Recently, various research projects have focused on developing remote vehicle control based on Mobile Edge Computing (MEC). In these projects, the number of vehicles connected to the edge server, which affects vehicle control performance, and computational resources of the edge server are not considered. We have proposed a remote vehicle control system that controls vehicles coordinated with edge and cloud servers [1]. The previous work can process the remote vehicle control with low delay and suppress the computational resources of the edge servers. However, the previous work does not take into account the number of connected vehicles and the actual delay of the Internet. In this paper, we propose a layered vehicle control architecture that is composed of vehicles and multiple edge servers, referred to as the "Upper edge server (UpES)" and "Lower edge server (LowES)," and apply the architecture to our remote vehicle control system. The UpES and LowES have different characteristic regarding the number of connected vehicles, network delay and computational resources. To evaluate the proposed system, we assume a Japanese network model and estimate the number of vehicles connected to the UpES. Furthermore, we measure the delay of the Internet and evaluate the stability of the vehicle control using the measured delay. Finally, we evaluate the relationship between the number of vehicles connected to the UpES and computational resources of the LowES and UpES by deploying the UpES to various locations.
