Today, there exist several methods to compute a realistic human posture depending on tasks and geometries, with which a human being is interacting. Several approaches use probabilistic models in combination with inverse kinematics. These methods produce acceptable results concerning the reliability and accuracy. However, they are unsatisfying in terms of flexibility concerning application to various tasks, since adapting a probabilistic model to new requirements needs a high technical measurement effort. Another disadvantage is that it is challenging to derive values like strain and workload respectively or discomfort from these approaches.
Hence we have developed an approach for 1) computing autonomous postures, 2) assessing and considering joint load and 3) assessing discomfort. Integral part of this approach is an accurate physical description of test subjects. The physical characterization includes anthropometry, masses of body parts, centre of gravity, angle dependent resistant torques and maximum torques for each joint of the body. Our approach assumes that a human being wants to minimize the joint strain when taking a desired posture. This approach has been formulated within a Mathematica and RAMSIS simulation and exemplary evaluated for different tasks.
Our investigation shows, that we are able to compute feasible postures by minimizing joint load for various tasks. Furthermore, we can predict the individual load and discomfort based on a found correlation between load and discomfort assessment.