Computational modeling of neuromusculoskeletal systems : from filaments to behavior

• Main Author: Yeo, Sang Hoon
• Language: English
• Published: University of British Columbia 2012
• Online Access: http://hdl.handle.net/2429/43756

This thesis describes computational approaches to modeling and simulating aspects of the neuromusculoskeletal system. We make contributions to models at three different levels of detail. We first investigate the mechanics of shortening muscle and evaluate two forms of the traditional Hill-type muscle model, force scaling and f-max scaling, and show that the f-max scaling model is significantly better at predicting experimental results. We hypothesize a new model called the winding filament model that incorporates the role of titin during active force development. Based on the proposed hypothesis, we develop a computational model that is able to simulate residual force enhancement. The suggested model can qualitatively simulate the pattern of the force enhancement observed in previous studies. In order to model the higher levels of the system consisting of muscles and bones, we propose an optimal design framework for estimating parameters of the musculoskeletal model. The method finds a set of morphological and physiological parameters that can optimally simulate the measured force and moment at the point of action. We apply the suggested framework to modeling two rat hindlimb muscles, gracilis posticus and posterior part of biceps femoris, to see if the traditional line segment based muscle geometry model is valid for musculoskeletal system modeling. The result shows that even a complex muscle like biceps femoris can be well modeled as a line segment, but its estimated insertion point is far from that of the traditional model based on anatomy. Finally, this thesis addresses a behavioral aspect of biological movement; in particular, how a high level movement is planned and controlled, in coordination with perception. We present a fully generative model of object interception that can simulate realistic, human-like behavior of ball catching for given arbitrary ball trajectory. The model includes a simplified probabilistic model of vision, a model of eye movements combining saccades and pursuit, and corresponding head, hand and body movements. The movements are constructed from submovements. By combining these components, realistic interception behavior is simulated with minimal user intervention.