Information-Efficient Model Identification for Tensegrity Robot Locomotion

This paper aims to identify in a practical manner unknown physicalparameters, such as mechanical models of actuated robot links, which are critical in dynamical robotictasks. Key features include the use of an off-the-shelf physics engineand the data-efficient adaptation of a black-box Bayesian optimizationframework. The task being considered is locomotion with a high-dimensional, compliant Tensegrity robot. A key insight in this case is the need to project the system identification challenge into an appropriate lower dimensionalspace. Comparisons with alternatives indicate that the proposed method can identify the parameters more accurately within the given time budget, which also results in more precise locomotion control.