An ACO Look-Ahead Approach to QOS Enabled Fault- Tolerant Routing in MANETs

合集下载

一种MANET中基于位置的可信路由

第２８卷第５期
２１年９００月
应
用
科
学
学
报
Ｖｌ．２Ｎｏ．５０１８
Ｓｐ．００ｅ２１
ＪＯＵＲＮＡＬ０ＦＡＰＰＬＥＤＩＳＣＩＥＮＣＥＳ— ＥｌｃｒｎｉｓａｄＩｏｍａｉｎＥｎｇｎｅｉｇｅｔｏｃｎｎｆｒｔｏｉｅｒｎ
２ｏｌｅｏｖｒｅｄｃｔｎＮａｊｇＵｉｅｓｔ｛ｓｓａｄＴｌｏ．ＣｌｇｌＯｅｓａＥｕａｉ．ｎｉｎｖｒｉｏＰｏｔｎｅｅｍｍｕｉａｉｎ。ｅｏｎｙｃｎｃｔｓｏ
Ｎａｊｎ２０３ｎｉｇ１０，Ｃｈｎａ０ｉ
３ｏｌｅｏｏｕｅｓＮａｊｎｎｖｒｉｆｏｔａｄＴｌｏ．ＣｌｇｆＣｍｐｔｒ，ｎｉｇＵｉｅｓｔｏｓｎｅｃｍｍｕｉａｉｎ。ｎｉｇ２００，ＣｉａｅｙＰｓｅｎｃｔｓＮａｊ１０３ｈｎｏｎ
ＡｂｓｒｔＴｏｄｅｌｗｉｈｔｅｓｆｔａｎｓｎｒｓｓｉｇａｔｖｔａｋｒｍａｉｉｕｏｅ，ｔｕｔｍｅｒｃｔａｃ：ａｔｈａｅｙｗｅｋｅｓｉｅｉｔｎｃｉｅａｔｃｓｆｏｍｌｏｓｎｄｓｒｓｔｉｃｉｎｒｄｃｄｔｏｂｔｔｅａｔｃｓｂｏｄｉｇａｔｕｔｍｏｅｎｔｒｐｓｄｄｓａｃ－ａｅｏａｉｎａｄｄｓｉｔｏｕｅｏｃｍａｈｔａｋｙｌａｎｒｓｄｌＯｈｅｐｏｏｅｉｔｎｅｂｓｄｌｃｔｏ — ｉｅ

一种改进的RFID动态帧时隙ALOHA算法

Ｋｙｗｒｓｌｍｌｔ；ＬＨｌｒｈａｔｃｌｓｎｔｒｕｈｕ；ａｉｆｑｅｃｅｔｃｔｎＲＩ）ｅｏｄ：ｆｅｓｔｄＡＯＡａｇｉｍ；ｎ —ｏｉｏ；ｈｏｇｐｔｒｄｅｕｎｙｉｎｆａｉ（ＦＤａｏｅｏｔｉｌｉｏｒｄｉｉｏ０引言
１０４
传感器与微系统（ｒｎｄｃｒｎｃｏｙｔｍＴｃｎｌｉｓＴａｓｕｅｄＭｉｓｓｅｈｏｏｅ）ａｒｅｇ
２１０２年第３卷第７期ｌ
一
种改进的ＲＩ动态帧时隙ＡＬＦＤＯＨＡ算法
潘峥嵘，尚凯
（兰州理工大学电气与信息工程学院。甘肃兰州７０５）３００
ｒｓｎｅｏｅｄｒａｈｅｓｍｅｔｍｅ，ｃｌａｓｌｃｒｎｉａｏｌｓｏＡｎｅｈｎｃｄｄｎｍｉｒｍｅｓｏｔｄｅｐｏｓｎｅｒａｅｔｔａｉｗｈｉｈｗｉｌｃｕｅｅｅｔｏｃｔｇｃｌｉｉｎ．ｎａｅｙａｃｆａｌｔｅＡＬＯＨＡｌｏｉｈｍｏｎｉｃｌｉｉｎｉａｇｒｔｆｒａｔ— ｏｌｓｏｎＲＦＩｓｔｍｓｉｒｐｓｄｙｎａｙｉｇｆａｓｏｔｄＡＬＯＨＡｌｏｔＤｙｓｅｓｐｏｏｅｂａｌｚｎｒｍｅｌｔｅａｇｒｈｍ．ｉＴｈｓａｇｒｔｍａｅｈｉｌｏｈｍｋｓｔｅＲＦＩｓｓｅｇｔｔｅｍａｉｍｈｏｇｐｎｍｐｏｅｔａｓｒｃｇｉｉｎｆｃｅｃｙｉＤｙｔｍｅｈｘｍｕｔｒｕｈｕｔａｄｉｒｖｈｅｔｇｅｏｎｔｏｅｉｎｙｂｉｅｔｍａｉｈａｕｂｒｗｉｈｎｈｃｐｆｔｅｄｒａｄｓｔｉｇｔｅｂｅｔｆａｅｌｎｔＳｉｌｔｏｅｓｌｓｓｏｓｉｔｎｇｔｅｔｇｎｍｅｔｉｔｅｓｏｅｏｈｅｒａｅｎｅｔｎｈｓｒｍｅｇｈ．ｍｕａｉｎｒｕｔｈｗｔａｈｏ￣ｙｏｈｒｐｅｌｏｔｍｓｉｏｅｃｍｐｒｄｗｉｒｄｔａｈｔｔｅｐｒｐｅｆｔｅｐｏｏｓｄａｇｒｈｉｍｐｒｖｄ，ｏａｅｔｔａｉｉｌＡＬＯＨＡｌｏｔ．ｉｈｏｎａｇｒｈｍｉ

无线传感器网络中异常读数检测算法研究

第２７卷第９期
２１００年９月
计算机应用研究
ＡｐｌａｉｎＲｅｅｒｈｏｏｕｅｓｐｉｔｓａｃｆＣｍｐｔｒｃｏ
Ｖｏ．７Ｎ．１２ｏ９Ｓｐ２０ｅ．０１
无线传感器网络中异常读数检测算法研究木
ＡｂｔａｃｉｎｌｏｔｍｓｉＳａｅｏｏｅｄｓｄａｔｇｓｓｃｓｌｗｅｔｃｉｎｐｅｉｓｒｃｕｄｔｔｔｅｅｉｔｎｕｌｅｅｅｔｏａｇｒｈｎＷＮｒｆｓｍｉａｖｎａｅｕｈａｏｒｄｅｅｔｏｒｃ— ｉｓｏｉｎ，ｈｇｅｏｉｈｒｃｍｍｕｉａｉｎｃｍｐｌｘｔｎｄｃｍｐａｉｎｌｃｍｐｌｘｔｕｅｔｏｎｕｈｃｎｓｄｅａｉｎｏｈｐｔｏｔｍｐｒｌｎｃｔｏｏｅｉａｏｕｔｔｏａｏｙｅｉｄｏｎｔｅｏｇｏｉｒｔｏｆｔｅｓａｉ－ｅｏａｙ
ＡｎｍａｙｒａｉｇｄｔｃｉｎａｇｒｔｍｎＷＳｏｌｅｄｎｅｅｔｏｌｏｉｈｉＮ
ＸＵＥＡｎ－ｏｇ，ＬＩＭｉｇｒｎｎ
（ｃｏｌｆＣｍｕｅＳｉｃＴｌｏｕｉｔｎＥｇｎｅｎ，ｉｎｓｎｅｓｙＺｅｉｕｍｎ１０３，ｈａＳｈｏｏｏｐｔｃｎｅ＆ｅｃｍｍｎｃｉｎｉｅｒｇＪａｇｕＵｉｒｔ，ｈｎａｇＪ￣ｕ２２１Ｃｉ）ｒｅｅａｏｉｖｉｊｎ
有算法。
关键词：无线传感器网络；异常检测；时空关联性；分布计算；隐私保护中图分类号：Ｔ３１Ｐ１文献标志码：Ａ文章编号：１０ —６５２１）９３５４０１３９（０００．４２０ｄｉ１．９９ｊｉｎ１０ —６５２１．９０７ｏ：０３６／．ｓ．０１３９．０００．６ｓ

