Joint user association and resource allocation in HetNets based on user mobility prediction

Zhipeng Cheng, Ning Chen, Bang Liu, Zhibin Gao, Lianfen Huang, Xiaojiang Du, Mohsen Guizani

Research output: Contribution to journalArticlepeer-review

17 Scopus citations

Abstract

Virtual small cell (VSC) formed by directional beams is seen as an alternative for the small base station (SBS) within the coverage of macro base station (MBS), to increase system capacity and reduce site cost. However, the flexibility of development for VSC poses challenges to user association and resource allocation in heterogeneous networks. In this paper, we consider the joint problem of user association and resource allocation in VSC aided multi-tier heterogeneous networks. To better analyze the formation of VSC and the impact of user mobility on the system, a user mobility prediction model is firstly constructed based on the Markov model. Then the joint user association and resource allocation problem is formulated to maximize the system capacity. Since the aforementioned problem is a coupling problem, two different solutions, namely, a decoupling solution and a coupling solution, are proposed based on the multi-agent Q-learning (MAQL) method, to find the optimal user association and resource allocation strategy. Moreover, to overcome the state and action space explosion in MAQL and accelerate convergence, the deep Q-network (DQN) is applied. Simulation results reveal that the deployment of VSC can increase the system capacity and spectrum efficiency. The coupling solution achieves better performance than the decoupling solution under a large number of users.

Original languageEnglish
Article number107312
JournalComputer Networks
Volume177
DOIs
StatePublished - 4 Aug 2020

Keywords

  • Deep Q-network
  • Heterogeneous networks
  • Multi-agent Q-learning
  • Resource allocation
  • User association
  • User mobility prediction
  • Virtual small cell

Fingerprint

Dive into the research topics of 'Joint user association and resource allocation in HetNets based on user mobility prediction'. Together they form a unique fingerprint.

Cite this