Vision performance measures for optimization-based posture prediction

Although much work has been completed with modeling head-neck movements as well with studying the intricacies of vision and eye movements, relatively little research has been conducted involving how vision affects human upper-body posture. By leveraging direct human optimized posture prediction (D-HOPP), we are able to predict postures that incorporate one's tendency to actually look towards a workspace or see a target. D-HOPP is an optimization-based approach that functions in real time with Santos^™, a new kind of virtual human with a high number of degrees-of-freedom and a highly realistic appearance. With this approach, human performance measures provide objective functions in an optimization problem that is solved just once for a given posture or task. We have developed two new performance measures: visual acuity and visual displacement. Although the visual-acuity performance measure is based on well-accepted published concepts, we find that it has little effect on the predicted posture when a target point is outside one's field of view. Consequently, we have developed visual displacement, which corrects this problem. In general, we find that vision alone does not govern posture. However, using multi-objective optimization, we combine visual acuity and visual displacement with other performance measures, to yield realistic and validated predicted human postures that incorporate vision.