Towards the autonomous animation of multiple human figures

High level tools to support the animation of multiple human figures make use of knowledge in a number of ways. Explicit knowledge, in the form of keyframes is supplied directly by the animator and procedural knowledge for repetitive movements like walking or grasping is built into the algorithms. However, the interaction of multiple figures in a complex environment requires a declarative knowledge base of rules and constraints. The most obvious way to add declarative knowledge to an animation system is to choose a well developed expert system and to set up communication channels between the two systems, but this “two monoliths” approach can be very inefficient. To avoid the problems associated with distinct expert and animation systems, we are implementing a blackboard architecture which allows integration of reasoning with the graphics algorithms. The result is a mixed initiative system where autonomously produced motion paths for multiple human figures are edited and constrained interactively by the animator. A partial implementation is being evaluated