Learning-based small cell base station selection scheme involving location privacy in service migration

Small cell base stations will be deployed at high density in next generation cellular networks. Mobile terminals need to select them more frequently to void the interruptions of location-based services in service migration. Consequently, some small cell base station selection schemes have been proposed. But these works left out location privacy protection and the feature that small cell base stations have cached some candidate result sets. Therefore, this paper first defines perceived latency, migration cost, and location privacy level, and then proposes a method that takes location privacy protection and candidate result sets into account in Actor-Critic with baseline algorithm. It takes the distance between mobile terminal and small cell base stations and the content feature vectors of cached content as environment state to select the small cell base station. The performance is evaluated on the AutoTel dataset after data engineering, and experimental results show that the overall effect of the proposed method outperforms three traditional algorithms and a DQN-based algorithm.