考虑网络吞吐量的异构无线传感器网络分簇路由算法

第４４卷第３期２０２２年５月沈　阳　工　业　大　学　学　报ＪｏｕｒｎａｌｏｆＳｈｅｎｙａｎｇＵｎｉｖｅｒｓｉｔｙｏｆＴｅｃｈｎｏｌｏｇｙＶｏｌ４４Ｎｏ３Ｍａｙ２０２２

收稿日期：２０２０－１１－１３．基金项目：陕西省科技计划项目（２０１８ＪＭ６０９９）．作者简介：许知博（１９８６－），男，河南尉氏人，高级工程师，硕士，主要从事电力系统信息化及通信等方面的研究．

ｄｏｉ：１０．７６８８／ｊ．ｉｓｓｎ．１０００－１６４６．２０２２．０３．１５考虑网络吞吐量的异构无线传感器网络分簇路由算法

许知博１，２，段　新３（１西安电子科技大学电子信息学院，西安７１０１２６；２陕西省地方电力（集团）有限公司信息化工作部，西安７１００７５；３陕西省地方电力物资有限公司，西安７１００３８）

摘　要：针对在异构无线传感器网络信息聚类过程中，当层数为３～５层时，存在网络吞吐量较低的问题，提出一种异构无线传感器网络分簇路由算法．分析异构无线传感器网络能耗的无线电一阶模式，构建异构无线传感器网络的能耗模型．当簇群请求节点接收到发送于簇头的码分多址编码与时分多址时隙后，转发数据并使其稳定传输；引入狼群算法建立路由路径，实现异构无线传感器网络分簇路由算法优化．结果表明，异构无线传感器网络层数为３～５时的网络吞吐量均得到提高．关　键　词：狼群算法；异构无线传感器网络；聚类路由算法；多路径衰落模型；自由空间模型；数据稳定性；多址编码；路径匹配中图分类号：ＴＰ３０１文献标志码：Ａ文章编号：１０００－１６４６（２０２２）０３－０３２６－０５

Ｃｌｕｓｔｅｒｉｎｇｒｏｕｔｉｎｇａｌｇｏｒｉｔｈｍｆｏｒｈｅｔｅｒｏｇｅｎｅｏｕｓｗｉｒｅｌｅｓｓｓｅｎｓｏｒｎｅｔｗｏｒｋｓｃｏｎｓｉｄｅｒｉｎｇｎｅｔｗｏｒｋｔｈｒｏｕｇｈｐｕｔ

ＸＵＺｈｉｂｏ１，２，ＤＵＡＮＸｉｎ３（１．ＳｃｈｏｏｌｏｆＥｌｅｃｔｒｏｎｉｃＥｎｇｉｎｅｅｒｉｎｇ，ＸｉｄｉａｎＵｎｉｖｅｒｓｉｔｙ，Ｘｉ’ａｎ７１０１２６，Ｃｈｉｎａ；２．ＩｎｆｏｒｍａｔｉｚａｔｉｏｎＤｅｐａｒｔｍｅｎｔ，ＳｈａａｎｘｉＰｒｏｖｉｎｃｉａｌＥｌｅｃｔｒｉｃＰｏｗｅｒ（Ｇｒｏｕｐ）Ｃｏ．Ｌｔｄ．，Ｘｉ’ａｎ７１００７５，Ｃｈｉｎａ；３．ＳｈａａｎｘｉＰｒｏｖｉｎｃｉａｌＥｌｅｃｔｒｉｃＰｏｗｅｒＳｕｐｐｌｉｅｓＣｏ．Ｌｔｄ．，Ｘｉ’ａｎ７１００３８，Ｃｈｉｎａ）

QoS

的智能网络，可以自动区分重要的应用数据和普通它
生哪种数据包。没有分类，网络就不能确定对特殊数据包要进行的处理。所有应用都会在数据包上留下可
以用来识别源应用的标识，分类就是检查这些标识，识别数据包是由哪个应用产生的。２标注功能。在识别数据包之后，对它进行标）要
维普资讯
Ｃｉｄｃｔｎｌｅｈｉｅ＆ＥｕｍｎｈａＥｕａｉａＴｃｎｕｎｏｑｑｉｅｔｐ
技术大
观
ＱＳ网络功能与设计ｏ
目前，经实现网络信息化的机构、位和企业，已单基本上都处于一种软件应用的环境中，计算机局域其网都必须拥有足够的带宽和强大的服务质量（ｏ）ＱＳ，以便在任何条件下都能提供可靠和可预测的应用性能。可想而知，如今应用多样化的情况下，有ＱＳ在没ｏ的网络会影响整个网络的运行。通常在是对关但键应用和多媒体应用就十分必要。尤其当网络过载或拥塞时，ｏＱＳ可以确保重要业务量不受延迟或丢弃，同时保证网络的高效运行。拥有ＱＳ的网络是一种新型ｏ
时间来实现。由于电子变焦，图像像素间会产生黑部分，用垂直内插和水平内插来消除它。电子变焦倍利数太大时，引起图像颗粒明显的缺点。视频展示台会

