Real-Time Motion Synthesis for Multiple Goal-Directed Tasks Using Motion Layers

Abstract

In this paper, a work in progress approach of layered motion interpolation method for designing realistic animation sequences for multiple goal-directed tasks is presented. The proposed solution is based on the ability to extract and synthesize different motions (in layers), while trying to efficiently reconstruct a natural-looking character's posture in real time. The proposed solution is examined for the case in which running, jumping, and reaching motions are combined. However, in addressing multiple goals fulfilled by a character in complex environments, as well as those involving complex motions, it is necessary to define the best way to handle and reconstruct the information from a motion capture database. Finally, because the character's posture should be as natural looking as possible, a simple centre of mass approach is proposed, to give desirable results at specific time steps.