Soft robots can be used in various spheres, such as agriculture, medicine, and defense. However, their complex physics means that they are hard to control. Current simulation testbeds are insufficient for taking the full advantage of elasticity.
A recent paper on arXiv.org proposes Elastica, a simulation environment tailored to soft robot context. It tries to fill the gap between conventional rigid body solvers, which are incapable to model complex continuum mechanics, and high-fidelity finite elements methods, which are mathematically cumbersome. Elastica can be used to simulate assemblies of soft, slender, and compliant rods and interface with major reinforcement learning packages. It is shown how most reinforcement learning models can learn to control a soft arm and to complete successively challenging tasks, like 3D tracking of a target, or maneuvering between structured and unstructured obstacles.
Soft robots are notoriously hard to control. This is partly due to the scarcity of models able to capture their complex continuum mechanics, resulting in a lack of control methodologies that take full advantage of body compliance. Currently available simulation methods are either too computational demanding or overly simplistic in their physical assumptions, leading to a paucity of available simulation resources for developing such control schemes. To address this, we introduce Elastica, a free, open-source simulation environment for soft, slender rods that can bend, twist, shear and stretch. We demonstrate how Elastica can be coupled with five state-of-the-art reinforcement learning algorithms to successfully control a soft, compliant robotic arm and complete increasingly challenging tasks.