An AI/OR hybrid expert system for data network design

Summarises the philosophy and results to date in building AI /OR hybrid expert systems to aid human users in designing large-scale (upwards of 250 nodes), low cost, packet-switched data communication networks using distributed, adaptive routing, where the goal design has constraints on node degree, node utilization, arc (channel) utilization, min-hop, node connectivity, end-to-end delay, and average packet delay. a The work described has been implemented and tested in DESI, the US Defense Communications Agency's (DCA's) network design system, which is DCA's principal tool for defense data network (DDN) design. A functional description of the expert system, known as XPRT, is given. Also provided are a short example, analysis of the system, and a discussion of areas for future work.<>