Adaptive Neural Network Control for Trajectory Tracking and Vibration Suppression in Flexible-Link Parallel Manipulators

Robotic manipulators are widely employed in applications that demand both high precision and high speed. Among the different configurations, parallel manipulators with flexible links have emerged as a promising alternative due to their higher energy efficiency, improved compliance, lighter structures, and smaller actuators. However, the reduction in mass increases the number of degrees of freedom, which makes mathematical modeling more complex and also induces undesired vibrations in the mechanical system. In this sense, an adaptive control based on a neural network is designed to both improve the tracking trajectory and attenuate the negative effects mainly caused by the low thickness of the links. The dataset to train the ANN is generated based on different trajectory configurations of the prototype. The end-effector is initially placed at different locations of the workspace to investigate the flexible characteristics over distinct scenarios. Once tested, additional signals (not used previously in the training phase) are randomly selected to validate the proposed strategy in experimental trials. The control strategy is then calculated by summing the pose loop, based on a PID strategy, and the adaptive control law, based on the ANN model. Experimental results demonstrate that the proposed control approach achieves at least a 27.70% reduction compared to the Pose loop scheme, highlighting its effectiveness in enhancing tracking accuracy in flexible-link parallel manipulators. This proposed control strategy is used to compensate in real-time for the overall dynamics regardless of the trajectory, model uncertainties, or undesired effects.