Integration of AI and OR Techniques in Constraint Programming

Abstract

Lazy reformulations of classical combinatorial optimization problems are new and challenging classes of problems. In this paper we focus on the Lazy Bureaucrat Problem (LBP) which is the lazy counterpart of the knapsack problem. Given a set of tasks with a common arrival time and deadline, the goal of a lazy bureaucrat is to schedule a least profitable subset of tasks, while having an excuse that no other tasks can be scheduled without exceeding the deadline. Three ILP formulations and their CP counterparts are studied and implemented. In addition, a dynamic programming algorithm that runs is pseudo-polynomial time and polynomial greedy heuristics are implemented and computationally compared with ILP/CP approaches. For the computational study, a large set of knapsack-type instances with various characteristics is used to examine the applicability and strength of the proposed approaches.