Reactive scheduling framework based on domain knowledge and constraint programming

Abstract

Industrial environments frequently face disruptive events. This contribution presents a support framework, aimed at addressing the repair-based reactive scheduling problem. It is based on an explicit object-oriented domain representation and a constraint programming (CP) approach. When an unforeseen event occurs, the framework captures the in-progress agenda status, as well as the event effect on it. Based on this information, a rescheduling problem specification is developed. Tasks to be rearranged are recognized and the set of the most suitable rescheduling action types (e.g. shift-jump, reassign, freeze) is identified. Since a given specification may lead to several solutions, the second stage relies on a CP model to address the problem just defined. To create such model, action types are automatically transformed into constraints. Provided that good quality schedules can be reached in low CPU times, alternative solution scenarios focusing on stability and regular performance measures can be posed for each problem.