A Hardware Implementation of an Expert System Shell Based on Belief Revision Concepts

We propose an Expert System Shell based on belief revision concepts who maintains the consistency of the knowledge base. In the first phase of this expert system shell we translate a classical rule-based system in an equivalent network representation where nodes are facts, and links stand for relationships. In the second phase we propagate any change in the uncertainty measures throughout the whole network. This expert system shell was first implemented in software. The initial knowledge base is given in a text file, as a set of rules. The text file is parsed and a belief network is generated, having in its nodes the facts (propositions), linked by edges which contain the main operators: AND, OR, NOT, AGGREGATE and DET. The network also contains the ONE-OF operator. This paper continues a previous research done by Cenan in [2], extending it to a hardware implementation. The hardware implementation achieves much better performance by exploiting the parallelism of triggering all the rules in the same time. The system was implemented and simulated on a XC4005E FPGA device from Xilinx Corporation. The main advantage obtained is an increased speed of the implementation, due to the high level of parallelism achieved and to the intrinsic hardware features.