-- AIMING AT LARGE-SCALE CONPLEX PLANNING EXPERT SYSTEMS

Exploiting AI multiple programming paradigms, a knowledge -based interactive train scheduling system is developed on EUREKA II t . The basic idea consists of the following framework for the flexible and efficient knowledge -base construction. (1) The hierarchical frame network called goal -strategy -net declaratively representing knowledge for scheduling, (2) scheduling experts models called actors cooperating each other through passing messages to invoke efficient procedures so called methods, (3) production rules traversing goal -strategy -net for actors efficiently to utilize various levels of scheduling knowledge, (4) and access -oriented human interface. The field prototype system developed for subway train scheduling is estimated satisfactory by experts, and is to be practically applied. The technology developed here is considered not only useful for practical train scheduling system but also beneficial for building l arge-scale complex planning expert systems involving the allocation of the needed personnel. Telex.