demonstra

QoS-Aware Cross-Layer Multicasting for Optical Packet-Switched Networks: Simulation Exploration and Test-BedDemonstrationCaroline P. Lai(1), Balagangadhar G. Bathula(2), Vinod M. Vokkarane(2), and Keren Bergman(1)(1)Dept.ofElectricalEngineering,ColumbiaUniversity,NewYork,NY;********************.edu(2) Dept. of Computer and Information Science, University of Massachusetts, Dartmouth, MAAbstract Cross-layer quality-of-service-aware packet multicasting is investigated for optical packet-switching network fabrics. We present both a numerical simulation exploration of the cross-layer routing algorithms and an experimental demonstration on an optical switching test-bed with 10×10-Gb/s wavelength-striped packets.IntroductionA novel Internet architecture will be essential toaccommodate the exploding bandwidthdemands faced by the current infrastructure.The next-generation design should leverageinnovative optical technologies to offer a moreintelligent, programmable optical layer withflexible bandwidth allocation and dynamicinteraction with higher network layers1. Weenvision an integrated platform for optical cross-layer (OCL) network communication and control(Fig. 1). OCL enhanced designs will facilitate theextraction of optical performance monitoring(OPM) measurements directly from the optical layer to optimize performance2. These OCL routing protocols must also invoke quality-of-service (QoS) classes on the optical layer. Ultimately, the OCL-optimized algorithms must provision for the data’s QoS as well as for the physical-layer performance and impairments2-5.The future Internet should also engage emerging physical-layer technologies, such as optical packet switching (OPS)6. OPS networks comprise a favourable technology approach to enable the flexible high-bandwidth, low-latency interconnections required by future Internet applications. OPS fabrics may be deployed within optical network routers to support high-bandwidth multi-wavelength packet streams between line cards. Additionally, OPS fabrics may achieve a high level of programmability to transparently route wavelength-division-multiplexed (WDM) packets entirely in the optical domain. A significant application that may leverage the greater functionality and programmable flexibility is broadband packet multicasting. We define packet multicasting as the ability to simultaneously transmit broadband multi-wavelength optical messages from a single source to multiple output destinations7. Multicasting may be advantageous in high-bandwidth applications, such as networked gaming and real-time diagnostic telemedicine.Broadband QoS-based packet multicasting constitutes an important functionality for future OPS networks. Notably, for bandwidth and latency sensitive applications, such as real-time collaboration, high-QoS packet transmission may be leveraged to provide a high-quality communication link. Here, we explore an OCL-enabled platform whereby a packet multicasting operation is realized accounting for both the message’s QoS and physical-layer degradation. The concept of cross-layer QoS-aware multicasting is investigated both in simulation and with a test-bed demonstration. We first provide a simulation-based comparative analysis between shortest distance and minimum hop routing algorithms using the NSF network. We then experimentally demonstrate the OPS fabric within one NSF node, validating the error-free operation of cross-layer QoS-based multicasting with bit-error rates (BERs) less than 10-12 and a power penalty of 2 dB. Simulation ValidationThe proposed OCL algorithms for QoS-aware packet multicasting are first investigated in simulation. One-way signaling is used to reduce the end-to-end packet transmission latency. The 14-node NSF network topology (Fig. 2) is Fig. 1: Cross-layer-optimized stack, indicating thebidirectional information flow between the application(top), network and routing (middle), and optical layers (bottom) enhanced to provide QoS guarantees.ECOC 2010, 19-23 September, 2010, Torino, Italy Th.9.A.5 978-1-4244-8534-5/10/$26.00 ©2010 IEEEnumerically simulated using a global control plane to track each node’s QoS performance. A centralized routing and wavelength assignment(RWA) scheme is realized. Packets areassumed wavelength-striped, using tenwavelength channels each at 10 Gb/s.Packets are simulated as discrete events 3. The packets follow a Poisson arrival rate and depart with exponential service times. Upon an arrival event, each packet is assigned to a request and then routed based on a minimum distance routing (MDR) or a minimum hop routing (MHR) algorithm. The necessary QoS parameters are retrieved from the application layer. Optical packets reaching the destination ensure that the threshold requirements imposed by the application layer are met 8. The QoS is embedded in the control signal and is updated as the packet propagates through the network. The QoS parameters consist of its BER, latency, priority, and the reliability of the link. At each node, the QoS of the routed packet is computed online and compared with the threshold requirement of the application. If the QoS parameters are violated, the packet is dropped or rerouted on an alternate path if available. Multicasting is initiated as required.An intelligent, efficient control plane acts as a middleware between the application and optical layers 8. Based on the control plane decision, the optical packet is routed on the link.Using the parameters in Tab. 1, the BER is estimated based on the optical-signal-to-noise-ratio (OSNR). Since the BER is a nonlinear function, we compute the link’s noise factor. The overall noise factor of the lightpath is computed as a product of the individual noise factors of the links 8. The overall latency of the packet is the sum of the individual latencies of the links. The reliability of the switch is based on the downtime and path restoration time. The priority parameter enables possible packet routing on alternate network paths.The performance of the proposed QoS-aware cross-layer multicasting is simulated using the NSF network with the distances scaled down by a factor of ten, due to the lack ofregenerators at the node’s switching fabrics. InFig. 3, we compare the performance of the NSFtopology in terms of packet loss, averagelatency of the packet, hop count, and execution time for the routing algorithm. The x-axis for all the plots in Fig. 3 is the offered network load in Erlang, defined as the ratio of the arrival rate to the departure rate. In Fig. 3(a), we observe that MHR offers lower loss compared to MDR at low network loads. This indicates that packets routed based on hopcount show a higher probability of successfully guaranteeing the QoS imposed by the application layer. The average latency (Fig. 3(b)) of MHR is higher than MDR; this may not be problematic if the latency threshold is still satisfied. Thus, the routing layer can adopt ahop-routing at lower network loads. As the load increases, the packet loss for both algorithms converges (Fig. 3(a)). In order to optimize performance, the application layer can instruct the routing layer to switch to distance routing at higher network loads. Thus, cross-layer communication helps to achieve design trade-offs and provide the necessary QoS. We also compare the average hop count for the two routing methods. It is evident that the hop countfor the MHR is lower than MDR (Fig. 3(c)). A decrease at higher loads indicates that providing QoS for optical packets that traverse longer hops is more problematic. Fig. 3(d) shows the execution time (in hours) required for the simulations using a 2.33-GHz Quad Core Xeon processor with Hyper-Threading and 8-GB RAM.Fig. 2: NSF topology with bidirectional links between thenodes, each carrying 10×10-Gb/s packets.Tab. 1: Simulation Parameters. Parameter Value Number of Packets 106BER 10-9Latency 1 msInput Optical Power -10 dBmInline Amplifier Gain 14 dBSwitch Crosstalk Ratio 25 dBStarting Wavelength 1537.4 nmWavelength Spacing 2.8 nmFig. 3:Performance of the scaled NSF network.Experimental DemonstrationThe QoS-based broadband packet multicasting operation is experimentally demonstrated on a multicast-capable OPS fabric test-bed 7 (Fig.4). The fabric test-bed represents the optical switching fabric deployed within one node of the NSF network. The multistage test-bed is implemented with 2×2 photonic switches, which use semiconductor optical amplifiers (SOAs). Wavelength-striped packets are supported, with control information (e.g. frame, address, QoS) encoded on a subset of wavelengths and the payload segmented and modulated at a high data rate (e.g. 10 Gb/s) on the rest of the band. The 2×2 switches detect the control information at the packet’s rising edge using filters and receivers. The packet’s header bits are processed electronically at each stage. The routing control logic gates the correct SOAs to provide the desired routing. No optical buffers are used. The multicast-capable fabric 7 is realized with a multistage design, using differing packet routing (PR) and packet multicasting (PM) stages. The stages have distinct control logic that depends on the recovered header bits. An SOA-based receiver is realized 4 whereby the real-time performance of optical packets can be monitored. Switching is triggered on the per-packet QoS and signal degradation (here, BER). Low-QoS/high-BER packets are detected by the cross-layer receiver and rerouted on an alternate path. The pseudo-BER signal is generated offline in place of an OPM, though a real-time packet OSNR monitor may be used 2. The QoS-aware packet multicasting is validated on the 4×4 optical fabric test-bed with two PR and three PM stages. The 10×10-Gb/s wavelength-striped packets are 120-ns long, analogous to the Ethernet MTU. The 1500-B packets are modulated by a LiNbO 3 modulator with 27-1 PRBS; the payload wavelengths range from 1537.4 to 1564.0 nm. Fig. 5 depicts the pattern of optical packets injected in the multicast-capable fabric with two QoS levels (high/low priority). The QoS and packet signal quality are assessed and a real-time decision is made to forward or reroute the message on aprotection path. At the output, we verify that error-free QoS-based packet multicasting is achieved. BERs<10-12 are obtained on all ten payload wavelengths. BER curves for the system show a 2-dB power penalty (Fig. 6). ConclusionsFuture networks will require a QoS-aware cross-layer protocol stack. This work confirms that broadband packet multicasting can be realized accounting for physical-layer access in a cross-layer-optimized approach. Numerical results and a demonstration on a fabric test-bed show that packet multicasting can be performed based on QoS and signal degradation. This exploration leverages an OCL-optimized platform and novel optical technologies to achieve performance gains for next-generation networks.We acknowledge support from the CIAN NSF ERC (subaward Y503160); BBN, GENI Project Office (agreement 1631); and the NSF SOON project (grant CNS-0626798).References1 CIAN, 2 i et al., Proc. OFC’10, OTuM2 (2010).3 F.Fidler et al., Proc. ECOC’09, 2.5.2 (2009).4 i et al., Proc. ECOC’09, 2.5.3 (2009).5 S.Azodolmolky et al., Proc. ONDM’09 (2009).6 E.W.M.Wong et al., JLT 27 (14) (2009). 7 i et al., Proc. OFC’10, OWI4 (2010).8 B.G.Bathula et al., TON 18 (1) (2010).Fig. 4: Experimentally implemented multicast-capablefabric architecture and test-bed photograph.Fig. 6: Sensitivity curves with insets of the 10-Gb/s eyediagrams (input: left, output: right).Fig. 5: Optical waveforms corresponding to the QoS-aware packet multicasting operation.。

一种简单的VOQ交换机时延确保分组调度算法_张福阳

4. 5 [5 6]
方法的突出问题是计采用两级交换结构 ,
2 调度算法
在 V O Q 中提供确定性 Q o S 保障 , 必须保证任意端口对在任意时间段中的最小服务量 . 具体体现在两个方面 , 一是输出端口必须提供传输带宽的保证 , 即在适当的时间与相应的输入端口匹配 ; 二是当与输出端口匹配时 , 相应输入端口中的分组要赢得输入竞争 , 使用传输机会 . 基于流的调度 , 还要裁决 V O Q 队列中的分组竞争 . 对于输入、输出竞争 , 为了算法易于物理实现 , 本文采用轮询的方式 , 保证每个 V O Q 队列都 i ,j 能在每个轮询周期内获得一次服务机会 . 裁决 V O Q 队列的竞争采用 E D F 策略 , 即分组在到达时被赋予一个时限 ( d e a d l i n e ) , 该时限用于竞争裁决. 本文假设所有分组都是在时隙开始后到达 , 时隙结束前从输入端口传输到输出端口 . 调度算法具体如下 : 1) 在 0 时刻 , 对端口匹配矩阵 Π初始化 πi ,j= 1 若 j =i 0 其它 ( 2)
DO I : 10 . 13700 / j. bh . 1001 5965 . 2008 . 11 . 027 2008 年 11 月北京航空航天大学学报第 34卷第 11期 J o u r n a l o f B e i j i n gU n i v e r s i t yo f A e r o n a u t i c s a n dA s t r o n a u t i c s
S i mp l es c h e d u l i n ga l g o r i t h mf o r d e l a yg u a r a n t e e i nV O Qa r c h i t e c t u r es w i t c h e s

