This paper presents a swing control method using echo state networks (ESNs) for a lower extremity exoskeleton. In order to realize the leg swing motion, the swing controller is essential while wearing a lower extremity exoskeleton which is composed by electrical actuators. The swing control method employed to control a lower extremity exoskeleton must take into account the complex interaction between human and exoskeleton. To handle the interaction, most of relevant studies have employed dynamics-based control approaches that require the mathematical representation of the lower extremity exoskeleton. However, obtaining a useful inverse dynamics model of the lower extremity exoskeleton can be challenging due to its nonlinearities. On this account, the inverse learning technique based on ESNs can be utilized to replace the dynamics-based inverse model. The prototype of a lower extremity exoskeleton with power-augmentation purposes is designed and built for a military use. Experiments conducted on the prototype of a lower extremity exoskeleton show the promising results that operate well even in the absence of the dynamics-based model.
展开▼
