Self-organized Locomotion with Multiple Stepping Frequencies in an Insect-Like Robot Under Decentralized Adaptive Neural Control

While walking robots typically employ a homogeneous stepping frequency across all legs to establish interlimb coordination, certain insects, such as locusts, interestingly exhibit adaptive interlimb coordination by employing heterogeneous stepping frequencies for different leg pairs. This adaptation arises from the distinct structures and lengths of their front, middle, and hind legs. Inspired by this biological phenomenon, this paper proposes decentralized adaptive neural control with multiple central pattern generators capable of realizing the adaptive interlimb coordination observed in insect walking. This neural control system can automatically generate gaits from heterogeneous stepping frequencies without the need for predefined interlimb coordination. Our preliminary results demonstrate that this neural control approach enables an insect-like robot, equipped with distinctly heterogeneous leg lengths, to achieve stable gaits autonomously and rapidly under various combinations of leg-stepping frequencies.