Validation of predicted posture for the virtual human Santos™

Digital human modeling and simulation plays an important role in product design, prototyping, and manufacturing: it reduces the number of design iterations and increases the safety and design quality of products. Posture prediction is one of the key capabilities. It is especially useful in the design of vehicle interiors for checking the reachability of buttons and determining comfort levels. This paper presents the validation of predicted posture for the virtual human Santos. The predicted posture is a physics-based model and is formulated as a multi-objective optimization (MOO) problem. The hypothesis is that human performance measures (cost functions) govern how humans move. We chose 12 subjects from four different percentiles, all Americans (female 5%, female 50%, male 50%, and male 95%). Four realistic in-vehicle tasks requiring both simple and complex functionality of the human simulations were chosen: reaching a point at the top of the A-pillar, the radio tuner button, the glove box handle, and a point on the driver's B-pillar seatbelt adjuster. The subjects were asked to reach the four target points, and the joint centers for wrist, elbow, and shoulder and the joint angle of elbow were recorded using a motion capture system. We used these data to validate our model. The validation criteria comprise R-square and confidence intervals. The results show that the predicted postures match well with the experiment results, and are realistic postures.