Software-DefinedMultilayeredAdmissionControlforQualityof …downloads.hindawi.com/journals/wcmc/2020/2989751.pdf · 2020-02-23 · 1.Introduction MobileAd-hocNetworks(MANET)formsaself-organized,

Research ArticleSoftware-Defined Multilayered Admission Control for Quality ofService Assurance in Mobile Ad-hoc Networks

Ponraj Appandairaj 1 and Kathiravan Kannan2

1Department of Electronics and Communication Engineering Easwari Engineering College Chennai 600089 India2Easwari Engineering College Chennai 600089 India

Correspondence should be addressed to Ponraj Appandairaj ponrajaa1gmailcom

Received 26 July 2019 Revised 19 October 2019 Accepted 9 November 2019 Published 11 January 2020

Academic Editor Tomas J Mateo Sanguino

Copyright copy 2020 Ponraj Appandairaj and Kathiravan Kannan +is is an open access article distributed under the CreativeCommons Attribution License which permits unrestricted use distribution and reproduction in any medium provided theoriginal work is properly cited

Mobile Ad-hoc Network has emerged as a key technology for next-generation networks +ough its rapid growth inspiresnumerous applications it is difficult to assure Quality of Service because of its immense scaling caused due to nodersquos mobilityfading radio signals and unreliable nature of the wireless channel To efficiently utilize network resources and accomplishguaranteed Quality of Service a novel Software-Defined Multilayered Admission Control model that embeds an intelligentNeurofuzzy Inference-based Admission Control service engine is proposed in this paper Each node makes use of the NeurofuzzyInference-based Admission Control service to learn manage prioritize and admit data traffic according to user requirement +eservice engine exploits fuzzy inference-based admission control process to assess nodersquos current status using Quality of Serviceparameters namely bandwidth queue load and Received Signal Strength Indicator to evaluate the prediction index +eprediction index not only helps in determining the strongly connected neighbors during reliable path selection process but alsosolely decides whether the admission control session can be admitted or rejected Moreover the Neuro-Multilayered Learningprocess of the service engine helps to self-organize and make the complete network intelligent for instantaneous decision making+e proposed mechanism not only improves the session admission between nodes but also reduces the packet drops assuringsuccessful session completion Performance analysis using the simulation model proves that the proposed system showspromising gains with assured throughput and low end-to-end delay and has the potential to be applied in real-world scenarios

1 Introduction

Mobile Ad-hoc Networks (MANET) forms a self-organizedinfrastructure-less and rapidly deployable wireless network[1] It is constructed by a set of mobile nodes that offersunique advantage of being created and used ldquoanywhereanytimerdquo but not relying on the complex infrastructure+ough MANET undergoes rapid progress inspiring nu-merous applications enabling multimedia communicationin mobile ad-hoc networks still remains a challenging taskPredominantly most real-time video and audio applicationsmakes use of stringent bandwidth typically requiring thenetwork to support such challenging and varying traffic rateswith high throughput and low end-to-end delay to guaranteeQuality of Service (QoS) [2ndash5]

Despite the efforts made to alleviate many issues [6]there still exist a number of barriers to the deployment ofreal-time applications in MANET +e most prominentones include maintaining connection state rapid alloca-tion of resources avoid scarcity of resources and effec-tively manage dynamic topological changes caused due tomobility and appropriate resource reservation between thesource and destination +ese challenges [7ndash9] has initi-ated the need to design and develop new QoS techniques[10ndash13] to support real-time data services in MANETFurthermore the new approach should be capable ofdifferentiating various service requirements taking intoconsideration all other essential elements carefully withan ultimate goal to guarantee best end-to-end servicedelivery

HindawiWireless Communications and Mobile ComputingVolume 2020 Article ID 2989751 23 pageshttpsdoiorg10115520202989751

+e Admission Control (AC) [14ndash17] is considered asthe most crucial mechanism for providing QoS in MANETIts core functionality is to estimate the nodersquos state based onresource availability and decide which applicationrsquos dataflow can be admitted along which node such that the QoS asper user requirement is met without violating previouslymade guarantees

To combat the aforementioned challenges and developan enhanced technique to effectively manage network re-sources for assuring QoS in MANET in this paper novelneurofuzzy logic-based Software-Defined MultilayeredAdmission Control (SD_MAC) model is proposed Aim ofthe proposed model is to assure guaranteed QoS support forreal-time traffic It exploits the fuzzy logic to ensure trafficadaptation characterized by dynamic changes (making itadequate to the uncertainty and the information incom-pleteness) in MANET environment +e model also incor-porates the benefits of Neural Networks (NN) to support themultimedia service in ad-hoc networks +e self-adaptivedata driven mechanism of the neural network allows fastlearning of different operations performed by the corefunctional blocks of the system thereby permitting a besttraffic class selection and delivery according to flow re-quirements After learning the data presented to them itoffers the capability to generalize ie it often predicts anevent occurrence even if the input data is noisy and theirunderlying relationships are unknown or difficult to de-scribe +e most important feature of the NN is that it canapproximate multivariate functions with superior predictionaccuracy and low computational complexity It can pri-marily be used when limited data sets are available and therelationships between them are vague Overall NN exploitsthe capability to learn and adapt to the system behaviourmore quickly by integrating the operations of variouscomponents (both in MAC and network layer) to make thenetwork self-organized and artificially intelligent

+e proposed SD_MAC system makes use of theNeurofuzzy Inference-based Admission Control (NFAC)service engine whose core functionality is to

(i) Evaluate the resource availability by using QoSparameters such as bandwidth queue load andRSSI in each node to generate the node status+e Fuzzy Inference-based Admission Control(FIAC) module in the NFAC engine is triggeredautomatically in each node to generate the nodestatus

(ii) Derive the prediction index which solely decideswhether the admission control session can be ad-mitted or rejected via the node for the requestedQoS

(iii) Self-learn the behavioural functionalities and trainand test the network to make it self-organized andartificially intelligent for effective decision makingusing Neuro-Multilayered Learning (NML) process

Our proposed methodology was modelled using pro-totype developed using MATLAB +e performance of theproposed model is evaluated by comparing it with other

existing approaches such as SWAN StAC DACME andFuzzyQoS +e results showed that the service exhibited byNFAC engine tends to automate the working procedure ofadmission control to achieve good throughput by selectingstrongly connected neighbors and reliable path for longerlive transmission

+e rest of the document is organized as follows Section2 presents related work Section 3 illustrates our proposedSD_MAC system Section 4 depicts the simulation resultsand analysis and finally Section 5 concludes the paper alongwith a discussion on future directions

2 Related Works

+is section describes briefly some of the main admissioncontrol model proposed for wireless MANET

A stateless service differentiation AC model (SWAN)[18] performs real-time traffic control using sender-basedadmission control by distinguishing real-time and best-ef-fort traffic +e source probes the path to the destinationwhen it wants to send a real-time traffic to destinationDuring the probe it identifies the available bandwidth andrelies on feedback received from the MAC layer by using therate and source-based admission control approach Eachnode uses the rate control algorithm to control the best-effort traffic +e bandwidth used for best-effort traffic isrestricted so that real-time applications can exploit the re-quired bandwidth the bandwidth not used by real-timeapplications can be exploited by the best-effort traffic +eprobe request in SWAN causes a lot of overhead and packetloss

In Staggered Admission Control protocol (StAC) [19]the source sends the session request packet +e session isrejected if the network layer does not support the flow else itbroadcasts the request Intermediate nodes add informationto request after verifying their local resources +e desti-nation receives the request and sends the response Sourcecaches all the route information and then sends an ad-mission request packet to two hop neighbors which sends anadmission denied back if it cannot support new informationflow Based on the observation the source transmits data at avery low rate and gradually increases the flow rate up to therequired level within the specified time During the gradualincreasing time if it is affecting the earlier admitted sessionsthen a session may be rejected Presence of interferencecauses link failures resulting in frequent route discoveries+is deteriorates the performance in StAC by incurring highrouting overheads and end-to-end delay

A distributed admission control for MANET environ-ments (DACME) [20] approach periodically assesses end-to-end conditions on the path Source node performs pathprobing to obtain various QoS measurements to assuretransmission of data traffic under good conditions +edestination responds to probes giving the source feedbackabout path conditions+ough this method tends to performwell in small to medium network taking advantage of theIEEE 80211e standard [21] it suffers from fairness issueunder high network load

2 Wireless Communications and Mobile Computing

+e FuzzyQoS [22] approach aims to improve the trafficregulation rate and congestion control of multimedia ap-plications It monitors the variation rate in queue to providea measure of the queue state and uses fuzzy thresholds toadapt the traffic transmission rate under dynamic condi-tions It makes the source node more responsive to anysudden changes in network traffic volume by using explicitrate congestion notification +ough this approach achievesstable end-to-end delay under different network conditionsit does not deal well with route failures It reduces the datarate of affected session while searching for new routes im-plying its support for real-time applications only with elasticthroughput requirements

To increase the userrsquos perceived QoS we apply an NFACservice mechanism in the SD_MAC model Detailed illus-tration of the proposed approach is elaborated in Section 3

3 System Description

31 Overview of the SD_MAC Model +e proposed Soft-ware-Defined Multilayered Admission Control (SD_MAC)model exploits a new cross-layer QoS solution for mobile ad-hoc networks +e overall architecture of the proposedSD_MAC model comprises of various functional blockssuch as routing service admission controller classifierscheduler and traffic regulator

Figure 1 illustrates the schematic representation of theSD_MAC model with its core functional blocks

+e Admission Controller module is considered as themost important functional block of the SD_MAC model Itscore functionality is to send a dedicated probe request from thesource node towards the destination node to detect mostreliable route by estimating intermediate nodersquos resourcesavailability Based on the information the admission controlmodule at the source decides whether to admit or reject thedata service request In the proposed model each node ini-tiates ldquoadmission controlrdquo process to evaluate its current re-source availability which is indicated using the ldquoservice offeredprediction index SOfdprediction_indexrdquo value Every node storesthe SOfdprediction_index value in its Local Aware Table (LAT)Next using the ldquoRouting Servicerdquo module each node broad-casts a beacon message (consisting of the signal qualities typeof service offered (SOfdprediction_index) etc) over the network+e neighbor nodes receive the beacon message and store thesenderrsquos id service offered (SOfdprediction_index) by the senderetc in its LAT Now let us assume that a source node has datato be transmitted to a destination node in the network Let usconsider that a reliable route from the source to the destinationdoes not exist and then during such occasions the proposedSD_MAC model performs the following steps

(i) +e source initiates admission control mechanismwithin itself to generate service-requested predic-tion index (SReqprediction_index) value

(ii) +e source then performs the ldquopath selectionrdquoprocess to reach the destination As part of theldquopath selectionrdquo process the source and the neighbornodes looks-up its LAT to find stronglyconnected neighbors SCN (nodes that have

SOfdprediction_indexgt SReqprediction_index) that has thecapability to process the data traffic as per userrsquosQoS requirement

(iii) +e source then sends a dedicated probe request(DPReq) message only to the selected SCNs which inturn repeats the process until destination is reached

Unlike other admission control techniques the sourcenode in the proposed model does not execute admissioncontrol process every time to select the strong neighborsrather it simply looks up its LAT to identify the SCNs(each node maintains the list of neighbor nodes with itsldquoprediction indexrdquo value in its LAT) +is mechanism ofvalidating nodersquos resource availability using ldquopredictionindexrdquo prevents the node from executing the ldquoadmissioncontrolrdquo process every time a resource verification requestarrives from other nodes In addition as every nodemaintains the list of neighbor nodes prediction index in itsLAT the neighbour which has the capability to processthe request is identified instantaneously during pathdiscovery Moreover to optimize traffic regulation andcontrol congestion over the network for both real-time(audiovideo) and nonreal-time services the proposedmodel embeds a Neurofuzzy Inference-based AdmissionControl (NFAC) service engine to exploit fuzzy inferenceapproach integrated with neural artificial intelligence +eaim is to guarantee best end-to-end service delivery for endusers +ough NFACrsquos core objective is to learn manageprioritize and admit data traffic according to userrsquos QoSrequirement it also keeps the network trained and testedon various behavioural changes making it self-organizedand artificially intelligent for effective decision makingover a period of time Significantly this not only improvesthe service flow acceptance between nodes but also reducesthe dropping of service among them +e other functionalblocks of the SD_MAC model such as the classifier trafficregulator and the scheduler are specifically used to dis-tinguish regulate and schedule data traffic appropriatelyWhenever a data service request is admitted the classifierprovides service differentiation by classifying and markingthe packets into 4 different classes It marks the packetseither as High Profiled (HP) or Medium Profiled (MP) orLow Profiled (LP) +e marked packets are then queued intheir respective HP or MP or LP queues and wait for thetransmission +e main objective of the classifier is todistinguish the traffic and determine the ones that shouldpass or bypass the Traffic Regulator +e Traffic Regulatormodule regulates MP and LP traffic by delaying MP and LPpacket processing and allows more bandwidth for HPpacket processing When a new real-time flow is allowedby admission control block the packets marked as HP bypasses the traffic regulator and are queued in HP queuesfor processing +e Scheduler module employs a newscheduling discipline referred to as the Strict and Prob-abilistic Priority Queuing (SPPQ) which determines theprobability with which the queue is served by assigning aparameter to each priority queue It offers variety of servicedifferentiation by providing service segregation amongvarious groups of traffic classes

Wireless Communications and Mobile Computing 3

+e functional blocks or components that belong to theSD_MAC system are as follows

Routing Service it discovers legitimate path between thesource and the destination It performs

(i) Path Selection(ii) Resource Reservation

Admission controller the Admission controller in theSD_MAC system incorporates a Neurofuzzy Inference-based Admission Control (NFAC) service engine to

(i) Evaluate and maintain the node status informationfor each node

(ii) Derive the certainity factor or prediction indexusing the node status information to solely decidewhether the admission control session can be ad-mitted or rejected

Primary functionality of NFAC is to perform the fol-lowing activities

(i) Fuzzy Inference-based Admission Control Process(ii) Neuro-Multilayered Learning Process

Classifier it classifies and stamps packets based on theirpriority (as platinum gold silver and bronze) when anew request is admitted by the admission controller

Scheduler it uses the scheduling algorithm in order topush the marked packets in their respective high- medium- and low-profiled queues for transmissionTraffic Regulator it regulates data traffic in MP and LPqueues if there is delay encountered in processing HPpackets and vice versa

+e summary of notations used in the SD_MAC systemis referred in Table 1

32 Detailed Specification of Software-Defined MultilayeredAdmission Control System Our work primarily focuses onRouting Service and Admission Control component

