Model Predictive Control for Human Motion Simulation

This paper describes a novel model-based controller designed to simulate human motion in dynamic virtual environments. The controller was tested on Santos^TM, the digital human developed at the Virtual Soldier Research Program at the University of Iowa.

A planar 3-degrees-of-freedom model of the human arm was used to test the hypothesis. The controller was used to predict on line, optimal torques required to move the end effector towards a target point. The control law was implemented using classical gradient-based optimization and the recently developed technique of model predictive control (MPC). An advantage of MPC is that it replaces intractable closed loop optimization problems with more easily implementable open loop problems.

The controller was used to produce physically consistent simulations of the motion of a human arm in a virtual environment in the presence of external disturbances that were not known in advance.