CHS-Soar: Introducing constrained heuristic search to the soar cognitive architecture

Abstract

Cognitive architectures aspire for generality both in terms of problem solving and learning across a range of problems, yet to date few examples of domain independent learning has been demonstrated. In contrast, constraint programming often utilizes the same domain independent heuristics to find efficient solutions across a broad range of problems types. This paper provides a progress report on how a specific form of constraint-based reasoning, namely Constrained Heuristic Search (CHS) can be effectively introduced into an integrated symbolic cognitive architecture (Soar) to achieve domain independent learning. The integration of CHS into Soar retains the underlying problem-solving generality of Soar, yet exploits the generalized problem representation and solving techniques associated with constraint programming. Preliminary experiments are conducted on two problems types: Map Colouring and Job Shop Scheduling, both of which are used to demonstrate a domain independent learning using texture based measures. Copyright © 2008, Association for the Advancement of Artificial Intelligence (www.aoai.org). All rights reserved.