321 Routing Service To realize service access anytime andanywhere the path between the source and destinationshould be strongly connected SD_MAC follows the Ad-hocon Demand Distance Vector Routing (AODV) protocol forpath establishment It uses Signal Strength (SS) metric tomake sure that the data traffic service is performed only onreliable links [23] rather than on weak diminishing links (ieselecting ldquoreliablerdquo links ensures that the path between thesource and the destination is strongly connected) +eldquorouting servicerdquo component performs key operation such aspath selection and resource reservation

PlatinumGoldSilver

BronzePacketclassification HPMPLP

Dedicated probing requestIP

layer

MAC

Classifier

User request

MAC layer

Admitreject flow admission

Request for flow admission

Routing service

Neurofuzzy inference based admission control(NFAC)

QoS resource reservation-node state

Neuro-multilayered learning(NML)

Testing Training

Admission controllerSOfdprediction_index SReqprediction_index

Physical layerShared media channel RSSI

Traffic regulator

Fuzzy inference-basedadmission

control (FIAC)

Defuzzification

Fuzzy inference engine

Fuzzification

Fuzzy rule base

Nstatus

SOfdprediction_index

Scheduler

Figure 1 Model of SD_MAC


322 Path Selection +e path selection is performed beforethe source Sn transmits data to the destination Dn +eprocess involves the following steps

Step 1 at regular intervals each node broadcasts theldquoP_beaconrdquo UDP packet to its immediate neighborsannouncing the link qualities and services provided+e ldquoP_beaconrdquo packet contains the followinginformation

P_beacon

Nid NRSSI NSNR NTS SOfdprediction_index1113960 1113961(1)

where Nid indicates the node id NRSSI indicates thenodersquos Received Signal Strength Indicator (RSSI) NSNRindicates the nodersquos Signal-to-Noise Ratio (SNR) NTSindicates the nodersquos time stamp and SOfdprediction_indexindicates the nodersquos service-offered prediction indexvalueStep 2 the neighbor nodes receive the P_beacon packetextracts the sender nodersquos information and adds theentry in its Local Aware Table (LAT) Each nodemaintains a LAT which contains the list of its one-hopneighbor information Entry in LAT has the followingfields

LATfields NHidNHAvRSSINHAvSNR1113858

NHTS SOfdprediction_index1113961(2)

Each node generates and maintains a service-offered(SOfdprediction_index) prediction index which indicatesthe type of service offered by neighbor nodes +eSOfdprediction_index derived using NFAC process iselaborated in detail in Section 33Step 3 in a similar way all the nodes in the networkshare among each other their signal qualities andservices offered using the ldquoP_beaconrdquo packet which in

turn is added and maintained in their respective LATNow using the information in LAT each node has thecapability to select its reliable neighbor nodes +isensures each node to selectively send service queriesonly to strongly connected neighbors (SCNs) whichmay in turn forward service queries to their SCNsStep 4 when a source node has data service (DS) to betransmitted to a destination and if it does not know aldquoreliable routerdquo to reach the destination it first looks upinto its LAT to verify if there are SCNs which offers therequired service Typically the data service consists ofthe following information (i) the type of data trafficTRType and (ii) measure of data traffic MDataUsing the information of DS such as TRType and MDataand the node status (Nstatus) the node derives aSReqprediction_index value where Nstatus indicates currentnodersquos status which is evaluated using FIACrsquos process ofNFAC and SReqprediction_index indicates the type of dataservice to be processed by the source which is evaluatedusing NMLrsquos process of NFAC Details regardingNstatusand SReqprediction_index are elaborated in detail inSection 33Step 5 the source looks up its LAT to verify if thereare neighbors which satisfies the criteriaSOfdprediction_indexge SReqprediction_index If there arenodes that satisfy the criteria then those nodes areconsidered as strongly connected neighbors (SCN)+esource then selectively sends a dedicated probing re-quest packet DPReq (Saddr Daddr Seqid Servicepath andSReqprediction_index) only to those SCNs +e fields SaddrDaddr and Seqid indicate the source address destinationaddress and sequence identifier +e Servicepath refersto the service path with first address filled with its ownaddress and the address of SCNs along with its signalstrength During the LAT look-up process the fol-lowing is performed

(i) If only one SCN satisfies the criteria then theDPReq packet is sent only to the SCN found

(ii) If multiple SCNs satisfy the criteria then theDPReq packet is send to ldquokrdquo SCNs whoseSOfdprediction_index value is high compared toothers Let ldquokrdquo indicate the optimal number ofpaths selected for data transmission Our scenarioconsiders ldquok 3rdquo ie maximum of three reliablepaths are selected during the ldquopath selectionrdquomechanism such that when a path discoveredfails during data transmission the next path canbe considered for transmission In cases where allthree routes selected fails during data transmis-sion the process of ldquopath selectionrdquo is initiatedonce again An optimal limit of three paths hasbeen chosen in our work as too many paths canincrease the complexity during maintenance

(iii) If no SCN satisfies the criteria then the node waitsfor a time period and then performs the LAT lookup once again+e time slot is not a fixed value andkeeps varying as it depends on the nodersquos LAT

Table 1 Summary of notations used in the SD_MAC system

Notation DescriptionSn Source nodeDn Destination nodeNid Node identityNHid Neighbor nodersquos identityNSS Nodersquos signal strengthNSNR Nodersquos signal to noise ratioNTS Nodersquos time stampNHAvSS Neighbor nodersquos average signal strengthNHAvSNR Neighbor nodersquos signal-to-noise ratioNHTS Neighbor nodersquos time stampDPRep Dedicated probe replayLATfields Local aware table fieldsHP High-profiled userMP Medium-profiled userLP Low-profiled userServicepath Service pathSReqprediction_index Service-requested prediction indexSOfdprediction_index Service-offered prediction index


update ie the LAT in each node is updatedwhenever a ldquoP_beaconrdquo packet from the neighboris received Due to dynamic change in topologythere is possibility that each node receivesldquoP_beaconrdquo packets from new or old neighbors andrecords the new or updated information in its LATAs soon as the LAT records a new entry or updatesthe existing entries the node embedded with SD-MAC functionality automatically reinitiates a LATlook-up process Moreover the influence that thisbehaviour across the network is low as in mostcases the occurrence of this behaviour in MANETis found to be minimal or negligible

Step 6 Upon receiving DPReq packet the SCN checksthe query_handler (QH) table +e QH table maintainsthe details related to query handled by each node IfSeqid in DPReq packet exists in QH table then therequest is considered duplicate and discarded otherwiseit is forwarded to its SCNs It also sends a DedicatedReplay Acknowledgement packet (DRAck) to thesource indicating the status on path selection and fordecision making by the source +e process continuesuntil DPReq packet reaches the destination node Al-gorithm 1 summarizes the steps involved in the pathselection process in the SD_MAC model

323 Resource Reservation On receiving the first DPReqpacket the destination initiates the resource reservationprocess It generates a Dedicated Probe Replay (DPRep)message and sends it along the reverse path +e intermediateSCNs which receives the DPRep packet revalidates its resourceavailability by checking its current SOfdprediction_index(SOfdprediction_indexge SReqprediction_index) and if satisfied itupdates its node state (Ns) as reserved Note each nodemaintains a ldquoNode State-Nsrdquo which helps in detecting andrecovering QoS violations At any particular instant of timethe Ns exists in any one of the following states

(i) checked if path is selected during path discovery(ii) reserved if node reserves resource during the re-

source reservation process(iii) allocated if node receives data packets and

processes(iv) liberated if resources are released by the node

Resources reserved by the node can be liberated if datapackets are not received by the node within a short allocatedperiod (the time slot is set specific to the application withinwhich if the neighbor node does not receive the data packetsfrom the source or from its neighbor it will release thereserved resource for other purposes +is ensures that theresource reserved for data transmission in the neighbornodes is utilized optimally rather than being locked forperforming a particular task) After sending DPReq alongmultiple SCNs the source may receive multiple DPRepresponse from different service paths It is now the re-sponsibility of the source to select the best service path todeliver the data traffic Our SD_MAC model selects the path

with least service path discovery time by always choosing thefirst DPRep response path because this is the path that hasexperienced the least delay and is probably less congested andshorter than the later arriving DPRep response path By usinga dedicated control session the source selects a reliable path (apath that is established for an acceptable period for trans-mission) with intermediate SCNs that have the capability toserve source nodersquos QoS requirement Now the source admitsdata traffic along the selected path to destination

Figure 2 represents the sequence diagram depicting thesteps performed in a routing service component

33 Neurofuzzy Inference-Based Admission Control (NFAC)Mechanism +e SD_MAC model makes use of the fuzzyneural approach based admission control to exhibit betterefficiency which leads to higher userrsquos satisfaction +eNeurofuzzy Inference-based Admission Control (NFAC)engine combines the linguistic control capabilities of thefuzzy logic controller and the learning capabilities of theneural networks to develop an intelligent system to handleincoming traffic keeping the resource utilization at an op-timal level

+e Neurofuzzy Inference-based Admission Control(NFAC) engine acts as a key component of the SD_MACmodel +e proposed model embeds the NFAC serviceengine an automated process triggered in each node to

(i) find the scheduling mechanism and the type ofservices being used to estimate the current datatraffic serviced and the actual bandwidth availablefor accommodating other data services

(ii) Find its current availability by observing and pre-dicting the network conditions

(iii) Identify the amount of resources utilized for high- medium- and low-priority packets Its core ob-jective is to generate the ldquonode status Nstatusrdquo andthereby derive the ldquoprediction indexrdquo which deter-mines whether the node has the capability to ac-commodate the type of services requested by othernodes in the network

In SD_MAC the Admission Control module maintainsthe prediction index value which helps each node to dy-namically regulate real-time data transmission and accu-rately allocate resources in the face of network dynamicssuch as mobility and traffic overloads Unlike other ACapproaches which executes the AC process to estimate thenodersquos current status in SD_MAC the decision regardingthe nodersquos current resource availability is made instanta-neously by merely verifying the ldquoprediction indexrdquo NFACrsquosfunctionality is categorized as follows


34 Fuzzy Inference-BasedAdmissionControl (FIAC) Process+e NFAC service engine is empowered with the FuzzyInference-based Admission Control (FIAC) process which


comprises of fuzzy logic (FIL) [15] that offers better adaptabilityunder varying network conditions by tuning rules withoutintervention of operators Aim of FIAC is to formulate acontrol decision by considering multiple QoS parameters +eStructure of FIAC used in NFAC is shown in Figure 3

Our scenario considers QoS parameters such as

(i) Type of user service packet type (PType)(ii) Varying queue load (QLoad)(iii) Bandwidth availability (BWavail)(iv) Signal intensity (RSSI)

as the input +ese inputs act as a vital sign to accuratelypredict the nodersquos current status in the network Next theinput variables are fuzzified through fuzzification process toform different Fuzzy Linguistic variables To fuzzify thevariables fuzzy sets have to be determined +e crisp inputsare classified with its corresponding membership functionsto form the fuzzy sets

In SD_MAC we have considered 9 fuzzy sets such as

(i) For the HP packet type the fuzzy sets considered areNSHlow NSHmoderate and NSHhigh

(ii) For the MP packet type the fuzzy sets consideredare NSMlow NSMmoderate and NSMhigh

(iii) For the LP packet type the fuzzy sets considered areNSLlow NSLmoderate and NSLhigh

where NSHhigh or NSMhigh or NSLhigh indicates theldquoexcellentrdquo state of the node with minimal traffic less con-gested HP or MP or LP queues and low mobility SimilarlyNSHmoderate or NSMmoderate or NSLmoderate indicates ldquogoodrdquostate of the node with allowable traffic moderate congestedHP or MP or LP queue and average mobility Finally NSHlowor NSMlow or NSLlow indicates ldquopoorrdquo state of the node withhigh traffic high mobility and highly congested HP or MP orLP queues respectively +reshold values ranges betweenTLlow TLmed and TLhigh for LP users TMlow TMmed andTMhigh for MP users and THlow THmed and THhigh for HPusers set as per the application scenario Table 2 describes theclassification of inputs and its fuzzy sets

Fuzzification is followed by the fuzzy rule base creationprocess which uses logical combination of input variableswith the AND (ampamp) operator Quality of results in a fuzzysystem depends on the fuzzy rules +e sample fuzzy rulebase is referred below

Node i broadcasts P_beacon packet to immediate neighborsNi⟶ P_beacon (NId NRSSI NSNR NTS SOfdprediction_index) Neighbor Node j receives P_beacon packet of Ni node iNj⟵P_beacon (NId NRSSI NSNR NTS SOfdprediction_index)extract the beacon of Node i and store in LAT of Node jLATj⟵ extract_data()Nj⟶ send P_beacon (NId NRSSI NSNR NTS SOfdprediction_index)Ni⟵ P_beacon (NId NRSSI NSNR NTS SOfdprediction_index)LATi⟵ extract_data()lowast entry in LAT contains the following field LATfields [NHid NHAvRSSI NHAvSNR NHTS SOfdprediction_index] if source node has data and does not know path to destinationif (Sn (data) ldquoTrdquo ampamp path_to_destination( ) ldquoFrdquo) thenSCNi Sn_lookup (LATsn)end ifif (is exist(SCNi)) then strongly connected neighbor node exists send DPReq packet to SCN foundSn⟶ send DPReq (Saddr Daddr Seqid Servicepath SReqprediction_index) SCN receivesDPReq packetSCNi⟵DPReq (Saddr Daddr Seqid Servicepath SReqprediction_index)Check_QHtable() SCN checks QH tableif (Seqid exists()) then

Discard DPReq()elseif (SOfdprediction_indexge SReqprediction_index) thenappend (Servicepath)extract_signal_data()store_into_LAT()forward DPReq() packet to SCNsend DRAck() to source send Ack to source

end ifend if

end if

ALGORITHM 1 Path selection process in the SD_MAC model


If (PType HP) ampamp (QLHP H) ampamp (QLMP M) ampamp(QLLP L) ampamp (RSSI L) ampamp (BWavail L) then

NStatus NSHlow

If (PType MP) ampamp (QLHP L) ampamp (QLMP H) ampamp(QLLP L) ampamp (RSSIH) ampamp (BWavail H) then

NStatus NSMhigh

If (PType LP) ampamp (QLHP L) ampamp (QLMP M) ampamp(QLLP M) ampamp (RSSIM) ampamp (BWavail M) then

NStatus NSLmoderate

341 Node Status and Prediction Index GenerationInference engine uses the rule base to generate the defuz-zified value node status Nstatus which typically reflects thedynamic behaviour of the traffic in nodersquos buffers and theavailable capacity of the current node Table 3 displays theNstatus derived using the QoS parameters

For example let us consider the packet type (PType) beHP the percentage of queue load for HP be 42 (M) for MPbe 35 (L) for LP be 63 (H) the RSSI be 67 (H) andbandwidth available (BWavail) be 28 (L) +e triangularmembership function representing the range of membership

