Unveiling decentralized feedback mechanisms in sea star locomotion using deep reinforcement learning

In recent years, there has been growing interest in understanding decentralized control mechanisms in biology, in part due to the increased adaptability they offer in robotics systems. Sea stars use hundreds of hydrostatic structures called tube feet for locomotion, making them an ideal model system for studying decentralized sensing and actuation. Individual tube feet are equipped with integrated sensing and actuation and their activity is orchestrated by a nerve net that is distributed throughout the body without a central brain. How the numerous tube feet are controlled and coordinated for locomotion through such a distributed nervous system remains mostly unknown.

Here, we study tube-feet driven locomotion of sea stars in the context of a decentralized control model. We first construct mathematical models of tube feet as soft actuators consisting of force elements. We propose a decentralized control framework where individual tube feet have local autonomy over their actuation, via peripheral sensori-motor feedback loops. We then employ a deep reinforcement learning algorithm to discover the most effective feedback loops and mechanosensory cues for optimal locomotion. We find that the interaction of locally controlled tube feet leads to stable and robust locomotion with minimal cost to the nervous system in terms of sensory integration. Due to the decentralized nature of our proposed framework, we easily generalize our model to 500 tube feet and we show that the robustness of the model increases with the number of tube feet. To conclude, we comment on the utility of this system as a new paradigm for robotic movement using distributed arrays of soft actuators.