一种基于蚁群优化的动态节能路由选择策略

一种基于蚁群优化的动态节能路由选择策略屈巍;赵晶;洪洋【摘要】针对无线传感器网络中寻找最优路径的问题,考虑网络的节能需求,提出了一种基于蚁群优化的动态节能路由选择策略.蚁群算法在进行过一段时间后,受转移概率公式影响易于陷入局部最优解,因此在提出的基于蚁群优化的动态节能路由选择策略中设计了动态状态转移优化规则,合理的增加了新节点的搜索概率,从而达到快速有效的寻找全局最优解的目的;此外,基于蚁群优化的动态节能路由选择策略设计了奖罚机制,进一步节省搜索时间的同时增加最优路径搜索概率,极大的延长了网络生存时间.仿真实验及分析表明,通过动态状态转移优化规则及奖惩机制的动态调整极大的增加了全局最优解的搜索概率,快速有效地实现了全局最优解的获得,节省了节点能量消耗,有利于延长网络生存时间.【期刊名称】《沈阳师范大学学报（自然科学版）》【年(卷),期】2016(034)002【总页数】6页(P234-239)【关键词】无线传感器网络;蚁群算法;状态转移优化规则;奖惩机制【作者】屈巍;赵晶;洪洋【作者单位】沈阳师范大学科信软件学院,沈阳 110034;沈阳师范大学科信软件学院,沈阳 110034;巴斯大学科学学院,巴斯 BA27AY【正文语种】中文【中图分类】TP393无线传感器网络由大量具有感知能力、计算能力和通信能力的传感器节点以自组织的方式构成,被广泛的应用于军事及民用领域[1-3]。

由于一般工作在无人值守的监控环境中,网络节点电池更换成本较高,因此无线传感器网络的路由设计更多关注于节点能量消耗的最小化[4-9]。

出于这些原因,节点在较短的时间内找到最短的路径进行通信不仅利于有效降低节点能耗、延长网络生存时间,也大大提高了无线传感器网络的工作效率,这是目前传感器网络路由设计的一个重要研究方向。

由Dorigo M等在1991年首次提出的蚁群算法成为解决此问题的一类典型研究方法,近年来受到广泛研究[10-13]。

满足QoS约束的自适应Ad Hoc网络路由算法

［ｓａｔＡｂｔｃ］ＡｄＨｏｅｗｏｋｉａｔｍｐｒｒｌ—ｏｉｌｓｃｍｐｔｒｏｒｃｎｔｒｅｏａｙｍｕｔｈｐｗｒｅｓｏｕｅｍｍｕｉｔｎｎｔｒａｏｏｅｆｂｌｈｓｏｓｉｅｃｎｃｉｅｗｏｋｔｔｓｍｐｓｄｏｉｏｔｃｍｍｕｉａｎａｏｈｉｃｍｏｅｓｎｃｔｇｉ
中提供ＱＳ保证是非常困难的。因此，Ｑｏｏ软Ｓ和自适应ＱＳｏ是在移动ＡｏｄＨｃ网络中提供ＱＳ保证的２个妥协准则…。ｏ基于Ｑｏ路由协议的基本目标是确定满足ＱｏＳＳ需求的一条自
附加的参数相关的约束问题。处理与附加参数相关问题的方法则要困难更多。因此，不失一般性，这里考虑附加的方法。定义１多约束路径（ｌ—ｏｓａｅａ，Ｐ问题：ＭｕｔＣｎｔｉｄｔＭＣ）ｉｒｎＰｈ考虑一个网络Ｇ（，）ＶＥ。每条边（ｆ ∈Ｅ与个附加权Ｗ（）ｋｊｉ ≥０ｉ１，Ｋ相关联。给定个约束，，， …，＝２＝１，Ｋ，问题， …，２
文献标识码：Ａ
中圈分类号ｌＴ３３２Ｐ９．０
满足ＱｏＳ约束的自适应Ａｃ网络路由算法ｄＨｏ
张书奎
（苏州大学计算机科学与技术学院，苏州２０）０６１５
摘
一
要：自组网是由一组带有无线收发装置移动节点组成的一个能够支持多跳的临时性计算机通信网络，其拓扑的动态变化是该类网络的
ｗｉａｈｏｅｒｕｈｗｉｌｓｎｓＤｙａｃｏｏｏｙｉａｙｉａｃａａｔｒＡｎｗＱＳｒｕｎｌｏｔｍｒｂｌＡｄＨｏｅｗｏｋ，ｈｃｈｔｅｃｔｒｏｇｒｅｓｉｋ．ｎｍｉｐｌｇｐｃｌｈｒｃｅ．ｅｏｏｔｇａｇｒｈｔｈｅｌｔｓｔｉｉｈｆｉｃｎｔｒｓｗｉｉｏｍｏｅｈｓｎｍｅＳＡｒｅｆＡａｔｅＲｏｔｇＡｌｏｉｍ（ＳＲＡ）ｉｒｐｓｄｉｉｐｐｒＩｉｔｇａｅｅｍｕｔｌＱｏａａｔｒｏａｉｔａｄＱｏ — ｗａｅＳｌｄｐｉｕｉｇｒ－ｖｎｈｔＱＡ—Ａ，ｓｏｏｅｔｓａｅ．ｔｎｅｒｔｓｌｐｅＳｐｍｅｅｓｆｒｆｃｉｏｐｎｈｈｔｉｒｔｎ

基于模糊QoS满意度的启发式多约束路由算法

第１２期
刘源旭等：于模糊ＱＳ基ｏ满意度的启发式多约束路由算法
・３・５
量，包括此条边上带宽、时延、时延抖动和丢包率等等，
估价函数：ｈｘ＝（）＝（）４
其中第ｉ个参数可以表示为ｃ Ⅱ ；，）ｐ＝（，。： … ，（ｓ，，
中图分类号：Ｐ０．Ｔ３１６文献标识码：Ａ文章编号：６３６９２１）２０５ — ３１７ — ２Ｘ（０１１— ０２０
ＡｕｉｔｃＭｕｔ－ｎｔａｎｓＲｏｔｎｇｒｔｍｓｄＨｅｒｓｉｌＣｏｓｒｉｔｕｉｇＡｌｏｉｉｈＢａｅ
是Ｎ — ｏｌｅ问题，ＰＣｍｐｔｅ多数ＱＳ路由算法都是根据ｏ
ｉｌｏｉｍ）ｓｒｇＡｇｒｈＥ算法，给定的多个加性约束合成为ｎｔ］将
个综合非线性约束，后再用最短路径算法求得然ＱＳ路由。还有一些智能方法，ｏ如遗传算法、］蚁群算
路由算法，模糊处理ＱＳ约束的基础上构造基于在ｏＱＳ满意度与路径距离相结合的启发式估价函数，ｏ通
过启发式搜索进行寻路，具有时间复杂度低、响应时间
［］３提出的算法只针对延迟和带宽两个约束，剪掉先
网络拓扑中不符合要求的链路，以延迟为权重利用再
基金项目：国家＂６ ” ８３计划资助项目（０９Ａ１２２２０Ａ０Ｚ０）作者简介：刘源旭（９８）男，１８一，安徽铜陵人，硕士研究生，研究方向为基于ＩＰ的下一代通信网络。

1、下载文档前请自行甄别文档内容的完整性，平台不提供额外的编辑、内容补充、找答案等附加服务。
2、"仅部分预览"的文档,不可在线预览部分如存在完整性等问题,可反馈申请退款(可完整预览的文档不适用该条件!)。
3、如文档侵犯您的权益，请联系客服反馈,我们会尽快为您处理(人工客服工作时间：9:00-18:30)。