Step 2 each node receives P_beacon extracts and stores the information in LAT

Strongly connectedneighbour node

Step 1 broadcasts P_beacon = (Nid NRSSI NSNRNTS SOfdprediction_index)

Step 3 Source looks up its LAT to select SCN Itsends a DPReq (Saddr Daddr Seqid Servicepath

SReqprediction_index)to SCN node

Step 5 SCN sends DRAck to source node

Destination nodeSource node

Source has data service to be

transmitted to destination

SCN checks QH table If Seqid exists in QH table SCN discards the request

otherwise it forwards it to its SCN or destination

Step 4 SCN looks up its LAT to forward theDPReq request to its SCNdestination node The destination

receives the DPReq and

initiates resource reservation

processStep 6 destination sends DPRep responsealong reverse path

SCN receives DPRep message revalidates its resource availability If its current

resource availability gt= requested resource then it updates its node state

(Ns) as reservedStep 7 the SCN forwards the DPRep message

to its SCN or source nodeSource receives multiple DPRep

message from multiple SCNs The source

selects the best service path to deliver data

packets

Figure 2 Sequence diagram depicting Routing Service in the SD_MAC model

Inference engine

Rule base Membership function

Node statusFuzzy outputFuzzy

inputsBW availability

DefuzzifierPacket type

Queue load RSSI Fuzzifier

Figure 3 Structure of FIAC process used in NFAC


degree is represented for each input as shown in the fol-lowing figures

If HPQueue Load 42 then the degree of membershipranges between 047 and 053 as shown in Figure 4

For MP Queue Load 35 then the degree of mem-bership ranges between 042 and 058 as shown in Figure 5

For LP Queue Load 63 then the degree of mem-bership ranges between 046 and 056 as shown in Figure 6

For RSSI 67 then the degree of membership rangesbetween 031 and 079 as shown in Figure 7

For bandwidth available 28 then the degree ofmembership ranges between 03 and 08 as shown inFigure 8

+e fuzzification process combines the input to generatethe defuzzified value node status Nstatus 79 (NSHmod) asrepresented in Table 4+eNSHmod status indicates the nodeis in good state and can admit allowable data traffic as itsqueues are moderately congested

+e SD_MAC model considers the Nstatus as one of theprimary factor [24] in making decision during admissioncontrol Followed by Nstatus each node generates a measureof prediction index an estimated value that reflects thenodersquos current resource availability It is used to validate thecurrent resource utilization of the node based on trafficmobility and congestion +e Admission Control modulemaintains two types of prediction indexes such as

(i) Service-offered prediction index (SOfdprediction_index)(ii) Service-required prediction index

(SReqprediction_index)

Table 2 Classification of input ldquoirdquo and its fuzzy set

Input field Range Values Fuzzy set

High-profileduser (HP)

TMhighgt iltTHlow 70gt ilt 78 NSHhighTHlowge iltTHmed 78ge ilt 85 NSHmoderateTHmedge ileTHhigh 85ge ile 100 NSHlow

Medium-profileduser (MP)

TLhighgt iltTMlow 30gt ilt 40 NSMhighTMlowge iltTMmed 40ge ilt 55 NSMmoderateTMmedge ileTMhigh 55ge ile 70 NSMlow

Low-xprofileduser (LP)

0ge iltTLlow 0ge ilt 8 NSLhighTLlowge iltTLmed 8ge ilt 18 NSLmoderateTLmedge ileTLhigh 18ge ile 30 NSLlow

Table 3 Crisp output Nstatus derived using the QoS parameter

PTypeQLoad ()

RSSI BWavail Node status NstatusQLHP QLMP QLLP

HPH M L L L NSHlowM M L H H NSHmodL M H H H NSHhigh

MPL H L H H NSMhighM L H M M NSMmodH H L H H NSMlow

LPH L L M H NSLhighM M H L M NSLlowL M M M M NSLmod

Mem

bers

hip

degr

ee

Low

0

05

1High

0 50 70 10030HP queue load

Medium

Figure 4 Triangular membership representation for HP QueueLoad

Mem

bers

hip

degr

ee

Low

0

05

1High

0 50 70 10030MP queue load

Medium

Figure 5 Triangular membership representation for MP QueueLoad

Mem

bers

hip

degr

ee

Low

0

05

1High

0 50 70 10030LP queue load

Medium

Figure 6 Triangular membership representation for LP Queueload

Mem

bership

degree

Low

0

05

1High

0 50 70 10030RSSI

Medium

Figure 7 Triangular membership representation for RSSI

Mem

bers

hip

degr

ee

MediumLow

0

1

Bandwidth available

High

0 10 28 40 50

08

03

Figure 8 Triangular membership representation for availablebandwidth


342 Service-Offered Prediction Index (SOfdprediction_index)It refers to the service currently offered by the node ie thenodersquos current resource availability is specified using thisvalue +e Nstatus combines with the current load status ofHP MP and LP queues to generate the SOfdprediction_index+e threshold limits for the HP queue (such as +Qhigh+Qmed and+Qlow) MP queue (such as TmQhigh TmQmedand TmQlow) and LP queue (such as TlQhigh TlQmed andTlQlow) load status are set as per application requirement+en SOfdprediction_index is evaluated by validating the Nstatusalong with the queue load status For example

if (Nstatus NSHhigh)ampamp(HPQloadgt+Qhigh)ampamp(MPQloadgtTmQmed)ampamp(LPQloadleTlQlow) thenSOfdprediction_index V where 0leVle 1end

where V ranges from 0leVle 1 provided ldquo1rdquo indicatesthat the node is completely occupied and is unavailable tooffer any service currently while ldquo0rdquo indicates that the nodersquosresources are free and is fully available to offer variedservices

343 Service-required prediction index (SReqprediction_index)It refers to the service requested by the node ie the nodeacts as a source and has a data service to be transmitted to adestination +e service to be supported by other nodes isspecified using this value +e type (TRType) and measure(MData) of data service along with the bandwidth required(Breq) are validated to generate the SReqprediction_index

For example

if (TRType ldquovoicerdquo)ampamp(MDatagtDhigh)ampamp(BWreqleBWmax) thenSReqprediction_index V where 0leVle 1end

where V ranges from 0leVle 1 provided ldquo1rdquo indicateshigh-profiled data service with high resource requirement isrequested by the source while ldquo0rdquo indicates low-profileddata service with very low resource availability is requestedby the source

Dynamic change in node behaviour (caused due to trafficcharacteristics mobility number of neighbors etc) tends toregenerate the dedicated probe request (prediction index)which is fed as part of the input to Neuro-MultilayeredLearning (NML) to accurately identify strongly connectedneighbors Detailed analysis on the NML process is dis-cussed in the following section

35 Neuro-Multilayered Learning (NML) Process +e pro-posed work uses the Artificial Neural Network (ANN) to

train the model using various input parameters such as thededicated probe request (prediction index) to accuratelypredict the strongly connected neighbor nodes +e aim ofthe proposed model is to recognize the weights of each inputparameter using machine learning specifically the ArtificialNeural Network +e ANN is a supervised mathematicallearning tool inspired by biological neural networks +enetwork is connected with a vast number of nodes calledneurons +e connections between the nodes imitate thesynaptic connections found in biological entities

ANNs are incredibly useful for analysis of mapped in-formation as it can be trained to recognize patterns in data

+e aim of implementing the ANN is to predict theweight factors for different nodes such that the best possiblenode with a weight factor can be selected +e proposedNeuro-Multilayered Learning (NML) mechanism imple-ments a multilayer perception (MLP) to learn analyze trainand adapt to the operational behaviour of various compo-nents and optimize decision according to the nodersquos currentresource availability +e NML maps various features toderive a set of output with good accuracy and high learningcapability It is composed of a vast number of connectedunits arranged in layers Among which the first layer is theinput layer and the last layer is the output layer +e innerlayers are referred to as the hidden layers which are re-sponsible for the bulk of the processing+e process involvesimportant activities such as

(i) Train the network(ii) Estimate and accurately admit data service

351 Train the Network NML initiates a training process toassure adaptive behaviour of the network Training thenetwork involves estimating the connecting weights byminimizing the overall error measures ie tasks where theweights of each unit (neurons) should be adjusted in such away that the error between the desired output and the actualoutput is reduced+e error derivative computed during thisprocess helps to accurately classify each instance +eprocess involves the following steps

Step 1 the NML receives input parameters (also re-ferred as instances) including the probe request con-stituting the SOfdprediction_index generated from FIACprocess Multiple instances (dynamic change in nodersquosbehaviour triggers NFAC engine which derives theSOfdprediction_index prediction index value) at variedtime slot are received by the NML process Let ldquoi1rdquo bethe instance received at time slot t1 Change in nodersquosbehaviour causes regeneration of an instance ldquoi2rdquo attime slot t2Step 2 the instances are fed into the NML process toestimate the error derivative using the steps elaboratedin ldquoMathematical Analysis of NMLrdquo It then classifiesthe instances appropriately for accurate decisionmaking Over a period of time the training processescraft each node to self-organize and adapt to dynamicchanges making the whole network intelligent


PTypeQLoad ()

RSSI BWavail

NodestatusNstatus

QLHP QLMP QLLP

HP (0) 42 (M) 35 (L) 63 (L) 67 (H) 28 (L) 79 NSHmod


Mathematical Analysis of NML aim of the NML is toselect the best strongly connected neighbor node basedon the Credibility Score (CS) +e credibility score wascalculated as

CS (αlowastNN) +(βlowastDP) +(clowastPF) (3)

where α β and c are the weight factors and NN DP and PFare the number of neighbors dedicated probe request andpacket forward rate factors respectively

+e proposed model considers 5 parameters as inputsuch as number of nodes average number of neighborsmobility dedicated probe request and packet forward rateA diagrammatic representation of the artificial neural net-work of the SD_MAC model is shown in Figure 9

+e input parameters such as number of nodes andmobility determine the kind of network we are dealing with[25] the packet forward rate determines the traffic char-acteristics number of neighbors and dedicated probe re-quest determines the strongly connected neighbor which iswhy these parameters has been included +e output layerhas 8 nodes resulting in 69 possible cases with a combinationof α β and c values +e sample values generated for thecases are shown in Table 5

+e cases with 0 values are excluded (the weight factorwith 0 relevance would mean that the factor would becomeirrelevant for the network) to generate 36 possible cases asdepicted in Figure 10 ie

α + β + c 1 (4)

+e NML was implemented to have 5 inputs 36outputs and 3 hidden layers with 10 neurons each Eachnode acts as a decentralized processing unit which basedon its own data activates nodes of the next layer A node isactivated by a weighted sum of all its parent nodes andtakes on probabilistic values For example let nodes bedenoted by the notation aij where i refers to the layercorresponding to the node while j refers to the nodenumber +us the first node in the input layer will benamed a00 while the last node in the same layer will benamed a04 +e first node in the next layer will similarly benamed a10 +e value a10 will be a weighted sum of all thevalues of its parents viz a00 a01 a0n where n is thenumber of nodes in the layer Since the sum will be greaterthan one in this work we use a normalizing functionReLU [26] (as ReLU achieves better result than sigmoid[27]) so that the value appears between 0 and 1 +eweights are the values assigned to the arrows linking eachnodes We use the notation wijk where the arrow connectsnode j of the layer i to node k of the layer i + 1 It is byadjusting these weights that we determine the strength ofthe connections and how firing of one neuron will activatethe next one To calculate the weights of the zeroth node(a10) on the first layer and its value is computed bynormalizing the sum of all the values in the previous layermultiplied with their weights Mathematically in case ofnode a10

a10 σ a00 lowastW000 + a01 lowastW010 + a02 lowastW020(

+ a03 lowastW030 + a04 lowastW0401113857(5)

For each node wijk may be written as wj +us theequation is rewritten as



In general it is expressed as

aij σ 1113944n

j0a(iminus 1)j times wj

⎛⎝ ⎞⎠ (7)

where aij is the jth node on ith layer andwj is the weight of thejth node in the i minus 1 layer In this manner all the nodes in theith layer are assigned values +is continues for all successivelayers and the neurons keep firing till finally the outputlayer is activated Initially the model starts with genericweights and in each successive case the weights are refinedto better suit the predetermined result A cost functiondetermines how close the result was to the actual value+isprocess occurs multiple times till the results converge andweights in each layer obtain the best possible value to suitthe output In our case each of the 36 output nodes wasassigned a probability value +e data was divided such that75 was used for training and 25 for testing To train ourmodel we use a simple mean squared error (MSE) as costfunction +e MSE is attractive because it is simple to useparameter-free and inexpensive to compute +us theMSE is computed as

MSE 1113944n

j1aLj1113872 1113873 (8)

where L refers to the last layer n is number of outputs andldquoardquo is the expected output value +e sum gives us the totalerror for the network For example if we consider only thefirst four of the 36 possible cases and assume value 03 is thecorrect answer we may obtain the values as shown inTable 6 in the first iteration+e error in this scenario wouldbe the sum of all the squares of differences between theobserved value and expected value

MSE (05 minus 00)2

+(02 minus 00)2

+(03 minus 00)2

+(04 minus 00)2

094(9)

Similarly the MSE is calculated for all possible cases andthe average is computed For each new dataset consisting ofall the weights and biases the cost function is minimized byreducing the error and thereby allowing themachine to learn+is model consisting of various parameters becomes capableof predicting the best weight factors (SCN) for the network

352 Estimate and Accurately Admit Data ServiceApart from training the network NML also helps in estimatingand accurately admitting data service using the following steps


Step 1 at a particular instance of time each node re-ceives multiple DPReq (Saddr Daddr Seqid ServicepathSReqprediction_index) packets from its neighbor nodes Atthe same time the nodersquos AC module maintains itscurrent SOfdprediction_index valueStep 2 the node gathers the SReqprediction_index (in-stances) from multiple DPReq packets and forwards itto the NML process First it categorizes instances basedon the traffic type (HP or MP or LP) and groups themtogether to form a similar set S (xi yi zi) 1le ile nconsisting of ldquonrdquo instances where (xi isinHP Yi isinMPand Zi isin LP) Next it classifies and labels instanceseither as ldquoAccept candidate-ACrdquo or ldquoReject candidate-RCrdquo or ldquoUnknown candidate-UCrdquo

(i) Accept candidateAC an instance is said to be labeledas AC provided the instance satisfies the predefinedrule (SOfdpredictionindexgt SReqpredictionindex)

(ii) Reject candidate RC an instance is said to be labeledRC provided the instance satisfies the predefined rule(SOfdpredictionindexlt SReqpredictionindex)

