Virtual humans and humanoid robots

**Figure 1.8:** Across the top, a motion computed by a planning algorithm, for a digital actor to reach into a refrigerator [498]. In the lower left, a digital actor plays chess with a virtual robot [544]. In the lower right, a planning algorithm computes the motions of 100 digital actors moving across terrain with obstacles [591].
$\begin{figure}\begin{center} \psfig{file=figs/kallmann.ps,width=\columnwidth} \\... ...g{file=figs/100chars.ps,height=2.0in,width=2.85in} \\ \end{center} \end{figure}$

Beyond video games, there is broader interest in developing virtual humans. See Figure 1.8. In the field of computer graphics, computer-generated animations are a primary focus. Animators would like to develop digital actors that maintain many elusive style characteristics of human actors while at the same time being able to design motions for them from high-level descriptions. It is extremely tedious and time consuming to specify all motions frame-by-frame. The development of planning algorithms in this context is rapidly expanding.

**Figure 1.9:** (a) This is a picture of the H7 humanoid robot and one of its developers, S. Kagami. It was developed in the JSK Laboratory at the University of Tokyo. (b) Bringing virtual reality and physical reality together. A planning algorithm computes stable motions for a humanoid to grab an obstructed object on the floor [561].
$\begin{figure}\begin{center} \begin{tabular}{cc} \psfig{file=figs/h7.ps,width=1.... ...igs/utokyo.ps,width=3.88in} \\ (a) & (b) \end{tabular}\end{center} \end{figure}$

**Figure 1.10:** Humanoid robots from the Japanese automotive industry: (a) The latest Asimo robot from Honda can run at 3 km/hr (courtesy of Honda); (b) planning is incorporated with vision in the Toyota humanoid so that it plans to grasp objects [448].
$\begin{figure}\begin{center} \begin{tabular}{cc} \psfig{file=figs/asimorun.ps,wi... ...igs/toyota.ps,width=1.95in} \\ (a) & (b) \end{tabular}\end{center} \end{figure}$

Why stop at virtual humans? The Japanese robotics community has inspired the world with its development of advanced humanoid robots. In 1997, Honda shocked the world by unveiling an impressive humanoid that could walk up stairs and recover from lost balance. Since that time, numerous corporations and institutions have improved humanoid designs. Although most of the mechanical issues have been worked out, two principle difficulties that remain are sensing and planning. What good is a humanoid robot if it cannot be programmed to accept high-level commands and execute them autonomously? Figure 1.9 shows work from the University of Tokyo for which a plan computed in simulation for a humanoid robot is actually applied on a real humanoid. Figure 1.10 shows humanoid projects from the Japanese automotive industry.