ＰＲＯＴＯＣＯＬＳ　ＡＮＤ　ＡＬＧＯＲＩＴＨＭＳ　Ａｎ　ＡＣＯ　Ｌｏｏｋ－－Ａｈｅａｄ　Ａｐｐｒｏａｃｈ　ｔｏ　ｑｏｓ　Ｅｎａｂｌｅｄ　Ｆａｕｌｔ－－　Ｔｏｌｅｒａｎｔ　Ｒｏｕｔｉｎｇ　ｉｎ　ＭＡＮ　ＥＴｓ　

Ｓｕｒｅｎｄｒａｎ．Ｓ“．Ｐｒａｋａｓｈ．Ｓ　Ｐｒｏｆｅｓｓｏｒ，Ｄｅｐａｒｔｍｅｎｔ　ｏｆ　Ｉｎｆｏｒｍａｔｉｏｎ　Ｔｅｃｈｎｏｌｏｇｙ，Ｔａｇｏｒｅ　Ｅｎｇｉｎｅｅｒｉｎｇ　Ｃｏｌｌｅｇｅ，Ｃｈｅｎｎａｉ，Ｔａｍｉｌｎａｄｕ，Ｉｎｄｉａ　Ｐｒｏｆｅｓｓｏｒ，Ｄｅｐａｒｔｍｅｎｔ　ｏｆ　Ｅｌｅｃｔｒｏｎｉｃｓ　ａｎｄ　Ｃｏｍｍｕｎｉｃａｔｉｏｎ　Ｅｎｇｉｎｅｅｒｉｎｇ，Ｊｅｒｕｓａｌｅｍ　Ｃｏｌｌｅｇｅ　ｏｆ　Ｅｎｇｉｎｅｅｒｉｎｇ，Ｃｈｅｎｎａｉ，Ｔａｍｉｌｎａｄｕ，Ｉｎｄｉａ　

Ａｂｓｔｒａｃｔ：ＭＡＮＥＴ　ｒｏｕｔｉｎｇ　１Ｓ　ｃｒｉｔｉｃａｌ　ａｎｄ　ｒｏｕｔｉｎｇ　ｄｅｃｉｓｉｏｎ　ｓｈｏｕｌｄ　ｂｅ　ｍａｄｅ　ｓｏｏｎｅｒ　ｂｅｆｏｒｅ　ｔｈｅ　ｎｏｄｅ　１ｅａｖｅｓ　ｔｈｅ　ｎｅｔｗｏｒｋ．Ｆａｓｔ　ｄｅｃｉｓｉｏｎｓ　ａｌｗａｙｓ　ｃｏｍｐｅｎｓａｔｅ　ｎｅｔｗｏｒｋ　ｐｅｒｆｏｒｍａｎｃｅ．　Ｉｎ　ａｄｄｉｔｉＯｎ，ｍｏ　ｓｔ　ＭＡＮＥＴ　ｒｏｕｔｉｎｇ　ｐｒｏｔｏｃｏｌｓ　ａｓｓｕｍｅ　ａ　ｆｒｉｅｎｄｌｙ　ａｎｄ　ｃｏｏｐｅｒａｔｉｖｅ　ｅｎｖｉｒｏｎｍｅｎｔ，ａｎｄ　ｈｅｎｃｅ　ａｒｅ　ｖｕｌｎｅｒａｂｌｅ　ｔｏ　ｖａｒｉｏｕｓ　ａｔｔａｃｋｓ．Ｔｒｕｓｔ　ａｎｄ　Ｒｅｐｕｔａｔｉｏｎ　ｗｏｕｌｄ　ｓｅｒｖｅ　ａｓ　ａ　ｍａｊｏｒ　ｓｏｌｕｔｉｏｎ　ｔｏ　ｔｈｅｓｅ　ｐｒｏｂｌｅｍｓ．　Ｌｅａｒｎｉｎｇ　ｔｈｅ　ｎｅｔｗｏｒｋ　ｃｈａｒａｃｔｅｒｉｓｔｉｃｓ　ａｎｄ　ｃｈｏｏｓｉｎｇ　ｒｉｇｈｔ　ｒｏｕｔｉｎｇ　ｄｅｃｉｓｉｏｎｓ　ａｔ　ｒｉｇｈｔ　ｔｉｍｅｓ　ｗｏｕｌｄ　ｂｅ　ａ　ｓｉｇｎｉｆｉｃａｎｔ　ｓｏｌｕｔｉｏｎ．Ｉｎ　ｔｈｉｓ　ｗｏｒｋ，　ｗｅ　ｈａｖｅ　ｄｏｎｅ　ａｎ　ｅｘｔｅｎｓｉｖｅ　ｓｕｒｖｅｙ　ｏｆ　ｆａｕｌｔ　ｔｏｌｅｒａｎｔ　ｐｒｏｔｏｃｏｌｓ　ａｎｄ　ａｎｔ　ｃｏｌｏｎｙ　ａｌｇｏｒｉｔｈｍｓ　ａｐｐｌｉｅｄ　ｔｏ　ｒｏｕｔｉｎｇ　ｉｎ　ＭＡＮＥＴｓ．Ｗｅ　ｐｒｏｐｏｓｅ　ａ　ＱｏＳ　ｃｏｎｓｔｒａｉｎｅｄ　ｆａｕｌｔ　ｔｏｌｅｒａｎｔ　ａｎｔ　ｌｏｏｋ—　ａｈｅａｄ　ｒｏｕｔｉｎｇ　ａｌｇｏｒｉｔｈｍ　ｗｈｉｃｈ　ａｔｔｅｍｐｔｓ　ｔｏ　ｉｄｅｎｔｉｆｙ　ｖａｌｉｄ　ｒｏｕｔｅ　ａｎｄ　１ｏｏｋ—ａｈｅａｄ　ｒｏｕｔｅ　ｐａｉｒｓ　ｗｈｉｃｈ　ｍｉｇｈｔ　ｈｅｌｐ　ｉｎ　ｃｈｏｏｓｉｎｇ　ｔｈｅ　ａｌｔｅｒｎａｔｅ　ｐａｔｈ　ｉｎ　ｃａｓｅ　ｏｆ　ｖａｌｉｄ　ｒｏｕｔｅ　ｆａｉｌｕｒｅ．Ｔｈｅ　ｒｅｓｕｌｔｓ　ｐｒｏｖｅ　ｔｈａｔ　ｔｈｅ　ｐｒｏｐｏｓｅｄ　ａｌｇｏｒｉｔｈｍ　ｔａｋｅｓ　ｂｅｔｔｅｒ　ｒｏｕｔｉｎｇ　ｄｅｃｉｓｉｏｎｓ　ｗｉｔｈ　２０－３０　ｐｅｒｃｅｎｔ　ｉｍｐｒｏｖｅｍｅｎｔ　ｃｏｍｐａｒｅｄ　ｗｉｔｈ　ｅｘｉｓｔｉｎｇ　ａｎｔ　ｃｏｌｏｎｙ　ａｌｇｏｒｉｔｈｍｓ．　

Ｋｅｙｗｏｒｄｓ：ｒｏｕｔｉｎｇ；ｆａｕｌｔ—ｔｏｌｅｒａｎｔ；ａｎｔ　ｃｏｌｏｎｙ　ａｌｇｏｒｉｔｈｍｓ；ＭＡＮＥＴｓ；ＱｏＳ；ｔｒｕｓｔ；ｒｅｐｕｔａｔｉｏｎ　