(iii) Unknown candidate UC all other instances arelabeled as UC

+e unknown candidate list is fed into NML for furtheroptimization +e multilayered process estimates theerror derivative using the steps elaborated in ldquoMathe-matical Analysis of NMLrdquo It then classifies instances withallowable errors as AC and otherwise as RC respectivelyStep 3 the instances (SReqprediction_index) labeled as ACare prioritized and considered as allowable data service(ie the node acts as SCN for its neighbors) +e nodereserves resources as per the request and triggers theNFAC to regenerate the SOfdpredictionindex Repeat theprocess from step 2

+e labeled candidates are stored into the local ManagedInformation Database (MID) maintained by each node +eMID has storage capacity for storing varied patterns oflabeled candidates +ese patterns are then used for datamining and classification during NMLrsquos training activity+e idea behind the NML model is to reduce the compu-tational complexity at each level and make the data miningefficient +e unique candidate list from the MID are col-lected and sent to its neighbors across the network +eneighbor node receives the candidate list and updates itsMID with unique candidate list Algorithm 2 summarizessteps to estimate and accurately admit data service

+e error (Err) between real and expected output iscomputed If Err is within the allowable range (AR) it in-dicates that instance is covered by an existing attractingbasin +en θij and αij are adjusted If Err is beyond the ARit means that instance is not covered by any existingattracting basins and need to find ldquohrdquo with minimum errorldquourdquo If Erc is in the AR it means that the internal outputclassification represented by unit ldquourdquo is applicable to theinstance Also the input to ldquojrdquo should be adjusted +us theunit whose AV is the maximum among those connectingwith unit ldquourdquo is selected If Erc is beyond the AR it meansthat both input classification and output classification isinadequate for the instance +us ldquojrdquo and ldquohrdquo are appendedto the hidden layers+e new second layer ldquosrdquo is connectedwith

Input layer Output layer

α = 01β ndash 01 γ ndash 08

α = 08β = 01 γ = 01

Number of nodes

Dedicated probe request

Packet forward rate

Number of neighbours

Mobility Hiddenlayers

Figure 9 Artificial neural network of the SD_MAC model

Table 5 Sample of α β and c value including 0 values

α β c

1 0 00 0 101 0 0902 0 0803 0 0704 0 0605 0 0506 0 0407 0 0308 0 02


all the input units +e new third-layer unit is connected withall the output units to generate a fine tuned expected output+e error derivative analysis using the internal layers of theNML process is represented in Algorithm 3

+e NML process considers the error (Err)(Erc) valuewhich is the difference between the real and expected outputto classify the instances as follows

(i) If error is within allowable range then they areclassified as ldquoAccept candidate ACrdquo

(ii) If error is beyond allowable range then they areclassified as ldquoReject candidate RCrdquo

After several execution of the training stage NML startstesting operation to test the correctness of the results +esuccess of network learning is to satisfy the user data flowrequirement without performing any operations at the nodelevel +is makes the entire network to offer data serviceaccurately and select the most suitable SCN nodes

36 Classifier Traffic Regulator and Scheduler Any givennetwork supports various types of data flows and categorizesit into two types

(i) nonreal-time traffic or best-effort flows such asmessages which requires the data to be reliablydelivered to the destination

(ii) real-time traffic or QoS flows such as voice and videowhich apart from reliability requires additionalconstraints such as available bandwidth and delay tobe satisfied

+e real-time traffic is normally expected to get betterservice than the others Hence priority for acquiring re-sources in a network is generally given to the real-timetraffic In addition to real-time flows nonreal-time flowssuch as messages play a very crucial role because of theirapplication in military battle field and emergency rescueoperations +ese applications require fast and assuredtransmission of emergency messages Hence in such ascenario priority for acquiring resources in a network has tobe given to the emergency nonreal-time traffic based on theuser profile QoS in MANET becomes a necessity due to itsapplications in critical scenarios +e SD_MAC modelmakes use of the classifier traffic regulator and the schedulermodules to distinguish regulate and schedule data trafficappropriately Figure 11 shows how these modules work

Case 7

Case 9

Case 13

Case 21

Case 4Case 3Case 2

1

Number of neighbors

Number of nodes

Mobility


Packet forward rate

1 1 1

Case 1

Case 5Case 6

Case 8

Case 11Case 10

Case 12

Case 19

Case 16Case 15Case 14

Case 17Case 18

Case 20

Case 22

Case 36

Figure 10 NML formed in SD_MAC with input hidden and output layers

Table 6 Sample values generated in first iteration

α β c Observed value Expected value01 01 08 03 001 02 07 02 101 03 06 09 001 04 05 04 0


Input +e instances i1 i2 i3 received from Neighbor nodesI generate_set (i1 i2 i3) I denotes the multi-instance setsInitialize icount 0 j 1icount getCount(I) get the number of instances (icount)

for j 1 to icountrepeat repeat and validate multi-instance set Xi Yi

if (ij Xi) thenif (SOfdpredictionindexgt SReqpredictionindex) then

AC_list Iiend

else if (ij Yi) thenif (SOfdpredictionindexlt SReqpredictionindex) then

RC_list Ii Instance labeled as reject candidateend

elseUC_list Ii instance labeled unknown candidate

err derive_Error() find error derivativecalssify_candidates(err) classify instance AC or RC or UCend

j j+1until (jlt icount) continue for all instancesend lowast end of for loopM label getMasterCandidate(AC_list RC_list UC_list)add_to_MID(Mlabel) add master list to DBsend_to_neighbor(Mlabel)Output Classify instances and generate error derivative

ALGORITHM 2 Estimating and admitting data service

Err DeriveError (Routput Eoutput) error Err

if (Err ltAlwrange) then Alwrange is allowable rangeAlwvalue AdjustSlayer (θij αij) θij and αij are adjustedIi classifyCandidate (Err CInstance Avalue)ACandidate_list Ii instance is set as acceptable candidateupdateMID (ACandidate_list)

else if (Err gtAlwrange) thenTlayer findMerrorTLUnit() find 3rd-layer min errorErc ComputeCharacteristicerror (sendtoTlayer(CInstance))if (Erc ltAlwrange) thenIi classifyCandidate (Erc CInstance)ACandidate_list IiupdateMID (ACandidate_list) Assigned as accept candidateelse if ( Err gtAlwrange) thenRC IOCinadequate(Ilist)if (RClt+value) thenIi classifyCandidate (Erc CInstance)ACandidate_list IiupdateMID (ACandidate_list)

elseIi classifyCandidate (Erc CInstance)RCandidate_list Ii Assigned as reject candidateupdateMID (RCandidate_list)

endendelseIlist addinformation (CInstance Err Erc)end

ALGORITHM 3 +e error derivative analysis of the NML process


together to provide the service +e classifier scheduler andtraffic regulator modules operate between the IP layer andthe MAC layer

In SD_MAC model after the source selects the bestservice path (a path that is established for an acceptableperiod for transmission with intermediate SCNs that havethe capability to serve source nodersquos QoS requirement) todeliver the data traffic it then admits data traffic along theselected path to the destination At each node the ad-mission control module maintains a SOfdpredictionindex toreflect the nodersquos current resource availability (in MANETthe nodersquos current resource utilization keeps changing dueto change in traffic mobility and congestion) using whichit revalidates if the SOfdpredictionindex gt SReqpredictionindex(indicates the data service to be supported by intermediatenodes) If the SOfdpredictionindex is high then the admissioncontroller admits the session for processing otherwiserejects it When a new session is admitted the classifier thetraffic regulator and the scheduler performs the followingprocess to support the desired QoS Based on the traffictype the classifier classifies the packets QoS profile asplatinum gold silver and bronze It then marks the pri-ority of those packets HP (High Profiled code is 0) MP(Medium Profiled code is 1) and LP (Low Profiled code is2) +us the classifier differentiates real-time and nonreal-time sessions forcing the traffic regulator to process MP andLP packets but not the HP packets For example if thetraffic type is ldquonetwork controlrdquo then the QoS profile is setto ldquoPlatinumrdquo and the user priority is marked as HP +enbased on marking the packets remain unregulated andbypass the traffic regulator module if it is HP and are forcedto pass through the traffic regulator if they are MP or LP+e goal of the traffic regulator is to regulate the data trafficof MP and LP packets (ie it tends to delay MP and LPpacket processing) and utilize the resource for processingHP packets (real-time sessions) +e stamped (marked)packets are then queued by the scheduler in their respectiveHP MP and LP queues and wait for transmission +escheduler implements a combination of Strict and Proba-bilistic Priority Queuing to impose service differentiationas per the userrsquos priority and serve packet in their respectiveHP MP and LP queues It permits users to utilize resourcesopportunistically by regulating high traffic imposed byother nodes

Detailed specification on classifier traffic regulator andscheduler modules are referred in the following section

361 Classifier To meet the userrsquos service requirements andprovide better service to selected network traffic the clas-sifier module in SD_MAC differentiates packets into fourclasses platinum gold silver and bronze as referred inTable 7

Based on the type of traffic the packets are stamped as

(i) High-Profiled user (HP) supports high bandwidthto ensure high quality of service

(ii) Medium-Profiled user (MP) supports nominalbandwidth for moderate QoS

(iii) Low-Profiled user (LP) supports lowest bandwidthfor bulk data service

+e first class ldquoplatinumrdquo and ldquogoldrdquo has the highestpriority and corresponds to applications with real-time (RT)traffic such as voice and network control packets +is classsupports high bandwidth to ensure high QoS +e class isreferred to as ldquoexpedited forwardingrdquo and in IntServ is re-ferred as guaranteed service +e second class ldquosilverrdquo hasless priority than the first class It is suitable for applicationsrequiring high throughput such as video applications +isclass supports nominal bandwidth for moderate QoS +eclass is referred to as ldquoassured forwardingrdquo in DiffServ andldquocontrolled loadrdquo in IntServ +e third class ldquobronzerdquo is theleast priority class and has no specific constraint +is classsupports lowest bandwidth for bulk data service +is class isreferred to as best effort in both DiffServ and IntServarchitectures

Based on Traffic type and QoS profile each user isassigned a static priority as per the classification HP MPand LP To meet service requirements a proportional weight(which favours fairness among the competing nodes) isassigned to each user +is is implemented by adding apriority field to the header of every packet +e codes (0 1and 2) representing the priorities are stored in the priorityfield +e intermediate node does enqueuing based on thepriority value

362 Traffic Regulator +e Traffic Regulator module reg-ulates MP and LP (best effort) traffic by delaying best-effortpacket processing and allows more bandwidth for HP (real-time) packet processing When a new real-time flow isallowed by the admission control block the packets markedas HP by passes the traffic regulator and are queued in HPqueues for processing

363 Scheduler +e Scheduler module employs a newscheduling discipline referred to as the Strict and Proba-bilistic Priority Queuing (SPPQ) which determines theprobability with which the queue is served by assigning aparameter to each priority queue It offers variety of servicedifferentiation among various groups of traffic classes Toprovide high QoS different priorities are assigned to packetsentering the nodes in the network +e packet schedulingalgorithm is employed to select each packet and transmit itto the data link layer+ough a number of packet schedulingalgorithm have been proposed the prominent ones amongthem include Strict Priority (SP) (ie servicing lower pri-ority packets only when higher-priority ones are not waitingto be serviced) [28] Weighted Fair Queuing (WFQ) (iehaving N data flows currently active with weights w1 w2 wn data flow i will achieve an average data rate of R lowast wi(w1 + w2 + middot middot middot + wN) where R is the data link rate) [29] andWeighted Round Robin (WRR) (ie assigning a portion ofthe available bandwidth to each priority queue) [30] areperhaps the most widely adopted disciplines Among thethree disciplines the SP is easy to implement and provideslarge differentiation among classes However the SP is


unfair to all classes other than the highest priority class +ismechanism of SP may introduce large periods of starvationto lower priority classes and usually cannot handle thisstarvation problem by itself To deal with starvation problemand make the SP discipline controllable we propose toassign a parameter to each priority queue which determinesthe probability with which the queue is served Hence a newscheduling discipline referred to as the Strict and Proba-bilistic Priority Queuing (SPPQ) is formed Similar to WFQand WRR variety of service differentiation can be achievedin SPPQ by setting the assigned parameters properly +enovel property of the SPPQ discipline is that it providesservice segregation among groups of traffic classes and hasthe following characteristics

(a) It is easy to implement and verify(b) It exploits well the available network bandwidth(c) It consumes very small memory and processing

power requirements(d) Network administrators find them easy to under-

stand and configure(e) It does not require time-stamping as in WFQ

+e performance of SPPQ algorithm has been evaluatedin this work by considering different priority classes andtraffic patterns+e evaluation shows that with high networkload ie the packet arrival probability λgt 06 the QoS of-fered to low-priority packets rapidly deteriorates with itsthroughput significantly dropping and delay sharplyincreasing

In this paper we consider the network to operate underthe following assumptions

(i) +e arrival process of each input of the network is asimple Bernoulli process that is the probability thata packet arrives within a clock cycle is constant andthe arrivals are independent of each other We willdenote this probability as λ +is probability can befurther broken down to λh λm and λl which rep-resent the arrival probability for high- medium- and low-priority packets respectively It holds thatλ λh+ λm+ λl

(ii) Under this mechanism when applications transmita packet to the network they specify its prioritydesignating it either as high medium or low +ecriteria for priority selection may stem from thenature of packet data

(iii) Each packet priority queue is statically assigned aweight which specifies the bandwidth ratio that willbe dedicated to the particular queue with their summust be equal to 1

(iv) Upon reception packets are first classifiedaccording to their priority and are then assigned tothe queue specifically dedicated to the particularpriority as depicted in Figure 12

During the network cycle the SPPQ algorithm evaluatesthe queues by observing the bandwidth ratio that has beenassigned to each queue to select the packet to be forwardedthrough the output link +e packet is always transmitted

HPMPLP

IPlayer

Classifier

User requestAdmitreject

flow admissionRequest for

flow admission

Routing service

Admission controller

Traffic regulator

Scheduler

MAC

Figure 11 SD_MAC model with classifier traffic regulator and scheduler modules

Table 7 QoS Profile supported in the SD_MAC model

Traffic type (TRType) QoS profile User priority (PType) Codes representing priorityNetwork control Platinum High profiled 0Voice Platinum High profiled 0Interactive video Gold High profiled 0Mission critical Gold High profiled 0Best effort Silver Medium profiled 1Bulk data Bronze Low profiled 2Scavenger Bronze Low profiled 2


when there is traffic waiting as opposed to nonwork con-serving algorithms which do not transmit a packet if thequeue is found to be empty+e bandwidth is divided amongthe queues that do have packets to transmit proportionallyto their weights in case if the queue does not use itsbandwidth ratio within a time window

Every node maintains a separate queue such as HP MPand LP for the four classes of users

