Other Academic Research Groups Using Soar

University of Michigan: Our research focuses on extensions to the Soar architecture and the cognitive capabilities made possible by Soar. Recent architectural research includes work on reinforcement learning, episodic memory, semantic memory, mental imagery, and motor control. Our research on cognitive capabilities includes work on learning by situated interactive instruction, learning relational and continuous action models, and robot control. We work on a variety of domains that include video games, linguistics tasks, mobile robots, and planning tasks.

Penn State: People in our lab help move the field of cognitive science. Our projects are focused on models that learn, ranging from how to provide models access to interfaces to analyzing the effects of caffeine on cognition to determining how children develop through the modeling of their development. Other projects include the development of the Psychological Soar Tutorial and the Soar FAQ. For more information contact frank.ritter at psu.edu.

University of Portland: Research is in improving the effectiveness of the episodic memory system. Much of the work is done external to Soar while honoring architectural requirements of Soar. Recent projects include: a survey of forgetting mechanisms, using memories to build plans, using sequences of episodes to overcome state aliasing and using experience to discover the relative importance of various WMEs that comprise an agent's episodes. Many of the insights from this research have been embodied in an advanced episodic memory system called Ziggurat. Throughout this process, the hardest part of creating an effective episodic memory system remains the same: creating an effective, domain-independent algorithm to select the best memory from a given cue.

Pace University: An Intelligent Soar Assistant for a Virtual World Abe Guerra has built an assistant for IBM's virtual world. His agent uses Soar to reason about what a user needs and guides the user through the virtual world. The assistant interacts with the user through typed natural language commands, and can also assist with commands for using the virtual world interface. Here is an introductory paper on his project. There is also a movie of his agent navigating and a movie of his agent assisting a customer. A Soar Agent for Playing Poker Bob Follek has implemented an architecture for an autonomous poker player. His SoarBot uses Soar to play poker with the poker server set up by the University of Alberta Computer Poker Research Group. Bob can be reached at bob@codeblitz.com Researchers use Soar to build an intelligent assistant for IBM's virtual world. Agent reasons about what a user needs and guides the user through the virtual world. The assistant interacts with the user through typed natural language commands, and can also assist with commands for using the virtual world interface. The research group also uses Soar to implement an architecture for an autonomous poker player. The agent uses Soar to play poker with the poker server set up by the University of Alberta Computer Poker Research Group.

University of Southern California and the Institute for Creative Technologies

Bar Ilan University, Israel is home to the MAVERICK group, which conducts research in social intelligence, multi-agent and multi-robot systems, and plan recognition. Work in Soar has focused on developing Soar-based teamwork capabilities for modeling para-military units, multi-agent plan recognition and mirroring, and social comparison for groups.

Prof. Dr. Claudia Meitinger's team at the Augsburg Technical University of Applied Sciences is utilizing Soar to enhance various aspects of human-machine interaction and problem-solving. In the B2X Beyond SalesBot project, the aim is to develop, validate, and generalize an AI-based tool for technical sales, using deburring tools as a case study. A Soar agent is used for the tool configuration process, determining the tool parameters based on the customer requirements, which are enquired step-by-step. Additionally, the research group on Human-Machine Interaction with Intelligent Systems focuses on developing a demonstrator for a use case where a collaborative grinding robot is delegated by shop floor workers. Soar is employed to facilitate human-machine interaction by providing an abstract communication level, contributing to knowledge base expansion and task inference based on delegated tasks.