A human preference-driven object-based scheduling technique

It is not easy to obtain optimal or sub-optimal solutions, if they exist, for a large complex scheduling job like timetable construction. In this paper, we confine a scheduling environment where objects have their own events competing for better slots on boards, objects have their own board slot preferences, and objects themselves belong to one or more classes of a society which globally constrains them. Neither conventional nor any improved backtracking alone is suitable for a complex scheduling job because of the huge search space, and central control alone is not enough because of the complexity of the scheduling problems. Recent developments in constraint languages like Prolog III may be considered as general approaches for complex scheduling problems, but only restrictive constraints are generally allowed, while humans also use preference constraints to solve these problems. We suggest a novel human preference-driven scheme: the added features include dynamic object switching with the class hierarchy considered, intelligent backtracking where temporal constraints are used for deciding backup points, and case-based revisions for refining or correcting old scheduling. This scheme has been implemented as a human-like constraint language and successfully tested for short-term lecture scheduling. Finally, to evaluate our approach, two degrees of satisfaction are defined and calculated for the tests