Deep recurrent neural network for optical fronthaul dimensioning and proactive vBBU placement in CF-RAN

IF 1.8 4区计算机科学 Q3 COMPUTER SCIENCE, INFORMATION SYSTEMS

Photonic Network Communications Pub Date : 2022-02-09 DOI:10.1007/s11107-022-00964-0

Matias R. P. dos Santos, Rodrigo I. Tinini, Tiago O. Januario, Gustavo B. Figueiredo

引用次数: 2

Abstract

In this paper, we solve virtualized passive optical network (VPON) assignment and virtualized baseband unit (vBBU) placement using an integer linear programming formulation, an approximated heuristic using linear relaxation, and a proactive heuristic based on a specific kind of recurrent neural network. We also studied the application of multi-step forecasting in Cloud-Fog Radio Access Network (CF-RAN) traffic demands for joint use with integer linear programming once it allows the solver more time to generate solutions. Also, we examine if the error in batch prediction impacts the final solution in terms of blocking and correctness.

查看原文本刊更多论文

基于深度递归神经网络的CF-RAN光学前传尺寸和主动vBBU放置

在本文中，我们使用整数线性规划公式、线性松弛的近似启发式方法和基于特定类型递归神经网络的主动启发式方法来解决虚拟无源光网络(VPON)分配和虚拟基带单元(vBBU)放置问题。我们还研究了多步预测在云雾无线接入网(CF-RAN)流量需求中的应用，它与整数线性规划联合使用，使求解器有更多的时间来生成解。此外，我们还检查了批预测中的错误是否会影响最终解决方案的阻塞和正确性。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

Photonic Network Communications 工程技术-电信学

CiteScore

4.10

自引率

5.90%

发文量

审稿时长

12 months

期刊介绍： This journal publishes papers involving optical communication networks. Coverage includes network and system technologies; network and system architectures; network access and control; network design, planning, and operation; interworking; and application design for an optical infrastructure This journal publishes high-quality, peer-reviewed papers presenting research results, major achievements, and trends involving all aspects of optical network communications. Among the topics explored are transport, access, and customer premises networks; local, regional, and global networks; transoceanic and undersea networks; optical transparent networks; WDM, HWDM, and OTDM networks and more.