Search Results

The purpose this paper is to delineate why there exists a trade-off between biomechanical realism and algorithmic efficiency for human motion simulation models, and to illustrate how empirical human movement data and findings can be integrated with novel modeling techniques to overcome such a realism-efficiency tradeoff. We first review three major classes of biomechanical models for human motion simulation. The review of these models is woven together by a common fundamental problem of redundancy—kinematic and/or muscle redundancy. We describe how this problem is resolved in each class of models, and unveil how the trade-off arises, that is, how the computational demand associated with solving the problem is amplified as a model evolves from small scale to large scale, or from less realism to more realism.

A prediction model for hand prehensile movements was developed and validated. The model is based on a new approach that blends forward dynamics and a simple parametric control scheme. In the development phase, model parameters were first estimated using a set of hand grasping movement data, and then statistically analyzed. In the validation phase, the model was applied to novel conditions created by varying the subject group and size of the object grasped. The model performance was evaluated by the prediction errors under various novel conditions as compared to the benchmark values with no extrapolation. Analyses of the model parameters led to insights into human movement production and control. The resulting model also offers computational simplicity and efficiency, a much desired attribute for digital applications.

A research effort was initiated to establish an empirical data base and to develop predictive models of normal human in-vehicle seated reaching motions while driving. A driving simulator was built, in which a variety of targets were positioned at typical locations a driver would possibly reach. Reaching motions towards these targets were performed by demographically representative subjects and measured by a state-of-the-art motion analysis system. This paper describes the experiment conducted to collect the movement data, and the new techniques that are being developed to process, analyze, and model the data. Some initial findings regarding the role of torso assistive motion, the effect of speed used in completing a motion on multi-segment dynamic postures, and illustrative results from kinematic modeling are presented.

For human motion studies, which are to be used for either dynamic biomechanical analyses or development of human motion simulation models, it is important to establish an empirical motion database derived from efficient measurement and well-standardized data processing methodologies. This paper describes the motion recording and data processing system developed for modeling seated reach motions at the University of Michigan's HUMOSIM Laboratory. Both electromagnetic (Flock of Birds) and optical (Qualysis) motion capture systems are being used simultaneously to record the motion data. Using both types of devices provides a robust means to record human motion, but each has different limitations and advantages. The amount of kinematic information (DOFs), external sources of noise, shadowing, off-line marker identification/tracking time, and setup cost are key differences.

Modeling normal human reach behavior is dependent on many factors. Anthropometry, age, gender, joint mobility and muscle strength are a few such factors related to the individual being modeled. Reach locations, seat configurations, and tool weights are a few other task factors that can affect dynamic reach postures. This paper describes how two different modeling approaches are being used in the University of Michigan Human Motion Simulation Laboratory to predict normal seated reaching motions. One type of model uses an inverse kinematic structure with an optimization procedure that minimizes the weighted sum of the instantaneous velocity of each body segment. The second model employs a new functional regression technique to fit polynomial equations to the angular displacements of each body segment. To develop and validate these models, 38 subjects of widely varying age and anthropometry were asked to perform reaching motions while seated in simulated vehicle or industrial workplace.