In this thesis, we firstly describe the properties of job shop scheduling problem and review the various approaches to this problem, the classical approaches and knowledge-based ones. The former includes the methods based on the Operations Research literature such as mathematical programming, priority dispatching rule, and man-machine interactive scheduling. The latter includes the approaches based on the Artificial Intelligence literature such as constraint-directed search method, multiple perspective opportunistic scheduling, executional opportunistic scheduling, nonlinear planning method, rule-based approach, and logic-based ones. Having discussed the previous approaches, we propose KAIS-2 for job shop scheduling.
KAIS-2 is a knowledge-based job shop scheduling system adopting the generate-and-compare procedure. This procedure is useful in solving the problems which have intrinsic combinatorial complexity and dynamic environments like job shop scheduling problems. The characteristics of KAIS-2 are the following: 1) The flexible and structured representation of diverse relevant knowledge in frame form, 2) the generate-and-compare strategy using the multiple knowledge sources, 3) The reflection of user``s preference in resolving the conflict of potential actions through interactive alternative comparison, 4) the learning capability to construct the user``s strategy base automatically, 5) the object oriented programming paradigm for the interactions between resources. The research prototype of KAIS-2 is developed, and illustrated how this system works to solve the job shop scheduling problems.