The foundation for all of Soar Technology’s projects is rooted in cognitive science. This primarily includes behavior, cognition, perception, memory, performance, learning, and emotion. Much of this scientific base is encoded within Soar, a computational cognitive architecture that enables the creation of autonomous software agents capable of sophisticated reasoning while utilizing large amounts of human-level knowledge. We use this basis to motivate and inform our work in training, decision aids, human-system interaction, and more fundamentally, how we develop autonomous systems. Our continuing research builds on this foundation; understanding how to integrate different cognitive architectures, how to better utilize external knowledge representations such as ontologies, and how to incorporate general learning algorithms into the Soar architecture.
We also conduct cognitive research using other cognitive and performance architectures, such as human performance and error modeling using with GOMS and the GLEAN architecture. Additionally, Soar Technology conducts research into how to better develop and engineer intelligent, autonomous systems for Soar and the broader agent community.
Adding Emotion to Knowledge-Rich Agents
We integrated a connectionist model of emotional processing with a synthetic force model. The intent of the project is to investigate improved realism in generating complex human-like behavior that is sensitive to emotional assessments related to fear, anger, joy, etc. The emotions model comes from a connectionist cognitive architecture called SESAME, which has been used to model a wide range of human behaviors, including navigation in large-scale spaces and the effects of emotions on learning. The synthetic force model is based on a special operations forces model that is implemented within the Soar architecture for cognition.
This project focuses on the interfaces required to integrate behavior moderators (represented by sub symbolic signal-processing systems) with higher cognitive processes (represented in a symbol-processing reasoning system). The interfaces include the transmission of emotional signals from the connectionist system to the symbolic system's perceptual and cognitive processes, plus high-level situation assessments and active concepts that must be communicated in an appropriate form to the behavior moderation system. We detail the interface solutions we have generated for our model, plus lessons learned about the more general problem of interfacing behavior moderators to the high-level reasoning systems in intelligent synthetic forces.
'
©2008 Soar Technology, Inc. All rights reserved
Legal Notice | Site Map
- Publications
- Scientists