Modeling and Control of Soft Robots Using the Koopman Operator and Model\n Predictive Control

Abstract

Controlling soft robots with precision is a challenge due in large part to\nthe difficulty of constructing models that are amenable to model-based control\ndesign techniques. Koopman Operator Theory offers a way to construct explicit\nlinear dynamical models of soft robots and to control them using established\nmodel-based linear control methods. This method is data-driven, yet unlike\nother data-driven models such as neural networks, it yields an explicit\ncontrol-oriented linear model rather than just a "black-box" input-output\nmapping. This work describes this Koopman-based system identification method\nand its application to model predictive controller design. A model and MPC\ncontroller of a pneumatic soft robot arm was constructed via the method, and\nits performance was evaluated over several trajectory following tasks in the\nreal-world. On all of the tasks, the Koopman-based MPC controller outperformed\na benchmark MPC controller based on a linear state-space model of the same\nsystem.\n