Using genetic algorithm to assess the robustness of project schedules with countable risks

By nature, space mission projects are ambitious endeavors with goals to expand the limits of our knowledge of the universe. Therefore, they involve the use, or development, of complex expertise as well as sophisticated technologies. This unique complexity coupled with large projects size make these missions prone to delays and overruns. To study the sensitivity of a project to perturbations, we developed a program based on a genetic algorithm to explore the impact of different project configurations on the project robustness. We assume some risks are known a priori. We account for these risks by allowing each activity to be processed in several modes. Hence, we are solving the so-called multi-mode resource constrained project scheduling problem or MRCPSP. In addition, we allow certain modes of activities to dictate the use of other modes of subsequent activities. Our implementation can accommodate several independent objectives. The main idea is to use the multi-objective genetic algorithm to generate and compare different projects schedules using known quantities (resources needs, project duration, ...) and thus reveal unobvious risks from a complex set a dependences. This method allows us to extract information about the robustness of the project schedule without having to resort to the use of probabilistic quantities, which are usually poorly known and more arbitrary. We demonstrate the performances of this algorithm with a fictitious test example that possesses many characteristics of a typical mission developed at the Jet Propulsion Laboratory. Finally, we describe how our approach can be complemented by assigning different probabilities to the different project configurations.