On realistic human motion simulation for virtual manipulation tasks

Repetitive motion on industrial workstations is at the origin of musculoskeletal disorders (MSD). There is a need for new simulation tools that allows for human motion prediction. This paper addresses the problem of realistic human motion simulation on industrial workstations. It is based on motion analysis and control synthesis on a kinematics basis. Literature and motion captures are compared. Then, a scheme based on optimization in Cartesian and motor space is proposed and tested on 2 significant examples. Singular Value theory is used for determining the shape of hand paths. Finally, an industrial application is simulated. Simulations are realized with a 23 Degrees Of Freedom (DOF) virtual human and are also of interest for humanoid robots.