Publisher's Synopsis
Recent development in video games, simulation, training, and robotics has seen a push for greater visual and behavioral realism. As the reliance on high delity models in the education, training, and simulation communities to provide information used for strategic and tactical decisions rises, the importance of accuracy and credibility of simulated behavior increases. Credibility is typically established through veri cation and validation techniques. Increased interest exists in bringing behavior realism to the same level as the visual. Thus far validation process for behavioral models is unclear. With real world behavior a major goal, this research investigates the validation problem and provides a process for quantifying behavioral correctness. We design a representation of behavior based on kinematic features capturable from persistent sensors and develop a domain independent classi cation framework for the measuring of behavior replication. We demonstrate functionality through correct behavior comparison and evaluation of sample simulated behaviors.
