An innovative four-layer heuristic for scheduling multi-mode projects under multiple resource constrains

Abstract

In this paper, an innovative four-layer heuristic is presented for scheduling multi-mode projects under multiple resource constraints. For this purpose, a biased-random sampling technique, a local search, a decomposition method, and an evolutionary search mechanism, each in a separate layer, are combined, with each layer passing its output to the next layer for improvement. The procedure has been designed based on the fact that what makes the scheduling of multi-mode projects hard to solve is a massive search space of modes compounded with the starting times of activities. That is why the procedure is aimed at balancing explorationversusexploitation in searching a massive search space. On the one hand, it exploits promising areas further and, on the other hand, it searches unexplored areas for expanding its range. Since the first layer provides an initial solution, and each of the other three layers can either improve the result of its previous layer or keep it unchanged, solutions never deteriorate and hence promising areas are exploited. Moreover, unexplored areas are searched effectively because each layer explores solution space differently than its previous layer. Based on whether or not an improvement each layer can make to the result of its previous layer, the effect of the corresponding layer on the performance of the procedure has been measured.