The quality of the road infrastructure plays a major role in the road safety, especially for the autonomous vehicles (AV). The AV contains cameras and lidars able to detect the road markings, obstacles and the other vehicles. The road markings help the AV to identify the path runway and to understand their localization. Thus, the AV must interact with the road infrastructure to drive around without any human interactions with a high automatization level. The failures of the components of the road infrastructure (pavement and road markings) are incompatible phenomena with the operation of AV. To ensure a good evolution of this kind of vehicles, the infrastructure and the vehicles must coordinate, each providing a certain level of service. Thus, an efficient road maintenance must be considered. The proposed paper suggests a maintenance policy for the road infrastructure by grouping the maintenance strategies of the road markings and the pavement. This strategy considers the road infrastructure as serial system. A genetic algorithm is used to group the maintenance activities. This methodology is applied to feedback datasets from both the French National Road 4 and the American pavement using the Long-Term Pavement Performance (LTPP) database.