Novos algoritmos heurísticos e híbridos para o Problema de Escalonamento de Projetos com Restrição de Recursos Dinâmicos

Abstract

This Thesis presents new methods for solving Dynamics Resource-Contrained Project Scheduling Problem (DRCPSP). This kind of resource is different from others because it is consumed when a project task is activated, but is also produced at the end of this activation. Its maximum amount is not bounded like the renewable resources, which are very common in project scheduling problems. The objective of DRCPSP is to maximize the amount of resources at the end of a planning horizon, through the activation of tasks considered profitable. The DRCPSP may be used to model expansion projects of companies, where the main objective is to obtain the greatest possible amount of resources .It is proposed in this thesis a new mathematical model for the problem, as well as meta-heuristic algorithms and hybrid methods. Some tests showed that the evolutionary algorithms that use a specific form of representation of the solutions are quite efficient compared with other meta-heuristcs. Hybrid methods that use these evolutionary algorithms with the CPLEX optimizer had very good performance in several instances.