SoftCon: Simulation and Control of Soft-Bodied Animals with Biomimetic Actuators

摘要

We present a novel and general framework for the design and control ofunderwater soft-bodied animals. The whole body of an animal consistingof soft tissues is modeled by tetrahedral and triangular FEM meshes. Thecontraction of muscles embedded in the soft tissues actuates the body andlimbs to move. We present a novel muscle excitation model that mimics theanatomy of muscular hydrostats and their muscle excitation patterns. Ourdeep reinforcement learning algorithm equipped with the muscle excitationmodel successfully learned the control policy of soft-bodied animals,which can be physically simulated in real-time, controlled interactively, andresilient to external perturbations. We demonstrate the effectiveness of ourapproach with various simulated animals including octopuses, lampreys,starfishes, stingrays and cuttlefishes. They learn diverse behaviors suchas swimming, grasping, and escaping from a bottle. We also implementeda simple user interface system that allows the user to easily create theircreatures.