOF) flexible manipulator. Simulation and experimental validations demonstrate that compensating for system delays in training data substantially improves tracking accuracy. However, issues with smooth trajectory generation persist. These findings emphasize the utility of neural networks in adaptive control and suggest future avenues for refining input-output modeling to close the gap between theory and implementation.