According to the priority the packets are enqueued anddequeued from the queues proportionally based on theweights and the percentage of packets waiting in their re-spective queues Let a b and c represent the percentage ofHP MP and LP packets waiting in their respective queuesand it is calculated as follows

a QLHP

QLHP + QLMP + QLLP( 1113857

b QLMP


c QLLP


(10)

where QLHP QLMP and QLLP represent the number ofpackets occupied in HP MP and LP queues and is dynamicFrom each queue the number of packets to be dequeued isdecided based on their access ratio and is calculated as follows

wHPa wMPb wLPc (11)

where ldquowrdquo represent the weight which is a constant+ewHPwMP and wLP are the user defined weights assigned for HPMP and LP queues such that wmax gtwHP gtwMP gtwLP gt 0+e wmax represents the maximum weight +us at anypoint of time for any random data flow the priority of HPpackets over MP and LP and the priority of MP packets overLP is ensured

4 Simulation Results and Discussion

We compare in this section the performance of our proposedSD_MAC scheme with other well-known admission control

protocols such as SWAN StAC DACME and FuzzyQoSFor evaluation purpose the simulator was developed usingMATLAB +e simulation prototype provides a user-friendly Graphical User Interface (GUI) enabling the user toedit the network topology and set the parameter values of theprotocol module in just a few mouse clicks

+e GUI of SD_MAC is shown in Figure 13+e simulation environment was tested with 50 nodes

moving in an area of 1000times1000 sq m Each node imple-ments random waypoint mobility model +e transmissionrate of nodes is 2Mbps while their transmission range is250m Initially when the simulation is initiated the nodesare deployed at random location in the geographical area asshown in Figure 14

+e mobile nodes move randomly and freely with aspeed varying between 0ms to 35ms +e networktraffic model is the Poisson distribution model with bufferspace in each node configured as 100 packets +e ap-plication with mixture of traffics such as FTP Webmicroflows VoIP and real-time traffics +e voice andvideo flows represent real-time traffic with voice modelledas 32 kbps constant rate traffic with a packet size of 80bytes video modelled as 200 kbps constant rate trafficwith a packet size of 512 bytes and TCP flows are greedyFTP traffic type with packet size of 512 bytes are usedduring evaluation

+e parameters used as part of simulation are listed inTable 8

During simulation the nodes are randomly chosen to actas the source (node that generates message and transmits themessage to destination) with 20 sessionssource +e sourceand destination nodes are distributed among the mobilenodes in the MANET Figure 15 displays the network modelof the SD_MAC system

To better investigate the scope and performance [31ndash35]of the protocols the metrics are selected carefully for fairresult analysis With constant node density scenario byvarying the speed of nodes (as it affects the performance ofthe protocols due to frequent topology changes) the sim-ulation was run for 500 iteration to derive the average resultsand the observation on various metrics is analyzed

HP (high)

LP

ClassifyMP (medium)

LP (low)

60

30

10

Incomingpackets

HP MP

Figure 12 +e SPPQ mechanism of the SD_MAC model


Scenario 1 in this scenario the number of nodes isvaried from 50 to 300 while the nodes were set to movewith low mobility of 5msScenario 2 in this scenario the number of nodes isvaried from 50 to 300 while the nodes were set to movewith high mobility of 30ms

41PacketDeliveryRatio (PDR) PDR defines the ratio of thenumber of packets delivered to the destination successfullyagainst the total number of packets generated by the source

For better insight and comparative analysis amongvarious approaches simulation results for scenario 1 andscenario 2 are captured and presented in Tables 9 and 10

411 Scenario 1 Analysis In scenario 1 all the nodesexhibit low mobility across the network +e PDR datacaptured for various approaches are represented inFigure 16(a) From Figure 16(a) we observe that thoughthe FuzzyQoS scheme performs better in deliveringpackets successfully to destination when compared toDACME StAC and SWAN the proposed SD_MACscheme demonstrates better PDR than FuzzyQoS +e fact

Figure 13 GUI of the SD_MAC model

0 10 20 30 40 50 60 70 80 90 1000

10

20

30

40

50

60

70

80

90

100

501

23

45

6

7

8

9

10

11

12

13

14

1516

17

1819

20

21

22

23

24

25

26

2728

29

30

31

32

3334

35

36

3738

39

40

4142

43

44

45

46

47

48

49

Figure 14 Network model with nodes deployed at RandomLocation

Table 8 Simulation parameters

Parameters ValuesSimulator MATLABSimulation time 100 sTopology area 1000mtimes 1000mNumber of nodes 50ndash300Mobility speed 0 to 35msMobility model Random waypoint modelTransmission range 250mTransmission rate 2MbpsPacket size 80 and 512 bytesTraffic type Constant bit rate (CBR)Number of CBR connections 5Pause time 0 sChannel type Wireless channelAntenna type Omni antennaMAC type IEEE 80211Frequency 24GHz


that in the SD_MAC model the packets are admitted by theadmission controller through the SCN nodes delivered suc-cessfully to the destination +e path detected remains intactdue to low mobility Significantly it is observed that when thenode density increases the percentage of PDR also increasesbecause the number of SCNs identified during path selection ismore enabling SD_MAC to select the best among themresulting in high PDR With 100 nodes SD_MAC surpassedDACMErsquos StACrsquos and SWANrsquos performance by sim10 sim15and sim18 and FuzzyQoS by sim5ndashsim8 Nevertheless evenwhen the number of nodes is increased to 300 the proposedscheme was able to sustain the PDRgt 95 and remain stablein delivering packets successfully to destination when com-pared to other schemes

412 Scenario 2 Analysis In scenario 2 all the nodes exhibithigh mobility across the network In Figure 16(b) we can

observe the PDR of the proposed SD_MAC scheme outper-forms other schemes and maintains the PDR above 80 evenwhen the node density increases and exhibits high mobility+eNeurofuzzy Inference-based Admission Control engine inthe SD_MACmodel exploits the linguistic control capabilitiesof the fuzzy logic controller and the learning capabilities of theneural networks to handle incoming traffic and utilize theresources at an optimal level It is intelligent and fast detectionof SCNs helps in achieving high PDR even when mobilityincreases Additionally the neuro-multilayered learningtechnique ensures the network to adapt to dynamic behaviourand make accurate decisions spontaneously +ough DACMEhas better PDR compared to StAC and SWAN and its PDRstarts degrading when the number of nodes increases Fromthe result it is observed that when the node density is 125 theSD_MAC surpassed SWANrsquos and StACrsquos performance bysim12ndashsim15 andDACME by sim10+e PDR for SD_MAC is

0 20 40 60 80 1000

10

20

30

40

50

60

70

80

90

100

50

1

2

3

4

5

6

7

8

9

10

11

12 13

14

1516

17

18

19

20

21

22

23

24

25

26

27

28

2930

3132

33

3435 36

37

38

39

40 41

42

43

44

45

46

4748

49

Figure 15 Network model of the SD_MAC System

Table 9 Scenario 1 nodes varied from 50 to 300 with mobility of nodes 5ms

Scenario 1 packet delivery ratio ()Nodes50

Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300

SWAN 69 76 79 885 895 912 935 955 965 977 99StAC 69 75 80 885 915 922 945 970 975 987 991DACME 70 78 86 895 895 912 945 969 975 987 991FuzzyQoS 75 88 91 95 96 97 98 985 989 996 998SD_MAC 88 94 97 975 985 985 987 991 992 998 999



Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300



found to be sim3ndashsim5 more compared to FuzzyQoSMoreover the PDR is sustained above 90 when the numberof node increases above 125 It has been noticed that even withdynamic change in topology (considering scenario 1 andscenario 2) the nodes were able to successfully deliver thepackets from source to destination and sustain high percentageof PDR compared to other existing counterparts +e fact thatthe LAT look-up process performed by each node during pathselection resulted in identifying one or more SCNs preventingthe node from waiting for the time slot in order to perform aLAT look up once again It is from the PDR observation thatthe path selectedwas successfully utilized for data transmissionin the proposed model making it most suitable scheme forcritical applications

42 Session Admission Ratio (SAR) It refers to the ratio ofdata sessions admitted into the network to the total numberrequesting admission Figure 17 shows the SAR of theprotocols at different node speed

SAR for StAC is high compared to others as increase innode speed impacts earlier admitted sessions making sourceto admit more sessions Even with high-node speedSD_MAC surpassed FuzzyQoS performance on an averageby sim18ndashsim22 DACME and SWAN by sim31 and sim35respectively +ough FuzzyQoS uses explicit notification tomake the source aware of any sudden change in topology itdoes not handle well the route failures making it to generatemore routing overheads to recover the data route

While DACME does periodic thorough assessment ofresources to identify path conditions during session ad-mission change in conditions requires reassessment toassure fairness in this approach degrading its SAR Even inhigh node speed occasions SD_MAC is able to sustain SAR

and successfully serve the admitted data sessions against otherapproaches +e fact being each node in SD_MAC has thecapability to select its SCNs by LAT look up and share amongeach other the type of services offered During probe requestit selects SCNs instantaneously reducing the overall timetaken during the probing sessionmaking it fast and successful

43 Session Completion Ratio (SCR) It refers to the ratio ofthe number of data sessions completed to the applicationrsquossatisfaction Figure 18 represents the Session Completion

50 100 150 200 250 30065

70

75

80

85

90

95

100

Number of nodes

Deli

very

rate

()

SD-MACFuzzyQoSDACME

StACSWAN

(a)

50 100 150 200 250 300Number of nodes

60

65

70

75

80

85

90

95

100

Pack

et d

eliv

ery

ratio

()

SD-MACFuzzyQoSDACME

StACSWAN

(b)

Figure 16 (a) Nodes vs PDR with node speed 5ms (b) Nodes vs PDR with node speed 30ms

5 10 15 20 25 30Maximum node speed (ms)

03

04

05

06

07

08

09

Sess

ion

adm

issio

n ra

tio

SWANStACDACME

FuzzyQoSSD-MAC

Figure 17 Session admission ratio for varying node speed


Ratio (SCR) of the protocols +e excessive retransmissionresults in switching the data session to another route orinitiate new route discovery

+e session drops if its requirements are not fulfilled It isclear from figure that the SCR of SD_MAC varies from 943to 768 by increasing speed from 2ms to 32ms while forStAC it varies from 412 to 314 respectively In case ofStAC admitted sessions are dropped due to link failuresaffecting earlier admitted session resulting in frequent routediscoveries SCR of DACME is higher than FuzzyQos even athigher node speed because it takes advantage of periodicassessment of end-to-end conditions during path probing andadmitting traffic only when the conditions are appropriateassuring high SCR for sessions admitted SCR of FuzzyQos islow compared to DACME and SD_MAC as it does not dealwell with route failures while higher speed cause frequentroute failure making SCR to degrade+e SD_MAC performsbetter at different node speed among all the protocols It isfast intelligent and controlled admission of data session helpsin achieving high SCR Obviously the neuro-multilayeredlearning process makes the whole network learn and adapt todynamic node behaviour making the system to make ac-curate decisions spontaneously Local route repair carried outthrough LAT look up helps to maintain the agreedthroughput to the data session

44 Packet Loss Ratio (PLR) It refers to the ratio of thenumber of lost packets to the total number of sent packets

In Figure 19 it is observed that compared to other ap-proaches the packet loss ratio for SWAN mechanism ishigher and increase from 15 to 107 when node speedchanges from 2ms to 32ms Higher node speed makes thepreprobed routes stale in SWAN resulting in high packet lossPLR for FuzzyQoS is low compared to SWAN StAC andDACME as it incorporates methodology to adapt to traffictransmission rate under dynamic conditions which reducesdata rate for affected session and admits it in new routes

Observation shows that SD_MAC has the lowest PLR andranges from 015 to 061 when node speed changes from2ms to 32ms +e multilayered neurofuzzy-incorporatedAC in SD_MAC switches data session from one route toanother when current route fails +e capability to sponta-neously identify SCNs and predict the nodersquos current resourceavailability makes the system to admit sessions faster and onmore appropriate routes It not only increases SCR inSD_MAC but also reduces the overall PLR respectively

45 Average End-to-End Delay It refers to the time slotelapsed between the time at which the packet is generated atits source and the time it is delivered to its destinationFigure 20 shows the average end-to-end delay caused withincrease in node speed for various approaches

Increase in node speed deteriorates the performance oflink causing interference resulting in packet loss route fail-ures or delayed transmission Route failures in StAC makethe source to reject the affected session and readmit thesession via the backup routes Frequent route failures not onlyinitiates route discovery frequently but also introduces moreoverheads and delay which is the case indentified in StAC+eobservation indicates the proposed SD_MAC systemrsquos effi-ciency in sustaining low delay making it more suitable forreal-time applications Prior to data traffic admission theNFAC engine enables accuracy in decision making ieselecting SCNs allocating and utilization of available re-sources not only assures successful guaranteed session ad-mission but also reduces the overall end-to-end delay

46AverageJroughput It refers to the total amount of datasuccessfully delivered to the destination per unit time

In most approaches frequent route-failures caused dueto high mobility makes nodes fail to deliver packets suc-cessfully to destination In Figure 21 the throughput is high

SWANStACDACME

FuzzyQoSSD-MAC


03

04

05

06

07

08

09

Sess

ion

com

plet

ion

ratio

Figure 18 Session completion ratio for varying node speed

SWANStACDACME

FuzzyQoSSD-MAC


0

005

01

015

Pack

et lo

ss ra

tioFigure 19 Packet loss ratio vs node speed


when node speed (NS) is low (NSlt 10ms) Optimalamount of throughput is achieved for all approaches whenNS is between 2ms to 10ms +e throughput for all ap-proaches drops when the node speed varies from sim10msndashsim15ms and then upholds

Observation in Figure 21 shows that SD_MAC hasmaintained its highest average throughput due to its highestSCR as it is able to assure and uphold the throughput of eachsession which has been guaranteed at the time of sessionadmission Although the average throughput of SD_MACdecreases with the increase in node speed it still maintainsthe guaranteed throughput at higher ratio for all the ad-mitted session in the network Nevertheless NFAC utilizes

the opportunities proactively by trasmitting more violentlyto achieve better throughput +e effectiveness of SD_MACapproach provide opportunities for more nodes completeadmitted sessions both in normal as well as in high mobilityand traffic constrained environment +e combination ofneural intelligence and fuzzy logic makes each node to self-organize and adapt to the dynamic changes making thenetwork effective in decision making enhancing QoS inMANET

5 Conclusions

Despite the rapid progress and the large volume of researchactivities made in the MANET almost all research areas stillimpose many open issues In this paper we have proposed anovel ldquoSoftware-Defined Multilayered Admission Control(SD_MAC) modelrdquo predominantly suitable for real-timemission critical applications Primarily the proposedscheme makes use of Neurofuzzy Inference-based Admis-sion Control service engine which dynamically regulatesreal-time data transmission and accurately allocates re-sources in the face of network dynamics such as mobility andtraffic overloads Our protocol enhancement proves to makenetwork more intelligent Simulation results proves theproposed approach outperforms the existing approachessuch as SWAN StAC DACME and FuzzyQoS in terms ofsession admission ratio session completion ratio and lowdropping probability assuring QoS for end users Our futurework relies on challenges imposed in limitation on terrainmodels with interference posed due to environmentalconditions

