Automated real-time animation

Abstract

Producing animation for the computer game/entertainment industry is a time- consuming, difficult process. One approach to maximizing the utility of previously animated sequences is to blend and warp pre-existing sequences to produce new animations on the fly. The extent to which these approaches are feasible in realtime and are doable without any animator input is debatable. This research considers many of the approaches to blending and warping in the literature. Many of these require manual intervention to tune the transition parameters and are also computationally expensive. However, they serve as a foundation for developing an approach to the problem that is completely automatic and is computationally inexpensive. Predefined motions of walking, running, and jumping derived from motion capture data are considered and a novel approach to producing segues between these motions is implemented. Following the lead of research suggesting that transition times are of vital importance in the response of users to transition animations the transitions produced by the implementation were tested on a group of survey participants and the same general conclusion was supported. To conclude the work, suggestions for improvement of the algorithms for transition production are made and possible future developments are considered.