Ｉ－ＩＮＴＲＯＤＵＣＴＩＯＮ　Ｍｏｂｉｌｅ　Ａｄ　ＨＯＣ　Ｎｅｔｗｏｒｋｓ（ＭＡＮＥＴｓ）ｈａｖｅ　ｅｎｏｒｍｏｕｓ　ｎｅｔｗｏｒｋ　ｆｌｅｘｉｂｉｌｉｔｙ　ａｎｄ　ｎｏｄｅ　ｍｏｂｉｌｉｔｙ　ｗｈｉｃｈ　ｍａｋｅｓ　ＭＡＮＥＴ　ｒｏｕｔｉｎｇ　ｖｅｒｙ　ｃｈａｌｌｅｎｇｉｎｇ．Ｓｉｎｃｅ　ＭＡＮＥＴ　ａｓｓｕｍｅｓ　ａ　ｆｒｉｅｎｄｌｙ　ｅｎｖｉｒｏｎｍｅｎｔ．ｉｔ　ｉＳ　ｖｕｌｎｅｒａｂｌｅ　ｔｏ　ａｔｔａｃｋｓ　ａｎｄ　ｍａｌｉｃｉｏｕｓ　ｎｏｄｅｓ．Ｉｎ　ｓｕｃｈ　ａ　ｄｙｎａｍｉｃ　ｎｅｔｗｏｒｋ，　ｉｔ　ｉＳ　ｄｉｆｆｉｃｕｌｔ　ｔｏ　ｕｓｅ　ｍｕｌｔｉｍｅｄｉａ　ａｎｄ　ｏｔｈｅｒ　ａｄｖａｎｃｅｄ　ａｐｐｌｉｃａｔｉｏｎｓ　ｗｉｔｈｏｕｔ　ｑｕａｌｉｔｙ—ｏｆ＿　ｓｅｒｖｉｃｅ（ＱｏＳ）ｃｏｎｓｔｒａｉｎｔ．ＱｏＳ　ｃａｎ　ｂｅ　ｄｅｆｉｎｅｄ　ａｓ　ａ　ｓｅｔ　ｏｆ　ｓｅｒｖｉｃｅ　ｒｅｑｕｉｒｅｍｅｎｔ　ｔｈａｔ　ａ　ｇｉｖｅｎ　ｎｅｔｗｏｒｋ　ｓｈｏｕｌｄ　ｓａｔｉｓｆｙ　ｗｈｉｌｅ　ｔｒａｎｓｍｉｔｔｉｎｇ　ｐａｃｋｅｔｓ　ｆｒｏｍ　ｓｏｕｒｃｅ　ｔｏ　ｄｅｓｔｉｎａｔｉｏｎ．Ｉｔ　ｉｓ　ｄｉｆｆｉｃｕｌｔ　ｔｏ　ｄｅｓｉｇｎ　ａ　ｐａｔｈ　ｗｉｔｈ　ｍｕｌｔｉｐｌｅ　ＱｏＳ　ｃｏｎｓｔｒａｉｎｔｓ　ａｓ　ｔｈｅｒｅ　ｃａｎ　ｂｅ　ｃｏｎｆｌｉｃｔｉｎｇ　ｐａｒａｍｅｔｅｒｓ　ａｎｄ　ｒａｃｅ　ｃｏｎｄｉｔｉｏｎｓ　ａｍｏｎｇ　ｖａｒｉｏｕｓ　ｐａｒａｍｅｔｅｒｓ．　Ｍａｎｙ　ｏｆ　ｔｈｅ　ｐｏｐｕｌａｒ　ｐｒｏｔｏｃｏｌｓ　ｌｉｋｅ　ａｄ　ｈｏｃ　ｏｎ・ｄｅｍａｎｄ　ｄｉｓｔａｎｃｅ　ｖｅｃｔｏｒ（ＡＯＤＶ），ｄｙｎａｍｉｃ　ｓｏｕｒｃｅ　ｒｏｕｔｉｎｇ（ＤＳＲ、ａｎｄ　ｔｅｍｐｏｒａｒｙ　ｏｒｄｅｒｅｄ　ｒｏｕｔｉｎｇ　ａｌｇｏｒｉｔｈｍ（ＴＯＲＡ）ａｒｅ　ｄｅｓｉｇｎｅｄ　ｗｉｔｈｏｕｔ　ｃｏｎｓｉｄｅｒｉｎｇ　ＱｏＳ　ｃｏｎｓｔｒａｉｎｔｓ　ｆｏｒ　ｔｈｅ　ｐａｔｈ　ｔｈｅｙ　ｇｅｎｅｒｍｅ．Ｉｎ　ＭＡＮＥＴｓ，ｉｔ　ｉｓ　ｄｉｆｉｃｕｌｔ　ｔｏ　ｋｅｅｐ　ｕｐ—ｔｏ　ｄａｔｅ　ｉｎｆｏｒｍａｔｉｏｎ　ａｂｏｕｔ　ｔｈｅ　１ｉｎｋ　ｏｗｉｎｇ　ｔｏ　ｉｔｓ　ｄｙｎａｍｉｃ　ｎａｔｕｒｅ　ａｎｄ　ｄｅｐｌｅｔｉｏｎ　ｏｆ　ｅｎｅｒｇｙ　ａｔ　ｎｏｄｅ　ｗｈｉｃｈ　ｃａｕｓｅｓ　ｌｉｎｋ　ｂｒｅａｋａｇｅ．Ｉｎ　ａｄｄｉｔｉｏｎ，ｄｕｅ　ｔｏ　ｔｈｅ　ｍｏｂｉｌｉｔｙ　ａｎｄ　ｃｏｎｓｔａｎｔｌｙ　

