Motion generation of a soft sheet-like swimming robot using physical reservoir computing

Abstract

In recent years, swimming robots have been developed to achieve efficient propulsion and high maneuverability. Previous researches have tried to achieve fish-like swimming by control based on physical models and top-down architectures, but have encountered problems due to high complexity of the underwater environment. Other researches have tried to overcome these problems by mimicking the physical properties of fish. To achieve more intelligent swimming, we argue that a control architecture that uses physical dynamics as a computational resource is needed. In this study, a soft sheet-like swimming robot achieves motion emulation and environment estimation through a control architecture based on physical reservoir computing.