【作者】 黎峰；

【导师】 贾智平；

【作者基本信息】 山东大学 ， 计算机科学与技术， 2019， 博士

【摘要】 物联网作为新一代信息技术的重要组成部分,受到了世界各国的高度重视和积极布局。物联网在智慧交通、智慧城市、智慧医疗、智慧工业等领域都取得了大量的创新技术成果、示范性应用,以及规模化的产业发展。其中,无线工业控制网络(wireless industrial control network,WICN)在工业生产环境下通过无线通信技术对物理设备进行监测与控制。作为一种新兴的工业通信技术,无线工业控制网络为物联网在工业生产领域的快速应用提供了理论与技术基础。相比传统的基于有线方式连接的工业控制网络,无线工业控制网络具有低安装与维护成本、高可扩展性、可移动性等方面的优势,因此被越来越广泛的应用于日益复杂的智慧化工业生产环境中。由于节点规模、部署环境等方面的限制,无线网络节点通常采用电池供电的方式。节点能量消耗直接影响网络的运维成本、寿命、以及传输的可靠性,因此成为无线网络领域的首要研究问题之一。与传统的无线传感器网络主要面向环境数据的采集和监测的应用场景不同,当前的无线工业控制网络越来越多的被应用于工业生产环境下现场设备的闭环控制任务当中。实时性决定了工业自动化控制过程中关键信息采集和控制指令的传输是否在可预测的时间内完成。实时网络系统需要对节点间端到端的通信提供传输延迟的毫秒级上界保障,因此对于数据链路层及网络层网络体系架构层次的协议设计提出了新的要求。然而,经典的无线网络节能技术往往没有将实时性作为核心设计约束,因此无法直接适用于以实时采集和控制任务为主的无线工业控制网络中。以WirelessHART无线工业控制网络协议为例,为保障传输的可靠性与实时性,其通常基于链路质量作为单一的考量进行路径选择,并潜在的造成了节点能量黑洞以及调度拥塞等问题。本论文针对上述问题,面向无线工业控制网络路由协议设计过程中实时性、可靠性与低功耗等相互制约的设计与优化目标,在保障传输延迟实时性的基础上,研究高可靠与低功耗的网络通信协议,旨在为设计高可靠、低功耗的实时无线工业控制网络系统提供理论基础与技术支撑。面向上述研究目标,本文以WirelessHART无线工业控制网络协议为研究对象,主要的研究工作和贡献包括以下内容。论文首先针对WirelessHART网络路由算法在路径选择时仅考虑信道质量作为路径选择依据带来的能耗与可靠性方面的弊端,将层数及转发节点剩余能量等因素融入链路选择过程中,在此基础上提出了一种可靠的图路由构建算法和高能效的链路选择算法来提升网络的鲁棒性并延长网络寿命。首先,在路由图的构建过程中,论文提出为只有一个上层转发节点的节点增加一条基于同层节点的转发链路,从而保证在链路选择时,每个节点有足够多的选择。在此基础上,论文提出了在链路选择过程中综合考虑链路质量、能耗、跳数等因素,使用综合分析法决定权重系数,从而得到一种最优的链路选择方案。通过增加冗余链路并进行合理地链路选择,大大增加了传输的成功率并减少重传次数,从而提升网络的鲁棒性并延长网络寿命。最后,在模拟实验中,将本论文提出的算法与现有的JRMNL、RUG和Re-add三种WirelessHART路由协议进行了比较。实验结果表明,本论文提出的算法在延长网络寿命和提高数据传输可靠性方面具有明显的优势。数据融合方法通过减少整个网络中的数据传输次数以降低网络能耗,已成为无线传感器网络中广泛认可的节能技术。然而,现有的数据融合方法往往没有考虑网络通讯的实时性,在转发节点等待与其他数据包进行融合的过程中可能造成部分数据包错过端到端的通信截止时间约束。本文首次提出面向工业无线网络的实时通信的数据融合框架,将数据融合方法与WirelessHART网络路由算法的链路选择和超帧调度阶段相结合,通过数据融合减少数据包的传输次数,并同时保证网络的实时性和可靠性要求。首先,在上传路径的链路选择中,论文充分考虑了链路质量、能耗和剩余能量以及融合因素等关键因素。在融合因素的影响下,选择指向具有较高传输频率的父节点的链路作为上传路径,从而使网络中更多的数据包进行融合。在此基础上,本论文基于贪婪的启发式方法重新构建超帧,使得在中间路由节点处,将可以融合成单个数据包的数据包分配到相近的时隙,从而减少了数据融合的等待时间开销,使得使用数据融合后的网络仍然保持实时性特点。实验结果表明,本论文提出的框架,虽然数据融合需要一定的融合等待时间为代价,但仍然可以满足网络通信的实时性,同时网络在单个节点的能量消耗和整体寿命方面有显著的实质性改进。分簇路由方式作为无线传感器网络的通信能耗优化的经典技术之一,通过周期性选择簇头节点以负责簇内节点的数据转发,均衡网络节点剩余能量并提升网络生命周期。同时,分簇路由的方式可以在簇头转发数据时进行充分的数据融合工作,进一步降低网络通信开销。本论文首次将基于节点频率的分簇路由和超帧调度算法引入WirelessHART网络中。首先,将簇头选择过程引入到WirelessHART协议的上行链路选择阶段,从节点通信能耗、剩余能量、现场设备自身产生数据的频率等方面综合计算权重以决定上传链路的选择,以尽可能的满足超帧调度阶段的实时性约束。在此基础上,重新设计了 WirelessHART通信协议的超帧调度阶段,将超帧调度与基于分簇的数据融合相结合。同时针对数据包大小和节点频率进行优化,来保证数据包的端到端延迟问题。实验评估结果表明,与现有的WirelessHART通信协议相比,本论文提出的基于分簇的路由算法在满足端到端约束的条件下,可以将网络平均寿命提高40%。上述研究工作主要基于主客观相结合的层次分析法和综合分析法进行链路选择,在一定程度上依赖于网络设计过程中的主观判断。此外,上述方法均根据当前网络状态决定下一跳链路的选择方案,忽略了全局端到端路径的能耗状态以及实时性。论文针对上述问题,提出了一种基于LSTM(Long Short Term Memory)的节点能耗预测算法,利用过去一段时间内链路传输损耗及时间信息对节点之间的链路损耗进行了回归分析。然后,利用该分析得到的链路传输损耗,对设备网络中终端节点到网关的链路选择问题进行合理化建模,提出了基于蚁群算法的WirelessHART端到端路径选择和调度算法,降低设备网络数据传输延迟,并平衡网络中各个传输节点的能量损耗。基于LSTM的链路选择方法对网关的计算能力提出了一定的要求。与此同时,相比之前提出的方法,对于网络端到端路径的全局性预测使得网络节点的剩余能量更为均衡,在保障实时性的前提下,进一步提升了网络的寿命。更多还原

【Abstract】 The Internet-of-Things(IoTs)technology has been considered as the core component of the next-generation information technology,which has drawn more and more attentions for countries all over the world.IoT technology has been widely adopted in smart traffic,smart city,intelligent medical care,and intelligent industries,where a great amount of innovative and demonstrative applications and industries have been developed.In particular,wireless industrial control network(WICN)utilizes wireless communication technologies to monitor and control physical devices in an industrial production environment.As an emerging industrial communication technology,WICN enables IoT to be adaptively used in the industrial production with theoretical and technological foundations.Compared with the traditional wireless control networks based on the wired communication,WICN has various advantages including low installation and maintenance cost,high scalability and mobility.Therefore,WICN becomes a promising design choice in the intelligent industrial production systems.Due to the restrictions on the node size and deployment environment,wireless nodes are typically battery-powered.Energy consumption has direct impact on the operation and maintenance cost,lifetime and transmission reliability of the network.Therefore,it has become one of the most important research issues in the field of wireless networks.Unlike traditional wireless sensor networks,which are mainly used for environmental data acquisition and monitoring,the contemporary wireless