Controlling swarm dynamics is challenging and has long been an attractive research field because swarms provide a fundamental insight of locally interacting systems’ emergent behaviors. For example, a sheepdog type navigation control has been studied recently, where swarms consist of two different agents: passive sheep and active sheepdogs. In this paper, we focused on the swarm predator system with a swarm that has a number of passive agents and a single active predator agent. Recently, reservoir computing (RC) was introduced as a new way to control swarms. RC offers an easy and analyzable way to find optimal controllers. In this paper, we suggest a new way to read the swarms’ state for controlling swarm predator systems, named relatively ordered state (ROS), where the agents’ IDs are reordered at each time step by relative distances from the predator. The ROS is robust against the swarm’s initial condition’s difference, despite the simpleness and naturalness of the process of the ROS. We found that a swarm within the critical phases of order disorder phase transition like structures can bring out the swarm’s potential to be a reservoir both in open-loop and closed-loop experiments. In this closed-loop control, the predator determines its own movement via the collective dynamics of the swarm like a serpent eating its own tail in the classic “Ouroboros.”

In recent years, swimming robots have been developed to achieve efficient propulsion and high maneuverability that are possessed naturally by fish. Previous studies have attempted to achieve swimming similar to fish by control based on physical models and top-down architectures, but have encountered problems due to the high complexity of the underwater environment. Several research works have tried to overcome these problems by exploiting embodiment—that is, by mimicking the physical properties of fish. To achieve more intelligent swimming from the perspective of the embodiment, we focused on a framework called physical reservoir computing (PRC). This framework allows us to utilize physical dynamics as a computational resource. In this study, we propose a soft sheet-like swimming robot and a PRC-based architecture that can be used to emulate swimming motions by exploiting its own body dynamics for closed-loop control. Through experiments, we demonstrated that our system satisfies the properties required for learning swimming motion through supervised learning. We also succeeded in robust motion generation and environmental state estimation, opening up future prospects for more intelligent robot control and sensing.