ISSN 1000-1239 CN 11-1777/TP

Journal of Computer Research and Development ›› 2018, Vol. 55 ›› Issue (8): 1706-1716.doi: 10.7544/issn1000-1239.2018.20180310

Special Issue: 2018数据挖掘前沿进展专题

Previous Articles     Next Articles

A Reinforcement Learning Algorithm for Traffic Offloading in Dense Heterogeneous Network

Wang Qian1,2, Nie Xiushan1,Yin Yilong2   

  1. 1(Department of Computer Science and Technology, Shandong University of Finance and Economics, Jinan 250014);2(Software College, Shandong University, Jinan 250101)
  • Online:2018-08-01

Abstract: With the explosive growth of numbers of Internet users and network traffic, the capacity of cellular mobile communication is already limited. In order to solve the contradiction between the increasing demand for high capacity and the limited resource, traffic offloading technology makes full use of the existing network, which offloads part of traffic from the cellular network into the other network and carries on the cooperation between networks, to improve the capacity of the cellular network greatly. Traffic offloading becomes one of the hot topics in the future research of wireless communication technology. In this paper, based on reinforcement learning, we propose a novel reinforcement learning algorithm for traffic offloading in dense heterogeneous network. Based on the previous experience and performance gain of the user offloading, this algorithm considers the system throughput of each state, and finds the optimal WiFi network access point (AP) by calculating the reward value. We also derive the optimal policy of traffic offloading decision to maximize the throughput of the system. Simulation results show that the reinforcement learning for traffic offloading can effectively avoid the collision caused by over offloading and rapid deterioration of system performance. Our scheme can effectively implement the adaptive traffic offloading control policy and achieve the cooperation between LTE and WiFi network guaranteeing the quality of service for users. The overall throughput of the dense heterogeneous network also reaches the maximum.

Key words: reinforcement learning, dense heterogeneous network, traffic offloading, throughput, utility function

CLC Number: