Navigation of soft slithering bodies on 3D heterogeneous terrains through bio-inspired sensing and learning

We present a control framework for soft slithering bodies, integrating dynamic snake simulations, neuron-inspired sensory feedback, and reinforcement learning. The utility of this framework is demonstrated by deriving simple yet effective control policies for snakes traversing homogeneous terrains. These learned policies are then coordinated through a higher-level decision-making mechanism, facilitating adaptive gait modulation across various frictional environments. We further demonstrate the robustness of this approach by deploying the snakes on realistic 3D terrains, showcasing successful navigation despite environmental complexities.