Biped humanoid navigation system supervised through interruptible user-interface with asynchronous vision and foot sensor monitoring

Abstract

Biped robots are expected to explore in the unknown environment which has undulate floors and a lot of obstacles on the floors. Such kind of the environment are difficult for wheeled mobile robots. In addition to planning and recognition functionality, interruption from operators to the systems are important to change behavior of navigation system. In order to achieve the tasks safely, as the system detects the unexpected situation, operators need to direct suitable goal and parameters to the planner. In this paper, we show a navigation system capable of interruption by operators. The communication between the system and operators are not continuous instruction but interruption. The keys of the system are: 1) system integration across different machines according to supervised autonomy framework. 2) fast vision based environment modeling and 3-D footstep planning to support operator to direct goals. 3) interruptible scheduling of footstep execution to cancel walking according to monitoring environment and operator's instructions. We implement the system on a real robot, HRP-2, and show effectiveness of interruptible system integration and 3-D footstep planning through experiments.