Model and Data Based Approaches to the Control of Tensegrity Robots

Abstract

This letter proposes two approaches to control the shape of the structure or the position of the end effector for a soft-robotic application. The first approach is a model-based approach where the non-linear dynamics of the tensegrity system is used to regulate position, velocity and acceleration to the specified reference trajectory. The formulation uses state feedback to obtain the solution for the control (tension in the strings) as a linear programming problem. The other model-free approach is a novel decoupled data-based control (D2C) which first optimizes a deterministic open-loop trajectory using a black-box (no actual model) simulation model and then develops a linear quadratic regulator around the linearized open-loop trajectory. A two-dimensional tensegrity robotic reacher is used to compare the results for both the approaches for a given cost function. The D2C approach is also used to study two more complex tensegrity examples whose dynamics is difficult to model analytically.