Search-based Planning for Character Animation

Planning is the most generic AI technique to generate intelligent behaviour for virtual actors, in computer animation or computer games. In the animation domain, planning capabilities will consist in finding a right sequence of actions that let an agent achieve pre-defined goals. Each action generated can be played in the environment to produce an animation. Hence entire animations can be generated from first principles, by defining a set of actions and allocating high-level goals to the character. This makes possible not only to generate intelligent behaviour, but also to explore the diversity of courses of action. In recent years, several researchers have described the use of planning systems to control characters’ behaviours. Geib [5] has proposed the use of refinement planning following a detailed study of animation requirements [5][8]. Funge has used situation calculus to generate intelligent behaviours for virtual actors [4], and Cavazza et al. [2][3] have approached this problem with Hierarchical Task Networks (HTNs) for storytelling, considering the knowledge intensive nature of this kind of applications.