A simple stochastic local search for multi-mode resource-constrained multi-project scheduling

In the multi-mode resource-constrained multi-project scheduling (MMRCMPSP) problem we have to find execution modes and starting times for the jobs of multiple projects, such that the total project delay is minimized. The delay of a project is the difference between its makespan and a lower bound on its duration. A secondary objective is to minimize the total makespan of all projects. The jobs of a project compete for renewable and nonrenewable local resources, and for global renewable resources shared among the projects. The duration and resource requirements of a job depend on its execution mode. We present a simple stochastic local search method for this problem. It maintains a feasible permutation and mode selection of the jobs, and improves it by swapping the order of two jobs, or changing the mode of a job. We propose several techniques to reduce size of the analyzed neighborhoods. The method has ranked fourth during the qualification phase of the MISTA challenge, and was qualified for the second phase. We present computational experiments on the twenty instances of the challenge and compare them to the results of the qualification phase.