Virtual humans: Evolving with common sense


While the quality and robustness of animation techniques for virtual human have improved greatly over the past couple of decades, techniques for improving their intelligence have not kept pace. Ideally, agents would be smart without being all-knowing and their future behaviors would be affected by their acquired knowledge just as with their real human counterparts. In this paper we present a method that uses commonsense knowledge to establish a baseline of concepts and relationships between objects. An agent then learns environment specific knowledge through its own perception and communication with other agents. Ultimately, agents' commonsense knowledge is then refined by their own experiences. © 2012 Springer-Verlag Berlin Heidelberg.

Publication Title

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)