Comparative Analysis on Collaborative Cloud-Edge-End Computing Architecture of High-Speed Train.

In this paper, we propose two collaborative cloud-edge-end computing architectures for the intelligent services of the high-speed train (HST), in which the edge servers are deployed along the railway lines or in the HST, respectively. To minimize the latency of intelligent services, two average latency minimization problems for two architectures are formulated, in which the impacts of train-ground fast-time varying channels are considered. Then, we investigate the optimal bandwidth allocation strategy for parallel transmission of sensor data and design a resource allocation and task offloading algorithm based on differential evolution (RATODE) to realize the resource allocation and task offloading. In simulations, we compare the performance of cloud-edge-end collaborative computing with local computing, edge computing and cloud computing and analyze the reasons for the different performance of two architectures. The simulation results show that the RATODE algorithm can decrease average task latency than other strategies and the lower latency can be achieved when edge servers are deployed in the carriages of HST. The deployment suggestions of edge servers are also given based on the cost and performance of the two architectures.