Predictive Dynamics: is a term coined to characterize the prediction of human motion in a physics-based world. While many multi-body dynamics commercial code can generally integrate the equations of motion for a physical system, such methods are not applicable for predicting how humans move.  For example, if the digital human is to climb a ladder, the user specifies to the simulation system the initial and final configurations of Santos.  The motion in between is then "predicted" by the system.   If Santos falls, the user would know why he fell!  If Santos stumbles across an obstacle, the user will also know why!    In the mean time, a user can monitor all forces, torques, stress levels, and physiological parameters of Santos as he performs

Our method capitalizes on a novel optimization-based approach to motion prediction.  Using this new approach, motion is governed by human performance measures, such as speed and energy, which act as objective functions in an optimization formulation. In addition, constraints on joint torques and angles are imposed quite easily.  Predicting motion in this way allows one to use avatars to study how and why humans move the way they do, given a specific scenario.  It also enables avatars to react to infinitely many scenarios with substantial autonomy.  In addition, by using optimization, it is possible to predict dynamic motion without having to integrate equations of motion, which can be a cumbersome process.

The foundation for our work with the biomechanical digital model is Santos^TM, an advanced, newly developed virtual human at the University of Iowa.  Ultimately, Santos^TM will be capable of offering feedback and answering questions about a virtual prototype. 

Rather than solving the equations of motion, the Predictive Dynamics generalized method uses the mature field of optimization to solve for a continuous time-dependent curve characterizing joint variables (also called joint profiles) for every degree of freedom.

This is a very active area of research at VSR involving five graduate students.  We firmly believe that it will make a significant impact on how human motion is predicted because it takes into consideration human performance measure as objective functions to be minimized or maximized and many realistic constraints on the motion and various forces.

Areas of research under this effort include the following:- Balance and gait prediction - Task segmentation - Single chain motion prediction - Swinging motion - Climbing a ladder - Lifting

Selected references:

a. Kim, H-J., Horn, E., Arora, J.S. and Abdel-Malek, K., “An optimization-based methodology to predict digital human gait motion,” 2005-01-2710, Digital Human Modeling for Design and Engineering Symposium, Society for Automotive Engineering, Iowa City, IA, June 14-16, 2005.

b. Wang, Q., Xiang, Y-J., Kim, J-H., Arora, J.S. and Abdel-Malek, K., “Alternative formulations for optimization-based digital human motion prediction,” 2005-01-2691, Digital Human Modeling for Design and Engineering Symposium, Society for Automotive Engineering, Iowa City, IA, June 14-16, 2005.

c. Kim, J-H., Abdel-Malek, K., Yang, J., Farrell, K. and Nebel, K.J. “Optimization-based dynamic motion simulation and energy expenditure prediction for a digital human,” 2005-01-2717, Digital Human Modeling for Design and Engineering Symposium, Society for Automotive Engineering, Iowa City, IA, June 14-16, 2005.

d. Wang, Q. and Arora, J.S., “Alternate formulations for transient dynamic response optimization, AIAA Journal, 43 (10), 2202-2209, 2005.