A memetic algorithm to address the multi-node resource-constrained project scheduling problem

The multi-mode resource-constrained project scheduling problem (MRCPSP) is a very general scheduling model. The MRCPSP covers problems where activities can be executed in several ways or modes, and is affected by parameters such as their duration, temporary relationships with other activities, and renewable and non-renewable resource requirements. The objective of the MRCPSP is to select a combination of time/resources to minimize the duration of the project and complete all activities while satisfying all resource constraints and precedence relationships. Here, we describe a memetic algorithm to solve the MRCPSP. This algorithm uses the components of genetic algorithms and variable neighborhoods search to implement (1) an adaptation of the uniform crossover operator, and (2) a local search to assess agents' performance that appropriately guides the evolution of the algorithm and hence generate better solutions. We implement a metaheuristic strategy and compare its performance for solving different instances of the standard PSPLIB and MMLIB libraries. Overall, our memetic algorithm provides suitable solutions for the MRCPSP and shows outstanding performance in all tested instances.

Automated summary

The multi-mode resource-constrained project scheduling problem (MRCPSP) is a general model for selecting a combination of time/resources to minimize project duration while completing all activities and satisfying resource constraints and precedence relationships. This study proposes a memetic algorithm for solving MRCPSP, which integrates components of genetic algorithms and variable neighborhood search to adjust the uniform crossover operator and evaluate agents' performance through local search, showing outstanding performance in different instances of standard libraries.