In this work we present an algorithm to automatically generate eye-believable animations for virtual human-like characters evolving in cluttered environments. The algorithm has two components: (1) a motion planner for a simplified model of the character that produces admissible paths for this model and (2) a motion generator based on captured sequences to transform the planned path into a trajectory that can be followed by the complete model of the character. This second component uses a motion capture database and a classifier to choose which motions in the database are the most appropriate to follow the computed path. We give examples of successful plans generated with our method.
展开▼