Modelagem e otimização multiobjetivo do problema integrado de escalonamento de tarefas e alocação de recursos com curva de aprendizado em múltiplos projetos

Abstract

In a scenario of accelerated digital transformation, strict control of production processes is increasingly necessary. In this context, emerges the concept of project characterized as a set of temporary tasks for creating a single and unique product, service, or result. As it involves the execution of previously determined tasks with limited resources, project development requires planning and control actions, especially with regard to the optimization of business resources and productivity. In addition to allocating the most appropriate resources to the tasks in the different projects, such actions involve defining a schedule of tasks assigned to each of the allocated resources, which consists of determining a sequence for the execution of the tasks. Additionally, resources can have different skills that not only interfere in their allocation but also in the time required to perform each task. Despite the great influence of scheduling for the quality of allocation, problems are usually solved independently. Thus, this thesis proposes a new optimization algorithm integrated with a simulation model for a unified solution of task scheduling and resource allocation problems in multiple projects, considering the dependency between tasks and resources with multiple skills and learning curve. The proposed algorithm was evaluated and then compared with the correlated literature in relation to resource allocation in multiple projects, allocation of resources with multiple project skills, allocation of multiple resources, with multiple skills, with gain and loss of these skills in multiple projects in a unified way. The algorithm proved to be effective, because it can respond in a unified way to all the problems evaluated and may increase or decrease the complexity of the simulation according to the parameters used. However, However, the proposed.method requires evolution because it presented high execution time of the algorithm in certain simulated scenarios, compared to other evaluated algorithms.