Snake-like robots have been widely studied and developed to exploit their flexible mobility and versatility. However, when encouter-ing powerful damages, how to recover the functionality is seldom investigated. This paper proposed a trial-and-error learning approach for damage recovery for 3-dimensional snake-like robots. The proposed method can guide snake-like robots to find compensation behavior in the absence of the pre-specified damage models. Our proposed method is evaluated by experiments in real world and various simulations.
展开▼