９３　Ｃｈｉｎａ　Ｃｏｍｍｕｎｉｃａｔｉｏｎｓ・Ａｕｇｕｓｔ　２０　１　５　ｃｈａｎｇｉｎｇ　ｔｏｐｏｌｏｇｙ　ｏｆ　ｔｈｅ　ｍｏｂｉｌｅ　ａｄ　ｈｏｃ　ｎｅｔｗｏｒｋｓ，ｉｔ　ｉｓ　ｖｅｒｙ　ｄｉｆｆｉｃｕｌｔ　ｔｏ　ｖａｌｉｄａｔｅ　ａｌｌ　ｔｈｅ　ｒｏｕｔｅ　ｍｅｓｓａｇｅｓ．Ｄｅｖｅｌｏｐｉｎｇ　ａｎ　ｅｎｄ　ｔｏ　ｅｎｄ　ｓｅｃｕｒｅ　ｒｏｕｔｉｎｇ　ｐｒｏｔｏｃｏｌ　ｂｙ　ｅｎｃｒｙｐｔｉｏｎ　ａｎｄ　ｄｅｃｒｙｐｔｉｏｎ　ｔｅｃｈｎｉｑｕｅ　ｗｏｕｌｄ　ｓｏｌｖｅ　ｔｈｉｓ　ｉｓｓｕｅ．　Ｉｎ　ａｄｖｅｒｓｅ　ＭＡＮＥＴ　ｓｅｔｕｐ，ｂｏｔｈ　ｒｏｕｔｅ　ｅｓｔａｂｌｉｓｈｍｅｎｔ　ａｎｄ　ｄａｔａ　ｔｒａｎｓｍｉｓｓｉｏｎ　ａｒｅ　ｖｕｌｎｅｒａｂｌｅ　ｔｏ　ｖａｒｉｅｔｙ　ｏｆ　ａｔｔａｃｋｓ．Ｍｉｓｂｅｈａｖｉｎｇ　ｎｏｄｅｓ　ｃｏｕｌｄ　ｄｉｓｔｕｒｂ　ｒｏｕｔｅ　ｄｉｓｃｏｖｅｒｙ　ｂｙ　ｉｍｐｅｒｓｏｎａｔｉｏｎ，ｏｒ　ｂｙ　ｒｅｓｐｏｎｄｉｎｇ　ｗｉｔｈ　ｆａｌｓｅ　ｒｏｕｔｅ　ｉｎｆｏｒｍａｔｉｏｎ．Ｔｈｉｓ　ｍａｙ　ｉｎｄｉｒｅｃｔｌｙ　ｔｒａｎｓｆｅｒ　ｔｈｅ　ｅｎｔｉｒｅ　ｎｅｔｗｏｒｋ　ｃｏｎｔｒｏｌ　ｉｎｔｏ　ｔｈｅ　ｈａｎｄｓ　ｏｆ　ｉｎｔｒｕｄｅｒｓ．Ｔｈｅｒｅｆｏｒｅ，ｉｎ　ｏｒｄｅｒ　ｔｏ　ｐｒｏｖｉｄｅ　ｃｏｍｐｌｅｔｅ　ｓｅｃｕｒｉｔｙ，ｐｒｏｔｏｃｏｌｓ　ｍｉｇｈｔ　ｂｅ　ｈｅｌｐｆｕ１．　Ｒｅｌｉａｂｌｅ　ｔｒａｎｓｐｏｒｔ　ｐｒｏｔｏｃｏｌｓ　ａｒｅ　ｐｌａｉｎｌｙ　ｉｎｓｕｆｆｉｃｉｅｎｔ　ｔｏ　ｓｅｒｖｅ　ｔｈｅ　ａｂｏｖｅ　ｐｕｒｐｏｓｅ．Ｔｈｅ　ａｔｔａｃｋ　ａｎｄ　ｔｈｅ　ａｆｔｅｒｍａｔｈ　ｉｓ　ｆａｒ　ｂｅｙｏｎｄ　ｔｈｅ　ｌｉｍｉｔ　ｏｆ　ｓｕｃｈ　ｐｒｏｔｏｃｏｌｓ．　Ｉｎ　ｔｈｉｓ　ｐａｐｅｒ　ｗｅ　ｐｒｏｐｏｓｅ　ｔｏ　ｄｅｖｅｌｏｐ　ａ　ｆａｕｌｔ　ｔｏｌｅｒａｎｔ　ｒｏｕｔｉｎｇ　ａｌｇｏｒｉｔｈｍ　ｂａｓｅｄ　ｏｎ　ａｎｔ　ｃｏｌｏｎｙ　ｏｐｔｉｍｉｓａｔｉｏｎ　ｗｉｔｈ　ｌｏｏｋ—ａｈｅａｄ　ｐｒｏｐｅｒｔｙ．Ｔｈｅ　ａｎｔ—ｌｉｋｅ　ａｇｅｎｔｓ　ｃａｌｌｅｄ　ｆｏｒｗａｒｄ　ａｎｔｓ（ＦＡＮＴ）　ａｎｄ　ｂａｃｋｗａｒｄ　ａｎｔｓ（ＢＡＮＴ）ａｒｅ　ｕｓｅｄ　ｔｏ　ｍｅａｓｕｒｅ　ｖａｒｉｏｕｓ　ＱｏＳ　ｐａｒａｍｅｔｅｒｓ　１ｉｋｅ　ｎｅｘｔ　ｈｏｐ　ａｖａｉｌａｂｉｌｉｔｙ（ＮＨＡ），ｄｅｌａｙ　ａｎｄ　ｂａｎｄｗｉｄｔｈ　ａｓ　ｐａｒａｍｅｔｅｒｓ　ｆｏｒ　ｓａｒｉｓｆｙｉｎｇ　ＱｏＳ　ｃｏｎｓｔｒａｉｎｔｓ．　Ｕｓｉｎｇ　ｔｈｅｓｅ　ｐａｒａｍｅｔｅｒｓ　ｗｅ　ｃａｌｃｕｌａｔｅ　ｐａｔｈ　ｐｒｅｆｅｒｅｎｃｅ　ｐｒｏｂａｂｉｌｉｔｙ，ｔｈａｔ　ｉｓ，ｔｈｅ　ｐｒｏｂａｂｉｌｉｔｙ　ｔｏ　ｓｅｌｅｃｔ　ａ　ｐａｒｔｉｃｕｌａｒ　ｐａｔｈ　ｆｒｏｍ　ａｌｌ　ｔｈｅ　ａｖａｉｌａｂｌｅ　ｐａｔｈｓ．Ｐａｔｈ　ｗｉｔｈ　ｈｉｇｈｅｒ　ｐａｔｈ　ｐｒｅｆｅｒｅｎｃｅ　ｐｒｏｂａｂｉｌｉｔｙ　ｂｅｔｗｅｅｎ　ｓｏｕｒｃｅ　ａｎｄ　ｄｅｓｔｉｎａｔｉｏｎ　ｇｅｔｓ　ｓｅｌｅｃｔｅｄ　ｆｏｒ　ｔｒａｎｓｍｉｔｔｉｎｇ　ｄａｔａ．Ｔｈｉｓ　ｗａｙ，　ｔｈｅ　ａｌｇｏｒｉｔｈｍ　ｆｉｎｄｓ　ｆａｕｌｔ－ｐｒｏｎｅ　ｎｏｄｅｓ　ｄｕｒｉｎｇ　ｒｏｕｔｅ　ｄｉｓｃｏｖｅｒｙ　ｐｈａｓｅ　ｓｏ　ｔｈａｔ　ｉｔ　ｃａｎ　ｓｋｉｐ　ｔｈｅｍ　ｆｒｏｍ　ｓｅｌｅｃｔｅｄ　ｐａｔｈ　ｔｏ　ｓａｒｉｓｆｙ　ｔｈｅ　ｇｉｖｅｎ　ＱｏＳ　ｒｅｑｕｉｒｅｍｅｎｔ．　ｎ．ＲＥＬＡＴＥＤ　ＷＯＲＫ　２．１　Ｏｖｅｒｖｉｅｗ　ｏｆ　ｍａｎｅｔ　ｒｏｕｔｉｎｇ　ａｌｇｏｒｉｔｈｍｓ　Ｓｏｕｒｃｅ　ｉｎｉｔｉａｔｅｄ　ｒｏｕｔｉｎｇ　ｉｓ　ｔｈｅ　ｓｉｍｐｌｅｓｔ　ａｍｏｎｇ　ｔｈｅ　ｅｘｉｓｔｉｎｇ　ＭＡＮＥＴ　ｒｏｕｔｉｎｇ　ｐｒｏｔｏｃｏｌｓ．　Ｔｈｅｓｅ　ｐｒｏｔｏｃｏｌｓ　ｃｏｎｓｉｓｔ　ｏｆ　ｔｗｏ　ｐｈａｓｅｓ：ｒｏｕｔｅ　ｄｉｓｃｏｖｅｒｙ　ａｎｄ　ｒｏｕｔｅ　ｍａｉｎｔｅｎａｎｃｅ．Ｕｐｏｎ　ｔｈｅ　ｒｅｑｕｅｓｔ　ｏｆ　ｓｏｕｒｃｅ　ｎｏｄｅ　ａｎｄ　ｈｅｎｃｅ　ｏｎ—　ｄｅｍａｎｄ，ｔｈｅ　ｎｅｉｇｈｂｏｕｒｉｎｇ　ｎｏｄｅｓ　ｄｏ　ｒｏｕｔｅ　ｄｉｓｃｏｖｅｒｙ．Ｔｈｉｓ　ｍｉｇｈｔ　ｅｎｄ　ｕｐ　ｗｉｔｈ　ｍｕｌｔｉｐｌｅ　ｖａｌｉｄ　ｒｏｕｔｅｓ　ｆｒｏｍ　ｓｏｕｒｃｅ　ｔｏ　ｄｅｓｔｉｎａｔｉｏｎ，ａｎｄ　ｔｈｅｒｅｆｏｒｅ　ｃｈｏｓｉｎｇ　ｏｎｅ　ａｍｏｎｇ　ｔｈｅｍ　ｖａｒｉｅｓ　ｏｎｅ　ｐｒｏｔｏｃｏｌ　ｆｒｏｍ　ａｎｏｔｈｅｒ．