The origins of the Soar architecture can be traced back to the seminal research of Allen Newell and Herbert Simon on symbol systems, heuristic search, goals, problem spaces, and production systems. Since its official inception in 1982, Soar has evolved through six major releases, as both an AI architecture and as the basis for a unified theory of cognition. This paper traces this evolutionary path, starting with Soar's intellectual roots, and then proceeding through the stages defined by the six major system releases. Each stage is characterized with respect to a hierarchy of four levels of analysis: the knowledge level, the problem space level, the symbolic architecture level, and the implementation level.

AI is moving away from "toy tasks" such as block stacking towards real-world problems. This trend is positive, but the amount of preliminary groundwork required to tackle a real-world task can be staggering, particularly when developing an integrated agent architecture. To address this problem, we advicate real-world software environments, such as operating systems or databases, as domains for agent research. The cost, effort, and expertise required to develop and systematically experiment with software agents is relatively low. Furthermore, software environments circumvent many thorny, but peripheral, research issues that are inescapable in other environments. Thus, software environments enable us to test agents ina real world yet focus on core AI research issues. To support this claim, we describe our project to develop UNIX softbots (software robots) -- intelligent agnets that interact with UNIX. Existing softbots accept a diverse set of high-level goals, generate and execute plans to achieve these goals in real time, and recover from errors when necessary.

The query reformulation approach (also called semantic query optimization) takes advantage of the semantic knowledge about the contents of databases for optimization. The basic idea is to use the knowledge to reformulate a query into a less expensive yet equivalent query. Previous work on semantic query optimization has shown the cost reduction that can be achieved by reformulation, we further point out that when applied to distributed multidatabase queries, the reformulation approach can reduce the cost of moving intermediate data from one site to another. However, a robust and efficient method to discover the required knowledge has not yet been developed. This paper presents an example-guided, data-driven learning approach to acquire the knowledge needed in reformulation. We use example queries to guide the learning to capture the database usage pattern. In contrast to the heuristic-driven approach proposed by Siegel, the data-driven approach is more likely to learn the re...

In systems which learn a large number of rules (productions), it is important to match the rules efficiently, in order to avoid the machine learning utility problem --- if the learned rules slow down the matcher, the "learning" can slow the whole system down to a crawl. So we need match algorithms that scale well with the number of productions in the system. (Doorenbos, 1993) introduced right unlinking as a way to improve the scalability of the Rete match algorithm. In this paper we build on this idea, introducing a symmetric optimization, left unlinking, and demonstrating that it makes Rete scale well on an even larger class of systems. Unfortunately, when left and right unlinking are combined in the same system, they can interfere with each other. We give a particular way to combine them which we prove minimizes this interference, and analyze the worst-case remaining interference. Finally, we present empirical results showing that the interference is very small in practice, and that...

Efficient knowledge representation and reasoning is an important component of intelligent activity, and is a crucial aspect in the design of large-scale intelligent systems. This dissertation explores the design, analysis, and implementation of massively parallel knowledge representation and reasoning systems which can encode very large knowledge bases and respond to a class of queries in real-time, with reasoning episodes expected to span a fraction of a second. The dissertation attempts to design efficient, large-scale knowledge base systems by: (i) exploiting massive parallelism; and (ii) constraining representational and inferential capabilities to achieve tractability, while still retaining sufficient expressive power to capture a broad class of reasoning in intelligent systems. To this end, shruti, a connectionist reasoning system which models reflexive--- i.e., effortless and spontaneous---reasoning serves as the knowledge representation and reasoning framework. Shruti-based mas...

Many learning systems must confront the problem of run time after learning being greater than run time before learning. This utility problem has been a particular focus of research in explanation-based learning. In past work we have examined an approach to the utility problem that is based on restricting the expressiveness of the rule language so as to guarantee polynomial bounds on the cost of using learned rules. In this article we propose a new approach that limits the cost of learned rules without guaranteeing an a priori bound on the match process or restricting the expressibility of rule conditions. By making the learning mechanism sensitive to the control knowledge utilized during the problem solving that led to the creation of the new rule --- i.e., by incorporating such control knowledge into the explanation --- the cost of using the learned rule becomes bounded by the cost of the problem solving from which it was learned. 1 Introduction The identification of the utility prob...

What is the match cost of adding a new rule to a production system (rule-based system)? Two conflicting views have emerged. Research in EBL indicates that learned rules add to the match cost of a production system. Thus, as the production system size increases with learning, the match cost will also increase. There is much data in the literature to support this phenomenon. On the contrary, researchers in parallel production systems have concluded that the match effort in a production system is limited, independent of the size of the production system. Thus, an increase in the size of the production system will not lead to an increase in the match cost. There is much data to support this phenomenon as well. In this paper, we point out these contradictory views of production match in the two research communities. A direct analysis of these conflicting views is difficult, since the two communities have worked with vastly different systems. Therefore, we have developed some large production systems in Soar, to analyze the situation within a common framework. This common framework narrows down the possible causes for this conflict, and raises important questions for future work.

xi 1 Introduction 1 1.1 Semantic Query Optimization : : : : : : : : : : : : : : : : : : : : : : 3 1.2 High Utility Semantic Knowledge for SQO : : : : : : : : : : : : : : : 6 1.3 Learning Effective and Robust Rules : : : : : : : : : : : : : : : : : : 8 1.4 Closely Related Work : : : : : : : : : : : : : : : : : : : : : : : : : : : 10 1.5 Summary of Contributions : : : : : : : : : : : : : : : : : : : : : : : : 12 1.6 Organization of the Dissertation : : : : : : : : : : : : : : : : : : : : : 13 2 Robustness of Knowledge 15 2.1 Consistency of Rules and Database Changes : : : : : : : : : : : : : : 15 2.2 Definitions of Robustness : : : : : : : : : : : : : : : : : : : : : : : : : 18 2.3 Estimating Robustness : : : : : : : : : : : : : : : : : : : : : : : : : : 19 2.4 Templates for Estimating Robustness : : : : : : : : : : : : : : : : : : 26 2.5 Empirical Demonstration : : : : : : : : : : : : : : : : : : : : : : : : : 27 2.6 Related Uncertainty Measures : : : : : : : : : : : : : : : : : : : ...

Many learning systems must confront the problem of run time after learning being greater than run time before learning. This utility problem has been a particular focus of research in explanation-based learning. This research focuses on the expensive chunk problem in which individual learned rules are so expensive to match that the system suffers a slow down from learning. In past work,there has been an approach that is based on restricting the expressiveness of the rule language. Although this restriction allowed polynomial bounds on the cost of using learned rules, it brought about several negative side effects. This proposal presents a new approach, called controlled chunking, that limits the cost of learned rules without guaranteeing an a priori bound on the match process or restricting the expressibility of rule conditions. By making the learning mechanism sensitive to the control knowledge utilized during the problem solving that led to the creation of the new rule --- i.e., by i...

this article, we propose an alternative diagnosis for the cause of expensivechunks, along with a new approach for eliminating expensivechunks that is derived from this new diagnosis. The core idea is to focus on the relationship between the problem-space search upon which the learning is based and the search performed, during match, by the rule learned from this problem-space search. In the search of the problem space, some path --- that is, some sequence of operators --- is followed that eventually leads to a result. The actual path followed usually depends on metalevel control rules that determine which operators are selected for which states. These control rules should a#ect only the e#ciency with which the result is found, and not its correctness. As a result, when a new rule is acquired from a trace of this problem solving, the control rules are not included as part of the explanation of the result. This omission, which turns out to also be the approach taken in PRODIGY #Minton 1993#