Abstract
Soft robots are becoming extremely popular thanks to their intrinsic safety to contacts and adaptability. However, the potentially infinite number of Degrees of Freedom makes their modeling a daunting task, and in many cases only an approximated description is available.
This challenge makes reinforcement learning (RL) based approaches inefficient when deployed on a realistic scenario, due to the large domain gap between models and the real platform.
In this work, we demonstrate, for the first time, how Domain Randomization (DR) can solve this problem by enhancing RL policies with: i) a higher robustness w.r.t. environmental changes; ii) a higher affordability of learned policies when the target model differs significantly from the training model; iii) a higher effectiveness of the policy, which can even autonomously learn to exploit the environment to increase the robot capabilities (environmental constraints exploitation).
Moreover, we introduce a novel algorithmic extension of previous adaptive domain randomization methods for the automatic inference of dynamics parameters for deformable objects.
We provide results on four different tasks and two soft robot designs, opening interesting perspectives for future research on Reinforcement Learning for closed-loop soft robot control.
Authored by Gabriele Tiboni, Andrea Protopapa, Tatiana Tommasi, Giuseppe Averta.
A policy is trained on a simplified model of the soft MultiGait robot design with domain randomization (DR) on the Poisson's Ratio, Young's Modulus and mass of its constituent materials. Later, the policy is evaluated on both (left) the simplified model and (right) the full accurate model, which is about ~8x slower. Training with DR allows to train policies on affordable models and transfer their behavior effectively to more accurate models, which resemble the real robot dynamics.
Findings
Coming Soon
Citing
@misc{tiboni2023dr_soro,
doi = {10.48550/ARXIV.2303.04136},
title = {Domain Randomization for Robust, Affordable and Effective Closed-loop Control of Soft Robots},
author = {Tiboni, Gabriele and Protopapa, Andrea and Tommasi, Tatiana and Averta, Giuseppe},
publisher = {arXiv},
year = {2023}
}