Model-Predictive Control With Inverse Statics Optimization for Tensegrity Spine Robots

Abstract

Robots with flexible spines based on tensegrity structures have potential advantages over traditional designs with rigid torsos. However, these robots can be difficult to control due to their high-dimensional nonlinear dynamics and actuator constraints. This article presents two controllers for tensegrity spine robots, using model-predictive control (MPC) and inverse statics (IS) optimization. The controllers introduce two different approaches to making the control problem computationally tractable. The first utilizes smoothing terms in the MPC problem. The second uses a new IS optimization algorithm, which gives the first feasible solutions to the problem for certain tensegrity robots, to generate reference input trajectories in combination with MPC. Tracking the IS reference input trajectory significantly reduces the number of tuning parameters. The controllers are validated against simulations of 2-D and 3-D tensegrity spines. Both approaches show noise insensitivity and low tracking error and can be used for different control goals. The results here demonstrate the first closed-loop control of such structures.