Regenerating Soft Robots through Neural Cellular Automata

Abstract

Morphological regeneration is an important feature that highlights the environmental adaptive capacity of biological systems. Lack of this regenerative capacity significantly limits the resilience of machines and the environments they can operate in. To aid in addressing this gap, we develop an approach for simulated soft robots to regrow parts of their morphology when being damaged. Although numerical simulations using soft robots have played an important role in their design, evolving soft robots with regenerative capabilities have so far received comparable little attention. Here we propose a model for soft robots that regenerate through a neural cellular automata. Importantly, this approach only relies on local cell information to regrow damaged components, opening interesting possibilities for physical regenerable soft robots in the future. Our approach allows simulated soft robots that are damaged to partially regenerate their original morphology through local cell interactions alone and regain some of their ability to locomote. These results take a step towards equipping artificial systems with regenerative capacities and could potentially allow for more robust operations in a variety of situations and environments. The code for the experiments in this paper is available at: github.com/KazuyaHoribe/RegeneratingSoftRobots.

Growing 3D Artefacts and Functional Machines with Neural Cellular Automata

“…They also only predict the "outer" layer of the 3d structures, improving efficiency with the added sparseness, but the tradeoff is the lack of detail within the interior of a structure. Horibe et al (2021) equips simulated soft robots with the ability to partially regenerate and regain locomotive capacity using two NCAs, one for generating the initial morphology and the other for restoration once the robot is damaged. The transition for a voxel (cell) is determined by a neural network using cellular states from the neighboring four voxels.…”

“…Compared to previous literature, our work has higher dimensionality (3D instead of 2D) than Mordvintsev et al (2020) and Ruiz et al (2021), more construction unit types than Zhang et al (2021), and we capture all surrounding cells with a 3D convolution instead of only the four immediate neighbors as in Horibe et al (2021).…”

“…Works on simulating morphogenesis with NCAs have primarily focused on easilyreplicable artefacts, such as pixelated emojis (Mordvintsev et al, 2020) and 2D images (Ruiz et al, 2021). Recent progress has also demonstrated its capability in generating high-fidelity 3D shapes (Zhang et al, 2021) and regenerating damaged morphology (Horibe et al, 2021). However, significant gaps still exist between the current state-of-theart and real-life applications for morphogenesis simulation.…”