An architecture for the Behavioural Animation of Virtual Humans

Despite recent powerful tools aimed at creating virtual humans, the animation of such characters is still a major bottleneck in production, and there is a need to automate this task, especially when the number of involved agents grows. This work focuses on creating autonomous agents, capable of taking decisions based on their simulated perceptions and internal motivations. Since complete autonomy is still difficult to reach (and especially for such complex creatures as virtual humans), we favour a plan-based approach (which relies on some predefined behaviours), rather than Artificial Life techniques, in order to keep the results more or less “predictable”. In Virtual Storytelling, autonomy helps the agent reacting to unexpected events, but it more or less follows a predefined plot. This research discusses other related animation problems, such as creating, animating and coordinating movements, and especially how motions can be adapted through Motion Retargeting.