Soft robots harness their built‐in mechanical intelligence to respond directly to their environment. However, they typically still depend on predefined sequences to coordinate their limbs, and external centralized hardware is often used for coordination in changing circumstances. In contrast, in nature, invertebrates like echinoderms distribute behavioral control throughout their body. Inspired by this decentralized computation strategy, we present a modular soft robotic system in which each limb independently adjusts the timing of its actuation to achieve phototaxis via entirely local, stochastic feedback with limited memory. Through this embodied computation approach, coordination emerges from the interaction of the body and the environment. We show robust phototaxis for soft robots that have different morphologies, that undergo damage, and that exhibit highly nonlinear leg behavior, all without an internal body representation. These results, therefore, offer a blueprint for designing resilient, autonomous soft robots that exploit the potential intelligence of their soft adaptive bodies.
Schomaker et al. (Thu,) studied this question.