面向5G的M2M通信低功耗覆盖增强及资源调度的研究
发布时间：2018-08-02 17:37
【摘要】：机器对机器(Machine to Machine,M2M)通信是一种不需要人为干预的机器设备之间的通信。作为物联网(Internet of Things)的关键技术之一,M2M通信被广泛应用于交通、金融、智能家居、环境监测和智能电网等多个领域。移动蜂窝网络具有高速率传输、大范围覆盖、高可靠性、易于部署等特点,是物联网业务的理想载体。但是现有蜂窝网络主要针对人对人(Human to Human,H2H)通信进行优化和设计,而M2M通信独特的业务特点会对蜂窝网络造成挑战。比如低功耗广覆盖(Low Power Wide Area,LPWA)类业务,物联网网络中存在海量机器类通信(Machine Type Communication,MTC)连接需求,这些连接设备速率要求低、时延不敏感,但是对功耗和覆盖非常敏感,而蜂窝网容量有限不能满足大规模MTC设备频繁接入的需求。因此在第五代移动通信系统(the 5th Generation mobile communication technology,5G)中,解决大规模设备接入问题成为5G的关键场景之一。本文针对5G蜂窝网络中LPWA类物联网业务接入问题,提出了基于非授权频谱的覆盖性增强的窄带M2M系统设计方案。针对蜂窝网中M2M通信资源调度问题,提出了基于强化学习的分布式M2M调度算法。本文主要研究内容和创新点如下:1.针对LPWA类业务特性和授权频谱资源紧张问题,提出一种部署在非授权频谱的覆盖性增强的窄带M2M系统。本文详细介绍了系统的物理层设计方案,同时针对M2M通信覆盖增强的研究,提出了在发送端采用重传机制和低阶调制编码,接收端采用多种相应接收机制的方案达到低功耗、广覆盖目的。仿真结果证明了提出窄带的M2M系统相比LTE系统可以获得10～21dB的覆盖增强。2.针对蜂窝网中M2M通信资源调度问题,本文面向5G网络,根据M2M通信业务流量、时延等将M2M业务进行分类。根据M2M业务类型将类型相同的终端设备分成一个簇,基于位置信息将簇内设备分成多个接入组,然后选出组长设备代表全体组内成员申请调度资源。在分组的基础上,提出一种基于强化学习的分布式M2M调度算法,将调度问题建模为多智能体学习机,具有强化学习能力的组长设备们通过基于收集到的环境信息,通过试错的方式寻找最优的无线调度资源完成数据传输。通过与其他先存方法对比,仿真结果证明了该算法的可行性、公平性和优势。
[Abstract]:Machine to machine (Machine to machine M 2m communication is a kind of communication between machine and equipment without human intervention. As one of the key technologies of Internet of things (Internet of Things), M2M communication is widely used in many fields, such as transportation, finance, smart home, environment monitoring and smart grid. Mobile cellular network is an ideal carrier for Internet of things services because of its high speed transmission, wide coverage, high reliability and easy deployment. However, the existing cellular networks are mainly focused on the optimization and design of human (Human to human H2H) communications, and the unique service characteristics of M2M communications will challenge the cellular networks. For example, low power and wide coverage of (Low Power Wide area (Low Power Wide) class services, there are a large number of machine communication (Machine Type communication (Low Power Wide connection requirements in the Internet of things network, these connection devices require low speed, delay is not sensitive, but very sensitive to power consumption and coverage. However, the limited capacity of cellular network can not meet the needs of frequent access to large scale MTC devices. Therefore, in the fifth generation mobile communication system (the 5th Generation mobile communication technology 5G), solving the problem of large scale equipment access becomes one of the key scenarios of 5G. In this paper, we propose a design scheme of narrow-band M2m system based on unauthorized spectrum coverage enhancement for LPWA class Internet of things (IoT) service access in 5G cellular networks. A distributed M2m scheduling algorithm based on reinforcement learning is proposed for M2m communication resource scheduling in cellular networks. The main contents and innovations of this paper are as follows: 1. In order to solve the problem of LPWA traffic characteristics and resource shortage of authorized spectrum, a narrowband M2m system with enhanced coverage in unauthorized spectrum is proposed. In this paper, the physical layer design scheme of the system is introduced in detail. At the same time, aiming at the research of M2m communication coverage enhancement, a scheme of using retransmission mechanism and low order modulation coding in the transmitter is proposed, and the receiver adopts a variety of corresponding receiving mechanisms to achieve low power consumption. Wide coverage purpose. The simulation results show that the proposed narrowband M2m system can obtain the coverage enhancement of 10~21dB compared with LTE system. Aiming at the scheduling problem of M2m communication resources in cellular networks, this paper classifies M2m services according to M2m traffic and delay. According to the M2m service type, the terminal devices of the same type are divided into a cluster, the devices in the cluster are divided into multiple access groups based on the location information, and then the leader equipment is selected to apply for scheduling resources on behalf of all the members of the group. On the basis of grouping, a distributed M2m scheduling algorithm based on reinforcement learning is proposed. The scheduling problem is modeled as a multi-agent learning machine. To find the optimal wireless scheduling resource to complete data transmission by trial and error. The simulation results show that the algorithm is feasible, fair and superior.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN929.5
