Neural fidelity warping for efficient robot morphology design

Abstract

We consider the problem of optimizing a robot morphology to achieve the best performance for a target task, under computational resource limitations. The evaluation process for each morphological design involves learning a controller for the design, which can consume substantial time and computational resources. To address the challenge of expensive robot morphology evaluation, we present a continuous multi-fidelity Bayesian Optimization framework that efficiently utilizes computational resources via low-fidelity evaluations. We identify the problem of non-stationarity over fidelity space. Our proposed fidelity warping mechanism can learn representations of learning epochs and tasks to model nonstationary covariances between continuous fidelity evaluations which prove challenging for off-the-shelf stationary kernels. Various experiments demonstrate that our method can utilize the low-fidelity evaluations to efficiently search for the optimal robot morphology, outperforming state-of-the-art methods.

Efficient Hyperparameter Optimization for Physics-based Character Animation

“…Morphology design and optimization is a problem in both computer animation and robotics [Agrawal et al 2014;Bongard 2011;Geijtenbeek et al 2013;Ha et al 2017;Hu et al 2020;Huang et al 2020;Liao et al 2019;Lipson and Pollack 2000;Ma et al 2021;Park and Asada 1994;Paul and Bongard 2001;Pil and Asada 1996;Sims 1994;Spielberg et al 2017;Villarreal-Cervantes et al 2012;Wang et al 2018;Won and Lee 2019]. Traditional modelbased methods require accurate dynamic models, and only work for specific types of control algorithms such as trajectory optimization [Ha et al 2017;Spielberg et al 2017].…”