| The problem of Resource-Constrained Project Scheduling has been widely studied in the literature and several techniques, both exact and heuristic have been proposed. Exact techniques are not always capable of finding optimal solutions and in other cases, the high computation time required by the NP-hard problem, makes them worthless. Heuristic methods are the alternative for exact methods, and those based on priority rules were the first to be applied. Metaheuristic techniques are nevertheless overpowering the former, since excellent results are being obtained with them. Among metaheuristic techniques, genetic algorithms, tabu-search and simulated annealing are the most widely used.; In this Ph.D. Thesis, we have developed new genetic algorithms to solve the problem, both in its single-mode variant and multi-mode variant. We have designed a new type of codification for the solutions to the problem, adding new information related with the scheduling scheme used to schedule the activities: forward or backward. Moreover, we have developed new crossover and mutation operators, capable of effectively managing the information within the solution. These genetic algorithms have been compared with the best heuristic algorithms published, by using the standard project library PSPLIB. The computational experiment shows the superiority of our algorithms. |
