Efficient Model Identification for Tensegrity Locomotion

Abstract

This paper aims to identify in a practical manner unknown physical parameters, such as mechanical models of actuated robot links, which are critical in dynamical robotic tasks. Key features include the use of an off-the-shelf physics engine and the Bayesian optimization framework. 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 model identification challenge into an appropriate lower dimensional space for efficiency. 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.

Adaptive tensegrity locomotion: Controlling a compliant icosahedron with symmetry-reduced reinforcement learning

“…Hardware trials indicated a good but imperfect transfer of behavior from simulation to reality after moderate calibration effort. The reality gap could likely be further narrowed with additional effort upon physical parameter calibration for the simulated system, a topic of ongoing investigation (Zhu et al, 2018), as well as more detailed model formulations for compliant contact and cable spooling. One appealing possibility would be to execute the last few iterations of the policy search using samples taken from hardware trials.…”

Not Only Domain Randomization: Universal Policy with Embedding System Identification

“…Present methods for system dynamics identification usually require aligned trajectories for constructing the contrastive loss [10], [11], which we find not to be a necessity. In previous methods, the trajectories alignment can usually be achieved with setting same initial states and executing the same policy, or rollout the simulation environment on a give state and certain action for a single timestep at each time.…”

“…Thus only suboptimal policies can be performed in the target domain even for the case that target sample data falls in the domain of randomized distributions of the source. Traditional explicit SI methods [1] directly configure the system parameter, which is testified to be an ill-posed problem [8], [10] due to that the entangled effects of multiple parameters will lead to nonunique identification results. This problem no longer exists for implicit SI [2], [3].…”

“…OSI may not be feasible since the trajectories may not provide enough information for accurately predicting the dynamics parameters in a pointestimation manner, therefore distributions can better represent the uncertainty in the dynamics identification. The problem can be ill-posed to directly predict dynamics parameters, according to [10]. Multiple potential dynamics models can lead to the same trajectories as a non-injective function, due to the entanglement of dynamics parameters.…”

Video1.MP4

“…This is a valuable feature in robots for locomotion, where interaction with the environment is inevitable. Therefore, there have been active studies on its control (Caluwaerts et al, 2014;Kim et al, 2020), and the utilization of data-driven approaches has gained significant attention in recent years (Kim et al, 2015;Zhang et al, 2017;Zhu et al, 2018;Surovik et al, 2019). However, there are also studies utilizing broad tensegrity for this type of tensegrity robot.…”