Co-Learning of Task and Sensor Placement for Soft Robotics

Andrew Spielberg,Lillian Chin,Alexander Amini,Daniela Rus,Wojciech Matusik

doi:10.1109/lra.2021.3056369

Andrew Spielberg, Lillian Chin + Show 3 more

Open Access

https://doi.org/10.1109/lra.2021.3056369

Copy DOI

Abstract

Unlike rigid robots which operate with compact degrees of freedom, soft robots must reason about an infinite dimensional state space. Mapping this continuum state space presents significant challenges, especially when working with a finite set of discrete sensors. Reconstructing the robot's state from these sparse inputs is challenging, especially since sensor location has a profound downstream impact on the richness of learned models for robotic tasks. In this work, we present a novel representation for co-learning sensor placement and complex tasks. Specifically, we present a neural architecture which processes on-board sensor information to learn a salient and sparse selection of placements for optimal task performance. We evaluate our model and learning algorithm on six soft robot morphologies for various supervised learning tasks, including tactile sensing and proprioception. We also highlight applications to soft robot motion subspace visualization and control. Our method demonstrates superior performance in task learning to algorithmic and human baselines while also learning sensor placements and latent spaces that are semantically meaningful.

Highlights

R ECENT efforts in soft robotics research have led to breakthroughs in soft robot modeling and design [1]–[3]
In order to realize the dream of fully untethered soft robots, intrinsic, on-board information must be used. To address this need for intrinsic soft robotic modeling, we focus on the problem of co-learning optimal sensor placements and models for general supervised tasks
We propose a neural architecture for simultaneously learning soft robotic tasks and the optimal sensor placement for that task

Summary

Introduction

R ECENT efforts in soft robotics research have led to breakthroughs in soft robot modeling and design [1]–[3]. Most of these victories, have relied on virtual simulation environments where full state information of a system’s dynamics is visible to a controller. In the physical world, soft robots’ continuum bodies are high/infinite dimensional. Several approaches have been proposed for efficient reasoning about the high-dimensional state of soft robots [4], [5], they rely. Date of publication February 2, 2021; date of current version February 17, 2021. This letter was recommended for publication by Associate Editor H.

Methods

Results

Conclusion