State Classification for Humanoid Robots

In this paper, we decouple the motion-planning problem of humanoid robots into two sub-problems, namely topological state planning and detailed motion planning. The state classification plays a key role for the first sub-problem. We propose several basic states, including lying, sitting, standing and handstanding, abstracted from the daily exercises of human beings. Each basic state is classified further from the topological point of view. Furthermore, generalised function (GF) set theory is applied with the aim of analysing the kinematic characteristics of the end effectors for each state, and meaningful names are assigned for each state. Finally a topological state-planning example is given to show the effectiveness of this methodology. The results show that the large amounts of states can be described using assigned names, which leads to systematic and universal description of the states for humanoid robots.