industrial control network has been widely used in closed-loop control tasks of field equipment in industrial production environment.The real-time constraints determine whether the collection of key information and the transmission of control instructions in the process of industrial automation control can be completed in a predictable time.Real-time network systems need to provide upper bound guarantee for end-to-end communication delay between nodes in the milliseconds granularity,which leads to new challenges for the design of architecture and protocols in data link layer and network layer.However,traditional wireless network energy optimization technologies often do not take real-time as the core design constraints,therefore cannot be directly applied to the wireless industrial control network with real-time acquisition and control tasks as the main design considerationsTaking the standard WirelessHART protocol as an example,in order to ensure the reliability and real-time transmission,it usually chooses the path based on link quality as the key consideration,which potentially causes the problems of node energy black hole and scheduling congestion.Aiming at the above problems,this work aims at the design and multi-objective optimization of real-time,reliability and low power consumption in the design of routing protocols for wireless industrial control networks.On the basis of guaranteeing the real-time transmission delay,the network communication protocols with high reliability and low power consumption are studied in order to design networks with high reliability and low power consumption which provides theoretical basis and technical support for real-time wireless industrial control network system.Given the above research objectives,this work study the routing protocol design of WirelessHART network.The main research work and contributions include the following contents.The standard WirelessHART routing selection only considers channel quality as the input factor,which potentially leads to high energy consumption and low reliability in the network communication.In this work,we propose to integrate layer levels and residual energy of forwarding nodes into the process of link selection.On this basis,a reliable graph routing construction algorithm and an energy efficient graph routing algorithm are proposed.Energy efficient link selection algorithm improves network robustness and prolongs network lifetime.Firstly,in the process of constructing the routing graph,this paper proposes to add a forwarding link to a node at the same layer when a node has only one upper forwarding node,so as to ensure that each node has enough choice when selecting the uploading links.Furthermore,the paper puts forward an optimal link selection scheme by considering link quality,energy consumption,hop number and other factors in the process of link selection,and using comprehensive analysis method to determine the weight coefficient.By adding redundant links and choosing links reasonably,the success rate of transmission is greatly increased and the number of retransmissions is reduced,thus enhancing the robustness of the network and prolonging the network lifetime.Finally,in the simulation experiment the proposed algorithm is compared with the existing JRMNL and RUG WirelessHART routing protocols.The experimental results show that the algorithm proposed in this work has obvious advantages in prolonging network life and improving data transmission reliability.Data fusion has become a widely recognized energy-saving technology in wireless sensor networks by reducing the number of data transmission in the whole network to reduce network energy consumption.However,the existing data fusion methods often fail to consider the real-time requirements of network communication.In the process of data forwarding,data packets need to wait at some nodes to be aggregated with packets from other nodes,which potentially leads to some data packets missing the end-to-end communication deadline constraints.To the best of our knowledge,this paper proposes the first data fusion framework for real-time communication in industrial wireless networks.It combines the data fusion method with the link selection and superframe scheduling stage of WirelessHART network routing