Closed form and geometric algorithms for real-time control of an avatar

Abstract

In a virtual environment with multiple participants, it is necessary that the user's actions be replicated by synthetic human forms. Whole-body digitizers would be the most realistic solution for capturing the individual participant's human form, however the best of the digitizers available are not interactive and are therefore not suitable for real-time interaction. Usually, a limited number of sensors are used as constraints on the synthetic human form. Inverse kinematics algorithms are applied to satisfy these sensor constraints. These algorithms result in slower interaction because of their iterative nature, especially when there are a large number of participants. To support real-time interaction in a virtual environment, there is a need to generate closed-form solutions and fast searching algorithms. In this paper, a new closed-form solution for the arms (and legs) is developed using two magnetic sensors. In developing this solution, we use the biomechanical relationship between the lower arm and the upper arm to provide an analytical, non-iterative solution. We have also outlined a solution for the whole human body by using up to ten magnetic sensors to break the human skeleton into smaller kinematic chains. In developing our algorithms, we use the knowledge of natural body postures to generate faster solutions for real-time interaction.