Data Availability

No data were used to support this study

Conflicts of Interest

+e authors declare that they have no conflicts of interest

References

[1] P Gupta and P R Kumar ldquo+e capacity of wireless net-worksrdquo IEEE Transactions on Information Jeory vol 46no 2 pp 388ndash404 2000

[2] S Chakrabarti and A Mishra ldquoQoS issues in ad hoc wirelessnetworksrdquo IEEE Communications Magazine vol 39 no 2pp 142ndash148 2001

[3] J U Duncombe ldquoInfrared navigation part I an assessment offeasibility (periodical style)rdquo IEEE Transactions on ElectronDevices vol 11 p 3439 1959

[4] S Chen and N Klara ldquoAn overview of quality-of-servicerouting for the next generation high-speed networks prob-lems and solutionsrdquo IEEE Network vol 12 no 6 pp 64ndash791998

[5] T Bheemarjun Reddy I Karthigeyan B S Manoj and C SivaRam Murthy ldquoQuality of service provisioning in ad-hocwireless networks a survey of issues and solutionsrdquo vol 4no 1 pp 83ndash124 2006

[6] M Karimi and D Pan ldquoChallanges for quality of service inmobile ad hoc networksrdquo in Proceedings of the IEEE 10th


0

005

01

015

02

025

03

Aver

age e

nd-to

-end

del

ay (s

)

SD-MACFuzzyQoSDACME

StACSWAN

Figure 20 Average end-to-end delay Vs node speed


0

50

100

150

200

250

300

350

400

Aver

age t

hrou

ghpu

t (kb

ps)

SD-MACFuzzyQoSDACME

StACSWAN

Figure 21 Average throughput vs node speed


AnnualWireless andMicrowave Technology Conference vol 4no 2 pp 129ndash153 Clearwater FL USA April 2009

[7] Y Zhang and N Ansari ldquoNew Jersey institute of technologyhiroshi tsoumada wireless telemedicine services over inte-grated IEEE 80211 WLAN and IEEE 80216 wi-max net-worksrdquo IEEE Wireless Communications vol 10pp 1536ndash1284 2010

[8] C R Linn and J S Liu ldquoQos routing in ad hoc wirelessnetworksrdquo IEEE Journal on Selected Areas on Communica-tions vol 17 no 8 pp 1426ndash1438 1998

[9] L Hanzo and R Tafazolli ldquo+e effects of shadow fading onQoS-aware routing and admission control protocols designedfor multi-hop MANETsrdquo Wireless Communications andMobile Computing vol 11 no 1 pp 1ndash22 2011

[10] H Luo S Lu V Bharghavan J Cheng and G Zhong ldquoApacket scheduling approach to QoS support in multi-hopwireless networksrdquo Mobile Networks and Applications vol 9no 3 pp 193ndash206 2004

[11] A Bouchti A Haqiq and S Kafhali ldquoAnalysis of quality ofservice performances of connection admission controlmechanisms in OFDMA IEEE 80216 network using BMAPqueuingrdquo International Journal of Computer Science vol 9no 1-2 pp 302ndash310 2012

[12] Y Dong and P Du ldquoCross-layer design of 2D queuing modelfor multi-hop wireless networksrdquo Wireless Personal Com-munications vol 77 no 3 pp 1815ndash1832 2014

[13] V Vidhyasanker B S Manoj and C Siva Ram Murthy ldquoSlotallocation schemes for delay sensitive traffic support inasynchronous wireless mesh networksrdquo Je InternationalJournal of Computer and Telecommunications Networkingvol 50 no 15 pp 2595ndash2613 2006

[14] Y Lin and V W S Wong ldquoAn admission control algorithmfor multi-hop 80211e based WLANsrdquo Computer Commu-nication vol 31 no 14 pp 3510ndash3520 2008

[15] L Hanzo and R Tafazolli ldquoAdmission control schemes for80211-based multi-hop mobile ad hoc networks a surveyrdquoIEEE Communications Surveys amp Tutorials vol 11 no 4pp 78ndash108 2009

[16] G Ash and D McDysan ldquoGeneric connection admissioncontrol (GCAC) algorithm specification for IPMPLS net-worksrdquo RFC 6601 Internet Engineering Task Force FremontCA USA RFC 6601 2012

[17] C Wang W-J Yan and H-K Lo ldquoDynamic admissioncontrol and bandwidth reservation for IEEE 80216e mobileWi-MAX networksrdquo EURASIP Journal on Wireless Com-munications and Networking vol 2012 no 1 2012

[18] G H Ahn A T Campbell A Veres and L H Sun ldquoSWANservice differentiation in stateless wireless ad hoc networksrdquoin Proceedings of the 21st Annual Joint Conference of the IEEEComputer and Communications Societies pp 457ndash466NewYork NY USA June 2002

[19] M Asif Z Sun and H Cruickshank ldquoAdmission controlprotocol in mobile ad-hoc networks provisioning QoSrdquo inProceedings of the International Conference on Frontiers ofInformation Technology Abbottabad Pakistan December2009

[20] C T Calafate M P Malumbres J Oliver J C Cano andP Manzoni ldquoQoS support in MANETs a modular archi-tecture based on the IEEE 80211e technologyrdquo IEEETransactions on Circuits and Systems for Video Technologyvol 19 no 5 pp 678ndash692 2009

[21] B Zhang andH TMouftah ldquoQoS Routing for wireless ad hocnetworks problems algorithms and protocolsrdquo IEEE Com-munications Magazine vol 43 no 10 pp 110ndash115 2005

[22] L Khoukhi and S Cherkaoui ldquoIntelligent QoS managementfor multimedia services support in wireless mobile ad hocnetworksrdquo Computer Networks vol 54 no 10 pp 1692ndash17062010

[23] A Pandey M D Nasir Ahmed N Kumar and P Gupta ldquoAhybrid routing scheme for mobile ad-hoc networks withmobile backbonesrdquo in Lecture Notes in Computer Sciencevol 4297 pp 411ndash423 Springer Berlin Germany 2006

[24] A Gnana Soundari and V L Jyothi ldquoEnergy efficient machinelearning technique for smart data collection in wireless sensornetworksrdquo Circuits Systems and Signal Processing pp 1ndash342019

[25] F Cadger K Curran J Santos and S Moffet ldquoOpportunisticneighbour prediction using an artificial neural networkrdquoInternational Journal of Advanced Pervasive and UbiquitousComputing vol 7 no 2 pp 38ndash50 2015

[26] V Nair and E G Hinton ldquoRectified linear units improverestricted Boltzmann machinesrdquo in Proceedings of the 27thInternational Conference on International Conference onMachine Learning pp 807ndash814 Haifa Israel June 2010

[27] X Glorot A Bordes and Y Bengio ldquoDeep sparse rectifierneural networksrdquo in Proceedings of the 14th InternationalConference on Artificial Intelligence and Statistics PMLRvol 15 pp 315ndash323 Ft Lauderdale FL USA 2011

[28] L Kleinrock Queueing Systems Vol 2 Wiley Hoboken NJUSA 1975

[29] A K Parekh and R G Gallager ldquoA generalized processorsharing approach to flow control in intergrated servicesnetworks the single-node caserdquo IEEEACM Transactions onNetworking vol 1 no 3 pp 344ndash357 1993

[30] M Katevenis S Sidiropoulos and C CourcoubetisldquoWeighted round-Robin cell multiplexing in a general pur-pose ATM switch chiprdquo IEEE Journal on Selected Areas inCommunications vol 9 no 8 pp 1265ndash1279 1991

[31] N Arora and S Kaur ldquoPerformance evaluation and faultrepair mechanism for resilience of MPLS RSVPTE networkwith different application traffic load balancing across labelswitched pathrdquo International Journal of Engineering Tech-nology Science and Research vol 2 no 11 pp 28ndash33 2015

[32] N N Kale and S A Waichol ldquoPerformance analysis of MPLSnetwork with traditional IP network in service provider en-vironmentrdquo International Journal of Advanced Research inComputer Engineering amp Technology vol 3 no 4 pp 1311ndash1316 2014

[33] A R Sulaiman and O SalihAlhafidh ldquoPerformance analysisof multimedia traffic over MPLS communication networkswith traffic engineeringrdquo International Journal of ComputerNetworks and Communications Security vol 2 pp 93ndash1012014

[34] M Kumar and S Sangal ldquoImproving the usage of networkresources using MPLS traffic engineering (TE)rdquo InternationalJournal of Current Engineering and Technology vol 5 no 1pp 261ndash265 2015

[35] E Kafhaliabdelali and A KrimHaqiq ldquoPerformance analysisfor bandwidth allocation in IEEE 80216 broadband wirelessnetworks using BMAP queuingrdquo International Journal ofWireless amp Mobile Networks vol 4 no 1 pp 139ndash154 2012


+e Admission Control (AC) [14ndash17] is considered asthe most crucial mechanism for providing QoS in MANETIts core functionality is to estimate the nodersquos state based onresource availability and decide which applicationrsquos dataflow can be admitted along which node such that the QoS asper user requirement is met without violating previouslymade guarantees

To combat the aforementioned challenges and developan enhanced technique to effectively manage network re-sources for assuring QoS in MANET in this paper novelneurofuzzy logic-based Software-Defined MultilayeredAdmission Control (SD_MAC) model is proposed Aim ofthe proposed model is to assure guaranteed QoS support forreal-time traffic It exploits the fuzzy logic to ensure trafficadaptation characterized by dynamic changes (making itadequate to the uncertainty and the information incom-pleteness) in MANET environment +e model also incor-porates the benefits of Neural Networks (NN) to support themultimedia service in ad-hoc networks +e self-adaptivedata driven mechanism of the neural network allows fastlearning of different operations performed by the corefunctional blocks of the system thereby permitting a besttraffic class selection and delivery according to flow re-quirements After learning the data presented to them itoffers the capability to generalize ie it often predicts anevent occurrence even if the input data is noisy and theirunderlying relationships are unknown or difficult to de-scribe +e most important feature of the NN is that it canapproximate multivariate functions with superior predictionaccuracy and low computational complexity It can pri-marily be used when limited data sets are available and therelationships between them are vague Overall NN exploitsthe capability to learn and adapt to the system behaviourmore quickly by integrating the operations of variouscomponents (both in MAC and network layer) to make thenetwork self-organized and artificially intelligent

+e proposed SD_MAC system makes use of theNeurofuzzy Inference-based Admission Control (NFAC)service engine whose core functionality is to

(i) Evaluate the resource availability by using QoSparameters such as bandwidth queue load andRSSI in each node to generate the node status+e Fuzzy Inference-based Admission Control(FIAC) module in the NFAC engine is triggeredautomatically in each node to generate the nodestatus

(ii) Derive the prediction index which solely decideswhether the admission control session can be ad-mitted or rejected via the node for the requestedQoS

(iii) Self-learn the behavioural functionalities and trainand test the network to make it self-organized andartificially intelligent for effective decision makingusing Neuro-Multilayered Learning (NML) process

Our proposed methodology was modelled using pro-totype developed using MATLAB +e performance of theproposed model is evaluated by comparing it with other

existing approaches such as SWAN StAC DACME andFuzzyQoS +e results showed that the service exhibited byNFAC engine tends to automate the working procedure ofadmission control to achieve good throughput by selectingstrongly connected neighbors and reliable path for longerlive transmission

+e rest of the document is organized as follows Section2 presents related work Section 3 illustrates our proposedSD_MAC system Section 4 depicts the simulation resultsand analysis and finally Section 5 concludes the paper alongwith a discussion on future directions

2 Related Works

+is section describes briefly some of the main admissioncontrol model proposed for wireless MANET

A stateless service differentiation AC model (SWAN)[18] performs real-time traffic control using sender-basedadmission control by distinguishing real-time and best-ef-fort traffic +e source probes the path to the destinationwhen it wants to send a real-time traffic to destinationDuring the probe it identifies the available bandwidth andrelies on feedback received from the MAC layer by using therate and source-based admission control approach Eachnode uses the rate control algorithm to control the best-effort traffic +e bandwidth used for best-effort traffic isrestricted so that real-time applications can exploit the re-quired bandwidth the bandwidth not used by real-timeapplications can be exploited by the best-effort traffic +eprobe request in SWAN causes a lot of overhead and packetloss

In Staggered Admission Control protocol (StAC) [19]the source sends the session request packet +e session isrejected if the network layer does not support the flow else itbroadcasts the request Intermediate nodes add informationto request after verifying their local resources +e desti-nation receives the request and sends the response Sourcecaches all the route information and then sends an ad-mission request packet to two hop neighbors which sends anadmission denied back if it cannot support new informationflow Based on the observation the source transmits data at avery low rate and gradually increases the flow rate up to therequired level within the specified time During the gradualincreasing time if it is affecting the earlier admitted sessionsthen a session may be rejected Presence of interferencecauses link failures resulting in frequent route discoveries+is deteriorates the performance in StAC by incurring highrouting overheads and end-to-end delay

A distributed admission control for MANET environ-ments (DACME) [20] approach periodically assesses end-to-end conditions on the path Source node performs pathprobing to obtain various QoS measurements to assuretransmission of data traffic under good conditions +edestination responds to probes giving the source feedbackabout path conditions+ough this method tends to performwell in small to medium network taking advantage of theIEEE 80211e standard [21] it suffers from fairness issueunder high network load



























PlatinumGoldSilver



layer

MAC

Classifier

User request

MAC layer



Routing service




Testing Training



Traffic regulator


control (FIAC)

Defuzzification


Fuzzification

Fuzzy rule base

Nstatus


Scheduler





P_beacon











































end ifend if

end if




NStatus NSHlow


NStatus NSMhigh


NStatus NSLmoderate
























packets


Inference engine



























PTypeQLoad ()





Mem

bers

hip

degr

ee

Low

0

05

1High


Medium


Mem

bers

hip

degr

ee

Low

0

05

1High


Medium


Mem

bers

hip

degr

ee

Low

0

05

1High


Medium


Mem

bership

degree

Low

0

05

1High

0 50 70 10030RSSI

Medium


Mem

bers

hip

degr

ee

MediumLow

0

1

Bandwidth available

High

0 10 28 40 50

08

03







For example












PTypeQLoad ()

RSSI BWavail

NodestatusNstatus

QLHP QLMP QLLP

HP (0) 42 (M) 35 (L) 63 (L) 67 (H) 28 (L) 79 NSHmod








α + β + c 1 (4)








aij σ 1113944n


⎛⎝ ⎞⎠ (7)


MSE 1113944n

j1aLj1113872 1113873 (8)


