Implementation of heuristic search in an expert database system

摘要

This paper shows how a relational database management system (DBMS) may be extended in order to solve a complex heuristic search problem DBMS-internally, i.e., without using a host language. To achieve this goal, the query language has been modified to include abstract data types and sophisticated control structures for efficient rule processing. Compared with conventional expert systems shells, this expert database approach seems particularly promising for problems that involve large amounts of data. The example application considered in this paper is the overland search problem, where one tries to find the best path between two points on a geographic map. In our system the map is represented by a set of disjoint adjacent polygons and stored in an INGRES relational database. The path search was implemented in a rule-based manner using an extension of the INGRES query language QUEL. In order to improve the performance of our system we also describe how to utilize geographic knowledge and geometric algorithms to perform a hierarchical decomposition of the given search problem.