Development of a posture prediction model

Graduation date: 2002Biomechanical models have been used in designing human work\ud environments to evaluate potential risks to workers before a work environment is\ud constructed. In order for work environments to be modeled correctly, most\ud biomechanical models require as input, an accurate body posture of the worker.\ud This information can be obtained by, either measuring the posture of workers for\ud the task of interest, or estimating the posture.\ud This research explores methods to estimate working postures by developing\ud a model that can predict a worker's posture. The model in this thesis represents the\ud body of the worker with ten links: neck, left and right forearms, left and right upper\ud arms, body, left and right thighs, and left and right calves. The work task inputs\ud consist of the magnitude and direction of the force applied to the hands, and the\ud distances between the hands and the floor. By using these inputs, the model can\ud predict a posture by optimizing an objective function of two criteria: Total Squared\ud Moment and Balance. Model constraints also ensure that a predicted posture is\ud feasible for human.\ud The output of the model is the predicted posture in terms of ten body joint\ud angles: neck, left and right elbows, left and right shoulders, hip, left and right\ud knees, left and right ankles. These joint angles are defined as angles relative to\ud horizontal.\ud The prediction posture can be used as a base reference when inputting\ud into other biomechanical models. By predicting posture from the model, one can\ud obtain postures of the workers without direct measurement of postures from the\ud workers, which can be expensive and time consuming