MSE (05 minus 00)2

+(02 minus 00)2

+(03 minus 00)2

+(04 minus 00)2

094(9)













α = 08β = 01 γ = 01

Number of nodes


Packet forward rate





α β c

1 0 00 0 101 0 0902 0 0803 0 0704 0 0605 0 0506 0 0407 0 0308 0 02











Case 7

Case 9

Case 13

Case 21

Case 4Case 3Case 2

1

Number of neighbors

Number of nodes

Mobility


Packet forward rate

1 1 1

Case 1

Case 5Case 6

Case 8

Case 11Case 10

Case 12

Case 19


Case 17Case 18

Case 20

Case 22

Case 36








AC_list Iiend






































HPMPLP

IPlayer

Classifier



flow admission

Routing service


Traffic regulator

Scheduler

MAC








a QLHP


b QLMP


c QLLP


(10)


wHPa wMPb wLPc (11)











HP (high)

LP

ClassifyMP (medium)

LP (low)

60

30

10

Incomingpackets

HP MP








0 10 20 30 40 50 60 70 80 90 1000

10

20

30

40

50

60

70

80

90

100

501

23

45

6

7

8

9

10

11

12

13

14

1516

17

1819

20

21

22

23

24

25

26

2728

29

30

31

32

3334

35

36

3738

39

40

4142

43

44

45

46

47

48

49








0 20 40 60 80 1000

10

20

30

40

50

60

70

80

90

100

50

1

2

3

4

5

6

7

8

9

10

11

12 13

14

1516

17

18

19

20

21

22

23

24

25

26

27

28

2930

3132

33

3435 36

37

38

39

40 41

42

43

44

45

46

4748

49




Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300




Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300









50 100 150 200 250 30065

70

75

80

85

90

95

100

Number of nodes

Deli

very

rate

()

SD-MACFuzzyQoSDACME

StACSWAN

(a)

50 100 150 200 250 300Number of nodes

60

65

70

75

80

85

90

95

100

Pack

et d

eliv

ery

ratio

()

SD-MACFuzzyQoSDACME

StACSWAN

(b)



03

04

05

06

07

08

09

Sess

ion

adm

issio

n ra

tio

SWANStACDACME

FuzzyQoSSD-MAC












SWANStACDACME

FuzzyQoSSD-MAC


03

04

05

06

07

08

09

Sess

ion

com

plet

ion

ratio


SWANStACDACME

FuzzyQoSSD-MAC


0

005

01

015

Pack

et lo

ss ra






5 Conclusions


Data Availability




References








0

005

01

015

02

025

03

Aver

age e

nd-to

-end

del

ay (s

)

SD-MACFuzzyQoSDACME

StACSWAN



0

50

100

150

200

250

300

350

400

Aver

age t

hrou

ghpu

t (kb

ps)

SD-MACFuzzyQoSDACME

StACSWAN



























































PlatinumGoldSilver



layer

MAC

Classifier

User request

MAC layer



Routing service




Testing Training



Traffic regulator


control (FIAC)

Defuzzification


Fuzzification

Fuzzy rule base

Nstatus


Scheduler





P_beacon











































end ifend if

end if




NStatus NSHlow


NStatus NSMhigh


NStatus NSLmoderate
























packets


Inference engine



























PTypeQLoad ()





Mem

bers

hip

degr

ee

Low

0

05

1High


Medium


Mem

bers

hip

degr

ee

Low

0

05

1High


Medium


Mem

bers

hip

degr

ee

Low

0

05

1High


Medium


Mem

bership

degree

Low

0

05

1High

0 50 70 10030RSSI

Medium


Mem

bers

hip

degr

ee

MediumLow

0

1

Bandwidth available

High

0 10 28 40 50

08

03







For example












PTypeQLoad ()

RSSI BWavail

NodestatusNstatus

QLHP QLMP QLLP

HP (0) 42 (M) 35 (L) 63 (L) 67 (H) 28 (L) 79 NSHmod








α + β + c 1 (4)








aij σ 1113944n


⎛⎝ ⎞⎠ (7)


MSE 1113944n

j1aLj1113872 1113873 (8)


MSE (05 minus 00)2

+(02 minus 00)2

+(03 minus 00)2

+(04 minus 00)2

094(9)













α = 08β = 01 γ = 01

Number of nodes


Packet forward rate





α β c

1 0 00 0 101 0 0902 0 0803 0 0704 0 0605 0 0506 0 0407 0 0308 0 02











Case 7

Case 9

Case 13

Case 21

Case 4Case 3Case 2

1

Number of neighbors

Number of nodes

Mobility


Packet forward rate

1 1 1

Case 1

Case 5Case 6

Case 8

Case 11Case 10

Case 12

Case 19


Case 17Case 18

Case 20

Case 22

Case 36








AC_list Iiend






































HPMPLP

IPlayer

Classifier



flow admission

Routing service


Traffic regulator

Scheduler

MAC








a QLHP


b QLMP


c QLLP


(10)


wHPa wMPb wLPc (11)











HP (high)

LP

ClassifyMP (medium)

LP (low)

60

30

10

Incomingpackets

HP MP








0 10 20 30 40 50 60 70 80 90 1000

10

20

30

40

50

60

70

80

90

100

501

23

45

6

7

8

9

10

11

12

13

14

1516

17

1819

20

21

22

23

24

25

26

2728

29

30

31

32

3334

35

36

3738

39

40

4142

43

44

45

46

47

48

49








0 20 40 60 80 1000

10

20

30

40

50

60

70

80

90

100

50

1

2

3

4

5

6

7

8

9

10

11

12 13

14

1516

17

18

19

20

21

22

23

24

25

26

27

28

2930

3132

33

3435 36

37

38

39

40 41

42

43

44

45

46

4748

49




Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300




Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300









50 100 150 200 250 30065

70

75

80

85

90

95

100

Number of nodes

Deli

very

rate

()

SD-MACFuzzyQoSDACME

StACSWAN

(a)

50 100 150 200 250 300Number of nodes

60

65

70

75

80

85

90

95

100

Pack

et d

eliv

ery

ratio

()

SD-MACFuzzyQoSDACME

StACSWAN

(b)



03

04

05

06

07

08

09

Sess

ion

adm

issio

n ra

tio

SWANStACDACME

FuzzyQoSSD-MAC












SWANStACDACME

FuzzyQoSSD-MAC


03

04

05

06

07

08

09

Sess

ion

com

plet

ion

ratio


SWANStACDACME

FuzzyQoSSD-MAC


0

005

01

015

Pack

et lo

ss ra






5 Conclusions


Data Availability




References








0

005

01

015

02

025

03

Aver

age e

nd-to

-end

del

ay (s

)

SD-MACFuzzyQoSDACME

StACSWAN



0

50

100

150

200

250

300

350

400

Aver

age t

hrou

ghpu

t (kb

ps)

SD-MACFuzzyQoSDACME

StACSWAN















































PlatinumGoldSilver



layer

MAC

Classifier

User request

MAC layer



Routing service




Testing Training



Traffic regulator


control (FIAC)

Defuzzification


Fuzzification

Fuzzy rule base

Nstatus


Scheduler





P_beacon











































end ifend if

end if




NStatus NSHlow


NStatus NSMhigh


NStatus NSLmoderate
























packets


Inference engine



























PTypeQLoad ()





Mem

bers

hip

degr

ee

Low

0

05

1High


Medium


Mem

bers

hip

degr

ee

Low

0

05

1High


Medium


Mem

bers

hip

degr

ee

Low

0

05

1High


Medium


Mem

bership

degree

Low

0

05

1High

0 50 70 10030RSSI

Medium


Mem

bers

hip

degr

ee

MediumLow

0

1

Bandwidth available

High

0 10 28 40 50

08

03







For example












PTypeQLoad ()

RSSI BWavail

NodestatusNstatus

QLHP QLMP QLLP

HP (0) 42 (M) 35 (L) 63 (L) 67 (H) 28 (L) 79 NSHmod








α + β + c 1 (4)








aij σ 1113944n


⎛⎝ ⎞⎠ (7)


MSE 1113944n

j1aLj1113872 1113873 (8)


MSE (05 minus 00)2

+(02 minus 00)2

+(03 minus 00)2

+(04 minus 00)2

094(9)













α = 08β = 01 γ = 01

Number of nodes


Packet forward rate





α β c

1 0 00 0 101 0 0902 0 0803 0 0704 0 0605 0 0506 0 0407 0 0308 0 02











Case 7

Case 9

Case 13

Case 21

Case 4Case 3Case 2

1

Number of neighbors

Number of nodes

Mobility


Packet forward rate

1 1 1

Case 1

Case 5Case 6

Case 8

Case 11Case 10

Case 12

Case 19


Case 17Case 18

Case 20

Case 22

Case 36








AC_list Iiend






































HPMPLP

IPlayer

Classifier



flow admission

Routing service


Traffic regulator

Scheduler

MAC








a QLHP


b QLMP


c QLLP


(10)


wHPa wMPb wLPc (11)











HP (high)

LP

ClassifyMP (medium)

LP (low)

60

30

10

Incomingpackets

HP MP








0 10 20 30 40 50 60 70 80 90 1000

10

20

30

40

50

60

70

80

90

100

501

23

45

6

7

8

9

10

11

12

13

14

1516

17

1819

20

21

22

23

24

25

26

2728

29

30

31

32

3334

35

36

3738

39

40

4142

43

44

45

46

47

48

49








0 20 40 60 80 1000

10

20

30

40

50

60

70

80

90

100

50

1

2

3

4

5

6

7

8

9

10

11

12 13

14

1516

17

18

19

20

21

22

23

24

25

26

27

28

2930

3132

33

3435 36

37

38

39

40 41

42

43

44

45

46

4748

49




Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300




Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300









50 100 150 200 250 30065

70

75

80

85

90

95

100

Number of nodes

Deli

very

rate

()

SD-MACFuzzyQoSDACME

StACSWAN

(a)

50 100 150 200 250 300Number of nodes

60

65

70

75

80

85

90

95

100

Pack

et d

eliv

ery

ratio

()

SD-MACFuzzyQoSDACME

StACSWAN

(b)



03

04

05

06

07

08

09

Sess

ion

adm

issio

n ra

tio

SWANStACDACME

FuzzyQoSSD-MAC












SWANStACDACME

FuzzyQoSSD-MAC


03

04

05

06

07

08

09

Sess

ion

com

plet

ion

ratio


SWANStACDACME

FuzzyQoSSD-MAC


0

005

01

015

Pack

et lo

ss ra






5 Conclusions


Data Availability




References








0

005

01

015

02

025

03

Aver

age e

nd-to

-end

del

ay (s

)

SD-MACFuzzyQoSDACME

StACSWAN



0

50

100

150

200

250

300

350

400

Aver

age t

hrou

ghpu

t (kb

ps)

SD-MACFuzzyQoSDACME

StACSWAN




































P_beacon











































end ifend if

end if




NStatus NSHlow


NStatus NSMhigh


NStatus NSLmoderate
























packets


Inference engine



























PTypeQLoad ()





Mem

bers

hip

degr

ee

Low

0

05

1High


Medium


Mem

bers

hip

degr

ee

Low

0

05

1High


Medium


Mem

bers

hip

degr

ee

Low

0

05

1High


Medium


Mem

bership

degree

Low

0

05

1High

0 50 70 10030RSSI

Medium


Mem

bers

hip

degr

ee

MediumLow

0

1

Bandwidth available

High

0 10 28 40 50

08

03







For example












PTypeQLoad ()

RSSI BWavail

NodestatusNstatus

QLHP QLMP QLLP

HP (0) 42 (M) 35 (L) 63 (L) 67 (H) 28 (L) 79 NSHmod








α + β + c 1 (4)








aij σ 1113944n


⎛⎝ ⎞⎠ (7)


MSE 1113944n

j1aLj1113872 1113873 (8)


MSE (05 minus 00)2

+(02 minus 00)2

+(03 minus 00)2

+(04 minus 00)2

094(9)













α = 08β = 01 γ = 01

Number of nodes


Packet forward rate





α β c

1 0 00 0 101 0 0902 0 0803 0 0704 0 0605 0 0506 0 0407 0 0308 0 02











Case 7

Case 9

Case 13

Case 21

Case 4Case 3Case 2

1

Number of neighbors

Number of nodes

Mobility


Packet forward rate

1 1 1

Case 1

Case 5Case 6

Case 8

Case 11Case 10

Case 12

Case 19


Case 17Case 18

Case 20

Case 22

Case 36








AC_list Iiend






































HPMPLP

IPlayer

Classifier



flow admission

Routing service


Traffic regulator

Scheduler

MAC








a QLHP


b QLMP


c QLLP


(10)


wHPa wMPb wLPc (11)











HP (high)

LP

ClassifyMP (medium)

LP (low)

60

30

10

Incomingpackets

HP MP








0 10 20 30 40 50 60 70 80 90 1000

10

20

30

40

50

60

70

80

90

100

501

23

45

6

7

8

9

10

11

12

13

14

1516

17

1819

20

21

22

23

24

25

26

2728

29

30

31

32

3334

35

36

3738

39

40

4142

43

44

45

46

47

48

49








0 20 40 60 80 1000

10

20

30

40

50

60

70

80

90

100

50

1

2

3

4

5

6

7

8

9

10

11

12 13

14

1516

17

18

19

20

21

22

23

24

25

26

27

28

2930

3132

33

3435 36

37

38

39

40 41

42

43

44

45

46

4748

49




Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300




Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300









50 100 150 200 250 30065

70

75

80

85

90

95

100

Number of nodes

Deli

very

rate

()

SD-MACFuzzyQoSDACME

StACSWAN

(a)

50 100 150 200 250 300Number of nodes

60

65

70

75

80

85

90

95

100

Pack

et d

eliv

ery

ratio

()

SD-MACFuzzyQoSDACME

StACSWAN

(b)



03

04

05

06

07

08

09

Sess

ion

adm

issio

n ra

tio

SWANStACDACME

FuzzyQoSSD-MAC












SWANStACDACME

FuzzyQoSSD-MAC


03

04

05

06

07

08

09

Sess

ion

com

plet

ion

ratio


SWANStACDACME

FuzzyQoSSD-MAC


0

005

01

015

Pack

et lo

ss ra






5 Conclusions


Data Availability




References








0

005

01

015

02

025

03

Aver

age e

nd-to

-end

del

ay (s

)

SD-MACFuzzyQoSDACME

StACSWAN



0

50

100

150

200

250

300

350

400

Aver

age t

hrou

ghpu

t (kb

ps)

SD-MACFuzzyQoSDACME

StACSWAN
































































end ifend if

end if




NStatus NSHlow


NStatus NSMhigh


NStatus NSLmoderate
























packets


Inference engine



























PTypeQLoad ()





