Neural Control and Learning of a Gecko-Inspired Robot for Aerial Self-righting

Animals like geckos are able to reorient themselves while falling from a height and land safely, utilizing their flexible spines and tails for aerial self-righting. Researchers have studied this behavior to develop robots that can mimic such an animal locomotor skill. While quadruped robots have achieved impressive self-righting results, they still lack the utilization of a combination of a body, a tail, and legs as found in a sprawling posture animal like a gecko. In this study, three versions of gecko-inspired robots are developed, allowing investigation into how the spine (body and tail) and legs contribute to aerial self-righting behavior. A neural central pattern generator (CPG) with a radial basis function (RBF)-based premotor neuron network is used for the robots to learn self-righting behavior, along with continuous locomotion from falling to walking with a smooth transition. The results show that the robot with leg control but without active body control is limited to a 90-degree falling angle, while the robot with both leg and body control can achieve 110 degrees, and the one with full leg, body, and tail control can recover from a falling angle of 150 degrees to land successfully. Finally, we demonstrate the modular capability of the controller by enabling the robot to fall and walk simultaneously.