Jonathan Dinerstein, Parris K. Egbert, Dan Ventura, Michael Goodrich
We present a novel technique for behavioral animation through data-driven behavior synthesis that provides realistic and natural character behavior and has a programming-by-demonstration interface. The demonstrator's high-level behavior is recorded as a sequence of discrete actions. The virtual character synthesizes novel behavior by concatenating segments of action sequences, guided by simulations that predict fitness (thus producing a deliberative cognitive behavioral model). We empirically show that our O(log n) technique is scalable, robust, general, produces effective behavior, and is faster than a traditional cognitive model.
Subjects: 7.2 Software Agents; 6.1 Life-Like Characters
Submitted: Apr 8, 2008