Virtual humans dynamic control: robust balance and task management

Physical simulators do not possess animation tools efficient enough to let a virtual human keep its balance and realize tasks while interacting in a very general way with its environment. We propose a dynamical controller of virtual humans which offers a unified framework where all the constraints are treated simultaneously : unilateral forces of contact with friction, bilateral grasping efforts and motor joint torques. In order to assure the virtual human’s balance in a noncoplanar contact and grasp situation, the controller’s commands such as position of center of mass are determined statically. Our virtual human is stable and robust to disturbance. To increase its autonomy and richness of behaviour, each goal is divided into a sequence of elementary actions, translated into criteria and constraints for the control. In addition, the results of the simulation were validated thanks to a trial campaign realized on human subjects. Eventually, this control is applied to the case of a powered industrial truck in lateral fall, so as to obtain a predictive model of the forklift operator’s behaviour.