Solving the resource-constrained project scheduling problem by a variable neighbourhood search

Abstract

The well-known, challenging problem of resource-constrained project scheduling is addressed. A solution scheme based on variable neighbourhood search is presented. The solution is coded by using activity sequences that are valid in terms of precedence constraints. The sequences are turned into valid active schedules through a serial scheduler. The search of the solution space is carried out via generating valid sequences using two types of move strategy. Much of the power of the solution scheme is attributable to repeatedly employing effective lower bounding and precedence augmentation, both of which serve to reduce the solution space. The effectiveness of the solution scheme is demonstrated through extensive experimentation with a standard set of 2040 benchmark problem instances. The best-known solutions have been improved upon for 48 instances and the best-known lower bounds have also been improved upon for 148 problem instances. The results are inferior to the best-known for a relatively small number of problem instances, but even then the average deviation is very small indeed. © 2003 Published by Elsevier B.V.