Extended grasping behavior for autonomous human agents

Abstract In this paper we present an efficient technique allowing distantgrasp behaviors while preserving the static balance. This featuregreatly increases the autonomy of virtual humans. 1. Introduction Our goal is to provide an integrated reaching and graspingbehaviors for a virtual human agent. In most cases, grasping aclose object only involves the motion of one or both arms oftencomplemented with the independent motion of the head to looktowards the goal [RG91][MT94]. However the reach of distantobjects is a much more complex task as human beings can extendtheir field of action by various means. Let us imagine the processof grasping a relatively low object (Fig. 1a). To avoid a loss ofbalance the body posture is constrained so as to keep the verticalprojection of the center of mass inside the base of support (Fig.1b). However the object can be even more distant and additionalsupporting area has to be used.(Fig. 1c). In the case of Fig. 1c wecan reach the target with the left hand while leaning on the boxwith the right hand (Fig. 1d). The present approach solves thisclass of problems on a human body model with 95 degrees offreedom including the hands mobilbities [MT94].a b cdFigure 1. Reaching a close object (a,b)and a distant target (c,d)