Learning control knowledge through cases in schedule optimization problems

We have developed an integrated framework of iterative revision and knowledge acquisition for schedule optimization, and implemented it in the CABINS system. The ill-structuredness of both the solution space and the desired objectives make scheduling problems difficult to formalize and costly to solve. In CABINS, situation-dependent user´s preferences that guide schedule revision are captured in cases together with contextual information. During iterative repair, cases are exploited for multiple purposes, such as (1) repair action selection, (2) repair result evaluation and (3) recovery from revision failures. The contributions of the work lie in experimentally demonstrating in a domain where neither the human expert nor the program possess causal knowledge that search control knowledge can be acquired through past repair cases to improve the efficiency of rather intractable iterative repair process. The experiments in this paper were performed in the context of job-shop scheduling problems