Publications

Learning-In-The-Loop Optimization: End-To-EndControl And Co-Design of Soft Robots ThroughLearned Deep Latent Representations

Publication

NeurIPS 2019, 2019

Authors

Andrew Spielberg, Allan Zhao, Tao Du, Yuanming Hu, Daniela Rus, Wojciech Matusik

Abstract

Soft robots have continuum solid bodies that can deform in an infinite number of ways. Controlling soft robots is very challenging as there are no closed form solutions.  We present a learning-in-the-loop co-optimization algorithm in which a latent state representation is learned as the robot figures out how to solve the task. Our solution marries hybrid particle-grid-based simulation with deep, variational convolutional autoencoder architectures that can capture salient features of robot dynamics with high efficacy. We demonstrate our dynamics-aware feature learning algorithm on both 2D and 3D soft robots, and show that it is more robust and faster converging than the dynamics-oblivious baseline.  We validate the behavior of our algorithm with visualizations of the learned representation.

Paper

learning_in_the_loop.pdf

CDFG

© 2020 The Computational Design & Fabrication Group