ABSTRACT
A substantial advantage of predictive virtual human models is the ability to adapt to changes in a virtual environment automatically, and with respect to posture prediction and analysis, this ability hinges on collision avoidance. Collision avoidance must be robust enough to accommodate various types of geometry, must apply to the avatar (self-avoidance) as well as virtual objects, must not detract from real-time operation, and must be suitable for a variety of real-world scenarios. Thus, while leveraging optimization-based posture prediction and a unique method for collision avoidance with increase computational speed we present new developments in this arena. A new sphere-filling algorithm is presented with increased speed and fidelity for creating surrogate geometry, which is critical for any type of collision avoidance or detection. The collision avoidance algorithm is implemented for self-avoidance. And, the new capabilities are demonstrated on automotive and motorcycle examples for ergonomic analysis. The results not only involve realistic predicted postures and novel forms of human-performance feedback, but also reflect real-time operation.
