An automated project planner

This thesis describes an investigation into the possibility and practicality of construction project planning utilizing Knowledge Based Systems, Site Layout and Resource Schedule concepts.

Current Expert System planning tools are mainly based on network planning methods. A management technique is presented which does not rely on any network method. Artificial intelligence techniques are combined with this technique to provide a knowledge based computer model which can be used for the programming of construction projects.

The method is based on grid formatted project data in which the project site is divided into small square or rectangular elements. Each element forms a column of finished construction work. Engineering knowledge of construction methods and techniques is formulated in IF ... THEN rules and stored in knowledge files. This knowledge is then used to replace the finished construction work in the grid elements with the intermediate work activities that produce the final product. Another knowledge file is used to organise the tasks in the grid element into the logical construction order that is usually followed when tasks are performed on the site.

To produce a plan of work for the grid formatted project, the grid elements are grouped into workable areas according to the work type and location. A scheduler then produces a plan of work by allocating resource teams to the grouped elements. Resources are allocated according to a defined priority such as cost, availability and work type. Each resource team will be available to be rescheduled as soon as it finishes the required work.

The performance of the method and the program model are measured using several tests. These are designed to measure the ability of the program to generate and organize the required intermediate subtasks, measure the computer speed and memory requirements with respect to the grid size, verify the use of the knowledge base in different project types, discuss the feasibility of scheduling real projects and compare these schedules with human developed schedules.

Conclusions and Recommendations for future work are also presented.