This work concerns the application of physics-informed neural networks to the modeling and control of complex robotic systems. Achieving this goal required extending Physics Informed Neural Networks to handle non-conservative effects. We propose to combine these learned models with model-based controllers originally developed with first-principle models in mind. By combining standard and new techniques, we can achieve precise control performance while proving theoretical stability bounds. These validations include real-world experiments of motion prediction with a soft robot and of trajectory tracking with a Franka Emika manipulator.
@article{arxiv.2305.05375,
title = {Physics-informed Neural Networks to Model and Control Robots: a Theoretical and Experimental Investigation},
author = {Jingyue Liu and Pablo Borja and Cosimo Della Santina},
journal= {arXiv preprint arXiv:2305.05375},
year = {2023}
}