Resource-constrained Multi-project Scheduling Problem Genetic Algorithm Research

The resource-constrained project scheduling problem (RCPSP) is a widespread practical question relating to optimization. The problem studies how to schedule the activities to achieve some optimal managerial objectives under precedence and resource constraints. Due to the complication of the application backgrounds, many scholars paid attention to it and generated rich models. Most of these researches, however, have focused on the single-project scheduling problem. In fact, it is common that multi-projects are carried out simultaneously under the limited resources. Therefore, the paper focuses on how to schedule the activities of multi-projects simultaneously under limited resources.Heuristic algorithm is an increasingly researched method to solve this kind of problem recently due to the low efficiency and limited size of exact algorithm. The paper studies genetic algorithm which belongs to heuristic algorithm and has been effectively applied to single-project scheduling. The model of the paper makes some improvement. First, the model has no fixed value of the priority projects, but confers the value randomly when the initial population in the generation is formed. In this way, it not only increases the robustness of the genetic algorithm, but also finds out the priority of son projects with the projects' shortest total makespan in the follow-up process of the algorithm. Second, after once iteration, the algorithm uses the information of priority value carried by the excellent individuals to produce a part of new individuals and put them into the new population. This new part of the individuals on the one hand can reduce the possibility of premature convergence of the algorithm; on the other hand, their production has used genetic information to maintain the overall quality of the population.Finally, the paper takes an example of 6-projects problem and analyzes every parameters of the algorithm. The paper also compares with three other heuristic algorithms based on priority and finds out that the test results of this algorithm are better than the other three algorithms. So, the effectiveness of this algorithm is confirmed.