The work presented in this thesis focuses on the design and
implementation of parallel algorithms for problem solving tasks principally
in Rule-based Expert Systems and Artificial Intelligence (AI).
Rule-based Expert Systems are widely used in AI. Their use covers a
wide variety of application areas. However, in most cases, these systems
are computation intensive and run slowly. This increases the need for high performance
and real-time response.
Because of the convergence of parallelism in computer design and the
wide spread use of expert system in industry, the design of Parallel Expert
System has become of increasing importance. Parallel computation may
prove useful in shortening the processing time of the expert systems.
Expert systems are being designed for both distributed (loosely-coupled)
and shared-memory (tightly-coupled) multiprocessor machines. The work
presented here is an attempt to focus on the issues involved in designing a
rule-based expert system for a shared memory "multiprocessor system (the
Sequent Balance 8000).
Eight parallel Forward Chaining models and two parallel Backward
Chaining models are implemented. These models are presented in Chapter 5
and 6, together with a study of their efficiency.
A Doctoral Thesis. Submitted in partial fulfilment of the requirements for the award of Doctor of Philosophy of Loughborough University.