Designing and Implementing a Bridge Playing Program

Yossi Nygate
AT&T - Bell Laboratories 6200 E. Broad St. - Rm. 2B253 Columbus, OH 43213, U.S.A.
yossi@hercules.cb.att.com

Leon Sterling
Department of Computer Science, Case Western Reserve University, Cleveland, OH., 44106, U.S.A.
leon@ces.cwru.edu

Abstract

This paper describes BASSINET, the first proficient system for one of the major areas in Bridge that have not yet been successfully solved - declarer play in no trumps. BASSINET closes the gap between the poor performance of existing Bridge programs and other successful game playing programs.

The main reason for this disparity is that Bridge cannot be solved using a standard knowledge-based problem solving technique such as heuristic search, production rules, or means-end analysis.

The success of BASSINET is primarily due to its use of ASPEN, a structured approach for integrating multiple problem solving techniques in complex, knowledge based systems. ASPEN proposes that programs should be decomposed into five consecutive, well integrated stages: Abstraction, Synthesis, Planning, Execution and moNitoring. Each stage may use the knowledge representation and problem solving technique that is best suited to the problem domain. The interaction between the stages is through knowledge representation transformations and feedback.

This paper presents BASSINET and its use of ASPEN.

Conference Home Page