Locomotion transition scheme with instability evaluation using Bayesian Network

The applicative field of activities of robots which have only one locomotion strategy is limited. As a mean of enhancing the mobile range, it is necessary to have various locomotion modes. Therefore, we focus on dynamic transitions between several kinds of locomotion modes adapting to environmental changes. In this paper, we aim to realize a stable locomotion along some unknown test courses with transition between biped and quadruped walks. To achive this transition, we propose a method to get environmental information and internal conditions. Robot plans locomotion based on recognition of test courses and estimate stability of walking using Bayesian Network. The effectiveness of proposed method is verified by experiments.