algorithm.The proposed framework reduces the transmission times of data packets through data fusion,and guarantees the real-time and reliability requirements of the network.In the link selection of uploading path,the key factors such as link quality,energy consumption,residual energy and fusion factors are fully considered.Under the influence of fusion factors,the link pointing to the parent node with higher transmission frequency is chosen as the uploading link,so that more data packets in the network can be aggregated.On this basis,this work reconstructs the superframe based on the greedy heuristic method,so that at the intermediate routing node,the data packets that can be fused into a single packet can be allocated to similar slots,thus reducing the waiting time overhead of data fusion,so that the network using data fusion still maintains the real-time characteristics.The experimental results show that the framework proposed in this chapter satisfies the real-time requirements of network communication at the cost of a certain fusion waiting time.Meanwhile,the energy consumption of a single node and the overall lifetime of the network are substantially improved.Clustering-based routing protocol is also one of the well-established technologies for energy consumption optimization in wireless sensor networks.By periodically selecting cluster head nodes to be responsible for data forwarding within the cluster,clustering-based routing balances the residual energy of the network nodes and improves the network life cycle.At the same time,clustering routing intuitively makes data fusion easier when the cluster head forwards data,which further reduces the network communication overhead.In this paper,for the first time,cluster routing and superframe scheduling algorithm based on node frequency are introduced into WirelessHART network.Firstly,the cluster head selection process is introduced into the uploading link selection phase of WirelessHART protocol.The weights are calculated comprehensively from the aspects of node communication energy consumption,residual energy,frequency of data generated by field devices themselves to determine the selection of the upload link,so as to meet the real-time constraints of the superframe scheduling phase with the best efforts.Furthermore,the superframe scheduling phase of WirelessHART communication protocol is redesigned,which combines superframe scheduling with clustering-based data fusion.Meanwhile,the size of the packet and the frequency of the node are optimized to ensure the end-to-end delay of the packet.The experimental results show that,compared with the existing WirelessHART communication protocols,the clustering-based routing algorithm proposed in this paper can improve the average network life by 40%under end-to-end real-time constraints.The above-mentioned research work primarily considers the combination of subjective and objective based comprehensive analysis method for link selection,which to a certain extent depends on the subjective judgment in the process of network design.In addition.the above methods determine the next hop link selection scheme according to the local network state,ignoring the global end-to-end path energy consumption status and real-time performance.In order to solve the above problems,we further propose a node energy consumption prediction algorithm based on LSTM(Long-Short Term Memory).The link loss between nodes is analyzed by using link transmission loss and time information over the past period of time.Then,using the link transmission loss obtained from the analysis,the link selection problem from field devices to the gateway in the network system can be modelled.We further propose an end-to-end path selection and scheduling algorithm based on ant colony algorithm for WirelessHART,which reduces the data transmission delay in the equipment network and balance the energy consumption of the transmission nodes in the network.Link selection method based on LSTM puts forward some requirements for computing power of the gateway.At the same time,compared with the previous methods,the global prediction of network end-to-end paths makes the residual energy of network nodes more balanced,and further improves the network lifetime while guaranteeing real-time performance.更多还原