Generalizable and Fast Surrogates: Model Predictive Control of Articulated Soft Robots using Physics-Informed Neural Networks

TL;DR

This paper introduces physics-informed neural networks (PINNs) for articulated soft robots, achieving fast, generalizable, and data-efficient predictions that enable real-time model predictive control with high accuracy and speed.

Contribution

It demonstrates the use of PINNs for soft robot control, significantly reducing training data needs and surpassing traditional models in prediction speed while maintaining accuracy.

Findings

PINNs outperform recurrent neural networks in generalizability.

Prediction speed exceeds traditional FP models by up to 467 times.

Real-time control at 47 Hz achieved in dynamic experiments.

Abstract

Soft robots can revolutionize several applications with high demands on dexterity and safety. When operating these systems, real-time estimation and control require fast and accurate models. However, prediction with first-principles (FP) models is slow, and learned black-box models have poor generalizability. Physics-informed machine learning offers excellent advantages here, but it is currently limited to simple, often simulated systems without considering changes after training. We propose physics-informed neural networks (PINNs) for articulated soft robots (ASRs) with a focus on data efficiency. The amount of expensive real-world training data is reduced to a minimum -- one dataset in one system domain. Two hours of data in different domains are used for a comparison against two gold-standard approaches: In contrast to a recurrent neural network, the PINN provides a high generalizability. The prediction speed of an accurate FP model is exceeded with the PINN by up to a factor of 467 at slightly reduced accuracy. This enables nonlinear model predictive control (MPC) of a pneumatic ASR. Accurate position tracking with the MPC running at 47 Hz is achieved in six dynamic experiments.