Neural control and learning with a turning strategy of a gecko-inspired robot with a bendable body

Sprawling animals leverage flexible trajectories during turning to navigate challenging and narrow environments effectively. Motivated by this natural behavior, we employed a gecko-inspired robot with a bendable body to investigate and implement an efficient turning strategy. A neural central pattern generator (CPG) with a radial basis function (RBF)-based premotor neuron network (CPG-RBF network) with black-box optimization (BBO) was used to control and train the robot for walking and consequently exploring the turning strategies based on the established walking policy to facilitate turning without the need for additional learning time. Through simulated robot experiments in circular and track environments, we demonstrated that the robot with an undulated C-shaped body strategy achieved a lower radius of the turning curve and consequently gained faster locomotion speed by approximately 23% compared to a fixed C-shaped body strategy. This enhanced performance is attributed to the undulated body’s effectiveness in facilitating turning maneuvers.