Ｈｏｗｅｖｅｒ，ｉｆ　ｔｈｅ　ｒｏｕｔｅｓ　ｈａｖｅ　ｌｏｏｐ（ｓ）ｉｎ　ｂｅｔｗｅｅｎ，ｓｉｍｐｌｅ　ｏｎ－ｄｅｍａｎｄ　ＭＡＮＥＴ　ｒｏｕｔｉｎｇ　ｐｒｏｔｏｃｏｌｓ　ｓｉｍｐｌｙ　ｉｇｎｏｒｅ　ｔｈｅ　ｒｏｕｔｅｓ　ｃｏｎｔａｉｎｉｎｇ　ｌｏｏｐｓ　ａｎｄ　ｏｐｔ　ｆｏｒ　ｏｔｈｅｒ　ｌｏｏｐ　ｆｒｅｅ　ｒｏｕｔｅｓ．　Ｔｈｅ　ｃｌａｓｓｉｃａｌ　Ａｄ　Ｈｏｃ　Ｄｉｓｔａｎｃｅ　Ｖｅｃｔｏｒ　（ＡＯＤＶ）ｒｏｕｔｉｎｇ　ｐｒｏｔｏｃｏｌ［１］ｕｓｅｓ　ｔｈｅ　ｄｅｓｔｉｎａｔｉｏｎ　ｓｅｑｕｅｎｃｅ　ｎｕｍｂｅｒ　ｆｏｒ　ｅａｃｈ　ｒｏｕｔｅ　ｅｎｔｒｙ．Ｂｕｔ　ｉｔ　ｓｕｆｆｅｒｓ　ｆｒｏｍ　ｔｈｅ　ｄｉｓａｄｖａｎｔａｇｅ　ｏｆ　ｓｔａｌｅ　ｅｎｔｒｉｅｓ　ｃａｕｓｉｎｇ　ｔｈｅ　ｒｏｕｔｅ　ｔｏ　ｂｅ　ｉｎｃｏｎｓｉｓｔｅｎｔ．ＡＯＤＶ－ＢＲ［２］ａｎｄ　Ａ０ＤＶ－　ＡＢＲ［３］ｕｓｅｓ　ｂａｃｋｕｐ　ａｎｄ　ａｄａｐｔｉｖｅ　ｂａｃｋｕｐ　ｒｏｕｔｅｓ　ｒｅｓｐｅｃｔｉｖｅｌｙ　ｉｎ　ｃａｓｅ　ｏｆ　ｒｏｕｔｅ　ｆａｉｌｕｒｅｓ．　Ｄｙｎａｍｉｃ　ｉｄｅｎｔｉｆｉｃａｔｉｏｎ　ｏｆ　ｂａｃｋ－ｕｐ　ｒｏｕｔｅｓ　ｗｏｕｌｄ　ｉｎｖｏｌｖｅ　ｍｏｒｅ　ｃｏｓｔ［４］．Ｌｏｗ—ｏｖｅｒｈｅａｄ　ｄｙｎａｍｉｃ　ｒｏｕｔｅ—ｒｅｐａｉｒｉｎｇ　ｐｒｏｔｏｃｏｌ［５］ａｌｓｏ　ｄｅａｌｓ　ｗｉｔｈ　ｌｉｎｋ　ｆａｉｌｕｒｅｓ．ＬＢＡＱ［６］ｄｅａｌｓ　ｗｉｔｈ　ｌｉｎｋ　ｓｔａｔｅ　ｄｅｔｅｒｍｉｎａｔｉｏｎ　ｂｕｔ　ｓｅｌｅｃｔｓ　ｏｎｌｙ　ｌｉｎｋ　ｔｈａｔ　ｈａｖｅ　ｍａｘｉｍｕｍ　ＱｏＳ　ｑｕａｌｉｔｙ．Ｍｕｌｔｉ・ｐａｔｈ　ｒｏｕｔｉｎｇ　ｉｓ　ｓｅｎｓｉｂｌｅ　ｉｎ　ｃａｓｅ　ｏｆ　ｌｉｎｋ　ｆａｉｌｕｒｅｓ．ＭＰ・　ＡＯＤＶ［７］ｉｓ　ａ　ｐｌａｉｎ　ｍｕｌｔｉ－ｐａｔｈ　ｖｅｒｓｉｏｎ　ｏｆ　ｅｘｉｓｔｉｎｇ　ＡＯＤＶ．Ｈｅｒｅ，ｏｎｅ　ｍａｉｎ　ｐａｔｈ　ａｎｄ　ｏｎｅ　ｂａｃｋ—ｕｐ　ｐａｔｈ　ｔｏ　ｄｅｓｔｉｎａｔｉｏｎ　ａｒｅ　ｆｏｕｎｄ　ｄｕｒｉｎｇ　ｒｏｕｔｅ　ｄｉｓｃｏｖｅｒｙ．Ｈｏｗｅｖｅｒ，ｔｈｅ　ｐｒｏｔｏｃｏｌ　ｄｏｅｓ　ｎｏｔ　ｈａｎｄｌｅ　ｔｈｅ　ｒｏｕｔｉｎｇ　ｄｅｃｉｓｉｏｎｓ　ｗｉｔｈ　ｒｅｄｕｃｅｄ　ｏｖｅｒｈｅａｄ．Ｉｎ　ａｄｄｉｔｉｏｎ，ｏｔｈｅｒ　ａｌｔｅｒｎａｔｅ　ｎｏｄｅ　ｄｉｓｊｏｉｎｔ　ｐａｔｈ　ｔｏ　ｄｅｓｔｉｎａｔｉｏｎ　ｉｓ　ｎｏｔ　ａｔｔｅｍｐｔｅｄ．　Ｔｈｅｒｅｆｏｒｅ，ｉｎ　ｃａｓｅ　ｏｆ　ｌｉｎｋ　ｆａｉｌｕｒｅｓ　ｏｎ　ｂａｃｋｕｐ　ｒｏｕｔｅｓ，ｔｈｅ　ｐｒｏｔｏｃｏｌ　ｈａｓ　ｔｏ　ｓｔａｒｔ　ａｇａｉｎ　ｉｔｓ　ｒｏｕｔｅ　ｄｉｓｃｏｖｅｒｙ　ｐｒｏｃｅｓｓ　ｆｒｏｍ　ｔｈｅ　ｆｉｒｓｔ　ｓｔａｇｅ．　ＮＤＭ—ＡＯＤＶ［８］ｉｓ　ａｎ　ｉｍｐｒｏｖｅｍｅｎｔ　ｔｏ　ＭＰ—ＡＯＤＶ　ｗｈｅｒｅ　ａｌｌ　ｔｈｅ　ｎｏｄｅ　ｄｉｓｊｏｉｎｔ　ｐａｔｈｓ　ａｒｅ　ｆｏｕｎｄ　ａｎｄ　ｓｔｏｒｅｄ　ｉｎ　ｒｏｕｔｉｎｇ　ｔａｂｌｅｓ．Ｂｕｔ　ｔｈｉｓ　ａｃｃｕｍｕｌａｔｅｓ　ｔｈｅ　ｎｏｄｅｓ　ａｎｄ　ｔｈｅｒｅｆｏｒｅ　ｒｅｄｕｃｅｓ　ｔｈｅ　ｐｅｒｆｏｒｍａｎｃｅ　ａｓ　ｔｈｅ　ｎｅｔｗｏｒｋ　ｇｒｏｗｓ　ｏｒ　ｔｈｅ　ｆｒｅｑｕｅｎｃｙ　ｏｆ　ｄｙｎａｍｉｓｍ　ｉｎｃｒｅａｓｅｓ．ＡＯＭＤＶ　［９］，ＡＯＤＶＭ［１０］ａｎｄＡＯＤＶＭ—ＰＥＳ［１１］ａｌｓｏ　ｆｉｎｄ　ｍｕｌｔｉｐｌｅ　ｐａｔｈｓ　ｂｕｔ　ｉｎｓｔｅａｄ　ｏｆ　ｎｏｄｅ　ｄｉｓｊｏｉｎｔ　ｐａｔｈｓ；ｔｈｅｓｅ　ｐｒｏｔｏｃｏｌｓ　ｆｉｎｄ　ｔｈｅ　ｌｉｎｋ　ｄｉｓｊｏｉｎｔ　Ｄａｔｈｓ　ｔｏ　ｔｈｅ　ｄｅｓｔｉｎａｔｉｏｎ．Ｉｎ　ａｌｌ　ｔｈｅｓｅ　ｖａｒｉａｔｉｏｎｓ　０ｆ　Ａ０ＤＶ．ｄａｔａ　ｔｒａｎｓｆｅｒ　ｓｔａｒｔｓ　ｏｎｌｙ　ａｆｔｅｒ　ａｌｌ　ｔｈｅ　ｐａｔｈｓ　ａｒｅ　ｄｉｓｃｏｖｅｒｅｄ．ＮＤＭＰ－Ａ０ＤＶ［１２】　ｉｓ　ａｎｃＩｔｈｅｒ　ａｌｔｅｒｎａｔｅ　ａｐｐｒｏａｃｈ　ｔｏ　Ａ０Ｄ　Ｖ．Ｈｅｒｅ　ｔｈｅ　ｓｏｕｒｃｅ　ｎｏｄｅ　ｓｔａｒｔｓ　ｔｒａｎｓｍｉｓｓｉｏｎ　ｓｏｏｎ　ａｆｔｅｒ　ｔｈｅ　ｆｉｒｓｔ　ｎｏｄｅ．ｄｉ￣ｏｉｎｔ　ｒｏｕｔｅ　ｔｏ　ｄｅｓｔｉｎａｔｉｏｎ　ｉｓ　