Project management expert system (abstract only)

Effective management and technical support are required for the success of large-scale projects. PERT packaged programs [1], [2], [3] provide scheduling capability. However, many tasks must still be performed by a human being, including activity plan generation, construction of activity networks, modification of a schedule produced by PERT program, and project monitoring. To support these tasks, we believe that the application of artificial intelligence techniques to this area has great potential. Accordingly, we are developing an experimental project management expert system named PROMX.

The main tasks that PROMX supports are activity plan generation, activity scheduling, and project monitoring. The ordinary flow of these tasks is shown in Fig. 1.

Activity plan generation is supported using a knowledge base consisting of the activities and the constituent relationships between them in a given project domain. A user first selects the template of the activity necessary to attain the project goal and assigns values to the attributes. The decomposition of an activity into its constituent parts is recursively performed by PROMX in cooperation with the user. The activity plan generation is followed by the activity scheduling. During the first phase of the activity scheduling, an activity network is constructed using the knowledge of precedence constraints between the activities. Then, time and resources are assigned to each activity in the activity network. This activity scheduling is controlled by the heuristic search method to avoid combinatorial explosions. During project operation, the project is monitored to control its progress. The user supplies progress data to PROMX periodically. PROMX diagnoses whether the progress of the project is problematic (e.g., behind schedule). If some problems are found, PROMX makes a suggestion to deal with them (e.g., rescheduling). These project monitoring capabilities are realized using diagnostic and dealing heuristics.

PROMX is implemented in ESP [4], which is a Prolog-based object-oriented programming language. Various kinds of knowledge in the project domain are represented using the object-oriented feature and the logic programming feature of ESP. To model activities in the given project domain, the knowledge of activities is represented as objects. The knowledge of the constituent relationships between the activities and the precedence constraints between the activities is represented in the form of a Prolog Horn clause. The knowledge of diagnostic heuristics and dealing heuristics is also represented as a Horn clause.

This work is part of the activities undertaken in the Fifth Generation Computer Systems (FGCS) Project of Japan.