本文编号：2160174
[Abstract]:Machine to machine (Machine to machine M 2m communication is a kind of communication between machine and equipment without human intervention. As one of the key technologies of Internet of things (Internet of Things), M2M communication is widely used in many fields, such as transportation, finance, smart home, environment monitoring and smart grid. Mobile cellular network is an ideal carrier for Internet of things services because of its high speed transmission, wide coverage, high reliability and easy deployment. However, the existing cellular networks are mainly focused on the optimization and design of human (Human to human H2H) communications, and the unique service characteristics of M2M communications will challenge the cellular networks. For example, low power and wide coverage of (Low Power Wide area (Low Power Wide) class services, there are a large number of machine communication (Machine Type communication (Low Power Wide connection requirements in the Internet of things network, these connection devices require low speed, delay is not sensitive, but very sensitive to power consumption and coverage. However, the limited capacity of cellular network can not meet the needs of frequent access to large scale MTC devices. Therefore, in the fifth generation mobile communication system (the 5th Generation mobile communication technology 5G), solving the problem of large scale equipment access becomes one of the key scenarios of 5G. In this paper, we propose a design scheme of narrow-band M2m system based on unauthorized spectrum coverage enhancement for LPWA class Internet of things (IoT) service access in 5G cellular networks. A distributed M2m scheduling algorithm based on reinforcement learning is proposed for M2m communication resource scheduling in cellular networks. The main contents and innovations of this paper are as follows: 1. In order to solve the problem of LPWA traffic characteristics and resource shortage of authorized spectrum, a narrowband M2m system with enhanced coverage in unauthorized spectrum is proposed. In this paper, the physical layer design scheme of the system is introduced in detail. At the same time, aiming at the research of M2m communication coverage enhancement, a scheme of using retransmission mechanism and low order modulation coding in the transmitter is proposed, and the receiver adopts a variety of corresponding receiving mechanisms to achieve low power consumption. Wide coverage purpose. The simulation results show that the proposed narrowband M2m system can obtain the coverage enhancement of 10~21dB compared with LTE system. Aiming at the scheduling problem of M2m communication resources in cellular networks, this paper classifies M2m services according to M2m traffic and delay. According to the M2m service type, the terminal devices of the same type are divided into a cluster, the devices in the cluster are divided into multiple access groups based on the location information, and then the leader equipment is selected to apply for scheduling resources on behalf of all the members of the group. On the basis of grouping, a distributed M2m scheduling algorithm based on reinforcement learning is proposed. The scheduling problem is modeled as a multi-agent learning machine. To find the optimal wireless scheduling resource to complete data transmission by trial and error. The simulation results show that the algorithm is feasible, fair and superior.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN929.5
【参考文献】
相关期刊论文 前2条
1 张平;陶运铮;张治;;5G若干关键技术评述[J];通信学报;2016年07期
2 高阳,陈世福,陆鑫;强化学习研究综述[J];自动化学报;2004年01期
相关博士学位论文 前1条
1 陈兴国;基于值函数估计的强化学习算法研究[D];南京大学;2013年
相关硕士学位论文 前1条
1 虞靖靓;基于Q学习的Agent智能决策的研究与实现[D];合肥工业大学;2005年
，本文编号：2160174
本文链接：https://www.wllwen.com/kejilunwen/xinxigongchenglunwen/2160174.html
