Brief Description
The dynamic optimization approach is often used in the simulation of human movement to increase understanding of both normal walking and how patients with a loss of function can best provide compensatory movement patterns. To do an optimization, a simulated annealing algorithm searches to find the muscle activation parameters that will produce the walking pattern that most closely matches the pattern measured from the patient. One of the major limitations of the dynamic optimization approach is that it can be very computationally expensive; often taking days to obtain a solution on a single processor system. Higginson et al have recently developed a parallel version of a simulated annealing optimization (SPAN) for use in dynamic optimizations of human walking. The SPAN algorithm was found to scale linearly with the number of processors up to a limit of the number of steps in the algorithm’s search radius loop. The goal of this project is to develop computational methodologies as well as computational facilities to support research on dynamic optimization of muscle forces for clinical motion analysis.
This set of "walking-legs" is a computer-simulated skeleton and associated musculature. Although the bones and muscles were generated by computer, the data used to animate them was precisely recorded from a real patient using a “motion-capture system.” The simulation program also uses the data to optimize the parameters in modeling human walk. (Adapted from PDB website with permission)
Recent Accomplishments
HPCIO developed a Microsoft Windows-based dedicated computational cluster and a parallel biomechanics simulation system. This work involved adapting the parallelized simulated annealing optimization (SPAN) algorithm to Microsoft Windows platform and analyzing the current parallel decomposition to determine if other decompositions can improve load balance and scalability. Current and Future WorkHPCIO will provide various computational supports, including: the application performance profile; comparing performance of alternate implementations; and improving load balance and scalability.
Collaborators
- Thomas M. Kepple, M.A. (Physical Disabilities Branch, CC-NICHD, NIH )
- Steven J. Stanhope, Ph.D. ( Director, Physical Disabilities Branch, CC-NICHD, NIH)
Performance Metrics
Simulation system speedup: 15