Mem

bers

hip

degr

ee

Low

0

05

1High


Medium


Mem

bers

hip

degr

ee

Low

0

05

1High


Medium


Mem

bers

hip

degr

ee

Low

0

05

1High


Medium


Mem

bership

degree

Low

0

05

1High

0 50 70 10030RSSI

Medium


Mem

bers

hip

degr

ee

MediumLow

0

1

Bandwidth available

High

0 10 28 40 50

08

03







For example












PTypeQLoad ()

RSSI BWavail

NodestatusNstatus

QLHP QLMP QLLP

HP (0) 42 (M) 35 (L) 63 (L) 67 (H) 28 (L) 79 NSHmod








α + β + c 1 (4)








aij σ 1113944n


⎛⎝ ⎞⎠ (7)


MSE 1113944n

j1aLj1113872 1113873 (8)


MSE (05 minus 00)2

+(02 minus 00)2

+(03 minus 00)2

+(04 minus 00)2

094(9)













α = 08β = 01 γ = 01

Number of nodes


Packet forward rate





α β c

1 0 00 0 101 0 0902 0 0803 0 0704 0 0605 0 0506 0 0407 0 0308 0 02











Case 7

Case 9

Case 13

Case 21

Case 4Case 3Case 2

1

Number of neighbors

Number of nodes

Mobility


Packet forward rate

1 1 1

Case 1

Case 5Case 6

Case 8

Case 11Case 10

Case 12

Case 19


Case 17Case 18

Case 20

Case 22

Case 36








AC_list Iiend






































HPMPLP

IPlayer

Classifier



flow admission

Routing service


Traffic regulator

Scheduler

MAC








a QLHP


b QLMP


c QLLP


(10)


wHPa wMPb wLPc (11)











HP (high)

LP

ClassifyMP (medium)

LP (low)

60

30

10

Incomingpackets

HP MP








0 10 20 30 40 50 60 70 80 90 1000

10

20

30

40

50

60

70

80

90

100

501

23

45

6

7

8

9

10

11

12

13

14

1516

17

1819

20

21

22

23

24

25

26

2728

29

30

31

32

3334

35

36

3738

39

40

4142

43

44

45

46

47

48

49








0 20 40 60 80 1000

10

20

30

40

50

60

70

80

90

100

50

1

2

3

4

5

6

7

8

9

10

11

12 13

14

1516

17

18

19

20

21

22

23

24

25

26

27

28

2930

3132

33

3435 36

37

38

39

40 41

42

43

44

45

46

4748

49




Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300




Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300









50 100 150 200 250 30065

70

75

80

85

90

95

100

Number of nodes

Deli

very

rate

()

SD-MACFuzzyQoSDACME

StACSWAN

(a)

50 100 150 200 250 300Number of nodes

60

65

70

75

80

85

90

95

100

Pack

et d

eliv

ery

ratio

()

SD-MACFuzzyQoSDACME

StACSWAN

(b)



03

04

05

06

07

08

09

Sess

ion

adm

issio

n ra

tio

SWANStACDACME

FuzzyQoSSD-MAC












SWANStACDACME

FuzzyQoSSD-MAC


03

04

05

06

07

08

09

Sess

ion

com

plet

ion

ratio


SWANStACDACME

FuzzyQoSSD-MAC


0

005

01

015

Pack

et lo

ss ra






5 Conclusions


Data Availability




References








0

005

01

015

02

025

03

Aver

age e

nd-to

-end

del

ay (s

)

SD-MACFuzzyQoSDACME

StACSWAN



0

50

100

150

200

250

300

350

400

Aver

age t

hrou

ghpu

t (kb

ps)

SD-MACFuzzyQoSDACME

StACSWAN














































end ifend if

end if




NStatus NSHlow


NStatus NSMhigh


NStatus NSLmoderate
























packets


Inference engine



























PTypeQLoad ()





Mem

bers

hip

degr

ee

Low

0

05

1High


Medium


Mem

bers

hip

degr

ee

Low

0

05

1High


Medium


Mem

bers

hip

degr

ee

Low

0

05

1High


Medium


Mem

bership

degree

Low

0

05

1High

0 50 70 10030RSSI

Medium


Mem

bers

hip

degr

ee

MediumLow

0

1

Bandwidth available

High

0 10 28 40 50

08

03







For example












PTypeQLoad ()

RSSI BWavail

NodestatusNstatus

QLHP QLMP QLLP

HP (0) 42 (M) 35 (L) 63 (L) 67 (H) 28 (L) 79 NSHmod








α + β + c 1 (4)








aij σ 1113944n


⎛⎝ ⎞⎠ (7)


MSE 1113944n

j1aLj1113872 1113873 (8)


MSE (05 minus 00)2

+(02 minus 00)2

+(03 minus 00)2

+(04 minus 00)2

094(9)













α = 08β = 01 γ = 01

Number of nodes


Packet forward rate





α β c

1 0 00 0 101 0 0902 0 0803 0 0704 0 0605 0 0506 0 0407 0 0308 0 02











Case 7

Case 9

Case 13

Case 21

Case 4Case 3Case 2

1

Number of neighbors

Number of nodes

Mobility


Packet forward rate

1 1 1

Case 1

Case 5Case 6

Case 8

Case 11Case 10

Case 12

Case 19


Case 17Case 18

Case 20

Case 22

Case 36








AC_list Iiend






































HPMPLP

IPlayer

Classifier



flow admission

Routing service


Traffic regulator

Scheduler

MAC








a QLHP


b QLMP


c QLLP


(10)


wHPa wMPb wLPc (11)











HP (high)

LP

ClassifyMP (medium)

LP (low)

60

30

10

Incomingpackets

HP MP








0 10 20 30 40 50 60 70 80 90 1000

10

20

30

40

50

60

70

80

90

100

501

23

45

6

7

8

9

10

11

12

13

14

1516

17

1819

20

21

22

23

24

25

26

2728

29

30

31

32

3334

35

36

3738

39

40

4142

43

44

45

46

47

48

49








0 20 40 60 80 1000

10

20

30

40

50

60

70

80

90

100

50

1

2

3

4

5

6

7

8

9

10

11

12 13

14

1516

17

18

19

20

21

22

23

24

25

26

27

28

2930

3132

33

3435 36

37

38

39

40 41

42

43

44

45

46

4748

49




Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300




Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300









50 100 150 200 250 30065

70

75

80

85

90

95

100

Number of nodes

Deli

very

rate

()

SD-MACFuzzyQoSDACME

StACSWAN

(a)

50 100 150 200 250 300Number of nodes

60

65

70

75

80

85

90

95

100

Pack

et d

eliv

ery

ratio

()

SD-MACFuzzyQoSDACME

StACSWAN

(b)



03

04

05

06

07

08

09

Sess

ion

adm

issio

n ra

tio

SWANStACDACME

FuzzyQoSSD-MAC












SWANStACDACME

FuzzyQoSSD-MAC


03

04

05

06

07

08

09

Sess

ion

com

plet

ion

ratio


SWANStACDACME

FuzzyQoSSD-MAC


0

005

01

015

Pack

et lo

ss ra






5 Conclusions


Data Availability




References








0

005

01

015

02

025

03

Aver

age e

nd-to

-end

del

ay (s

)

SD-MACFuzzyQoSDACME

StACSWAN



0

50

100

150

200

250

300

350

400

Aver

age t

hrou

ghpu

t (kb

ps)

SD-MACFuzzyQoSDACME

StACSWAN



































NStatus NSHlow


NStatus NSMhigh


NStatus NSLmoderate
























packets


Inference engine



























PTypeQLoad ()





Mem

bers

hip

degr

ee

Low

0

05

1High


Medium


Mem

bers

hip

degr

ee

Low

0

05

1High


Medium


Mem

bers

hip

degr

ee

Low

0

05

1High


Medium


Mem

bership

degree

Low

0

05

1High

0 50 70 10030RSSI

Medium


Mem

bers

hip

degr

ee

MediumLow

0

1

Bandwidth available

High

0 10 28 40 50

08

03







For example












PTypeQLoad ()

RSSI BWavail

NodestatusNstatus

QLHP QLMP QLLP

HP (0) 42 (M) 35 (L) 63 (L) 67 (H) 28 (L) 79 NSHmod








α + β + c 1 (4)








aij σ 1113944n


⎛⎝ ⎞⎠ (7)


MSE 1113944n

j1aLj1113872 1113873 (8)


MSE (05 minus 00)2

+(02 minus 00)2

+(03 minus 00)2

+(04 minus 00)2

094(9)













α = 08β = 01 γ = 01

Number of nodes


Packet forward rate





α β c

1 0 00 0 101 0 0902 0 0803 0 0704 0 0605 0 0506 0 0407 0 0308 0 02











Case 7

Case 9

Case 13

Case 21

Case 4Case 3Case 2

1

Number of neighbors

Number of nodes

Mobility


Packet forward rate

1 1 1

Case 1

Case 5Case 6

Case 8

Case 11Case 10

Case 12

Case 19


Case 17Case 18

Case 20

Case 22

Case 36








AC_list Iiend






































HPMPLP

IPlayer

Classifier



flow admission

Routing service


Traffic regulator

Scheduler

MAC








a QLHP


b QLMP


c QLLP


(10)


wHPa wMPb wLPc (11)











HP (high)

LP

ClassifyMP (medium)

LP (low)

60

30

10

Incomingpackets

HP MP








0 10 20 30 40 50 60 70 80 90 1000

10

20

30

40

50

60

70

80

90

100

501

23

45

6

7

8

9

10

11

12

13

14

1516

17

1819

20

21

22

23

24

25

26

2728

29

30

31

32

3334

35

36

3738

39

40

4142

43

44

45

46

47

48

49








0 20 40 60 80 1000

10

20

30

40

50

60

70

80

90

100

50

1

2

3

4

5

6

7

8

9

10

11

12 13

14

1516

17

18

19

20

21

22

23

24

25

26

27

28

2930

3132

33

3435 36

37

38

39

40 41

42

43

44

45

46

4748

49




Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300




Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300









50 100 150 200 250 30065

70

75

80

85

90

95

100

Number of nodes

Deli

very

rate

()

SD-MACFuzzyQoSDACME

StACSWAN

(a)

50 100 150 200 250 300Number of nodes

60

65

70

75

80

85

90

95

100

Pack

et d

eliv

ery

ratio

()

SD-MACFuzzyQoSDACME

StACSWAN

(b)



03

04

05

06

07

08

09

Sess

ion

adm

issio

n ra

tio

SWANStACDACME

FuzzyQoSSD-MAC












SWANStACDACME

FuzzyQoSSD-MAC


03

04

05

06

07

08

09

Sess

ion

com

plet

ion

ratio


SWANStACDACME

FuzzyQoSSD-MAC


0

005

01

015

Pack

et lo

ss ra






5 Conclusions


Data Availability




References








0

005

01

015

02

025

03

Aver

age e

nd-to

-end

del

ay (s

)

SD-MACFuzzyQoSDACME

StACSWAN



0

50

100

150

200

250

300

350

400

Aver

age t

hrou

ghpu

t (kb

ps)

SD-MACFuzzyQoSDACME

StACSWAN





















































PTypeQLoad ()





Mem

bers

hip

degr

ee

Low

0

05

1High


Medium


Mem

bers

hip

degr

ee

Low

0

05

1High


Medium


Mem

bers

hip

degr

ee

Low

0

05

1High


Medium


Mem

bership

degree

Low

0

05

1High

0 50 70 10030RSSI

Medium


Mem

bers

hip

degr

ee

MediumLow

0

1

Bandwidth available

High

0 10 28 40 50

08

03







For example












PTypeQLoad ()

RSSI BWavail

NodestatusNstatus

QLHP QLMP QLLP

HP (0) 42 (M) 35 (L) 63 (L) 67 (H) 28 (L) 79 NSHmod








α + β + c 1 (4)








aij σ 1113944n


⎛⎝ ⎞⎠ (7)


MSE 1113944n

j1aLj1113872 1113873 (8)


MSE (05 minus 00)2

+(02 minus 00)2

+(03 minus 00)2

+(04 minus 00)2

094(9)













α = 08β = 01 γ = 01

Number of nodes


Packet forward rate





α β c

1 0 00 0 101 0 0902 0 0803 0 0704 0 0605 0 0506 0 0407 0 0308 0 02











Case 7

Case 9

Case 13

Case 21

Case 4Case 3Case 2

1

Number of neighbors

Number of nodes

Mobility


Packet forward rate

1 1 1

Case 1

Case 5Case 6

Case 8

Case 11Case 10

Case 12

Case 19


Case 17Case 18

Case 20

Case 22

Case 36








AC_list Iiend






































HPMPLP

IPlayer

Classifier



flow admission

Routing service


Traffic regulator

Scheduler

MAC








a QLHP


b QLMP


c QLLP


(10)


wHPa wMPb wLPc (11)











HP (high)

LP

ClassifyMP (medium)

LP (low)

60

30

10

Incomingpackets

HP MP








0 10 20 30 40 50 60 70 80 90 1000

10

20

30

40

50

60

70

80

90

100

501

23

45

6

7

8

9

10

11

12

13

14

1516

17

1819

20

21

22

23

24

25

26

2728

29

30

31

32

3334

35

36

3738

39

40

4142

43

44

45

46

47

48

49








0 20 40 60 80 1000

10

20

30

40

50

60

70

80

90

100

50

1

2

3

4

5

6

7

8

9

10

11

12 13

14

1516

17

18

19

20

21

22

23

24

25

26

27

28

2930

3132

33

3435 36

37

38

39

40 41

42

43

44

45

46

4748

49




Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300




Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300









50 100 150 200 250 30065

70

75

80

85

90

95

100

Number of nodes

Deli

very

rate

()

SD-MACFuzzyQoSDACME

StACSWAN

(a)

50 100 150 200 250 300Number of nodes

60

65

70

75

80

85

90

95

100

Pack

et d

eliv

ery

ratio

()

SD-MACFuzzyQoSDACME

StACSWAN

(b)



03

04

05

06

07

08

09

Sess

ion

adm

issio

n ra

tio

SWANStACDACME

FuzzyQoSSD-MAC












SWANStACDACME

FuzzyQoSSD-MAC


03

04

05

06

07

08

09

Sess

ion

com

plet

ion

ratio


SWANStACDACME

FuzzyQoSSD-MAC


0

005

01

015

Pack

et lo

ss ra






5 Conclusions


Data Availability




References








0

005

01

015

02

025

03

Aver

age e

nd-to

-end

del

ay (s

)

SD-MACFuzzyQoSDACME

StACSWAN



0

50

100

150

200

250

300

350

400

Aver

age t

hrou

ghpu

t (kb

ps)

SD-MACFuzzyQoSDACME

StACSWAN






































For example












PTypeQLoad ()

RSSI BWavail

NodestatusNstatus

QLHP QLMP QLLP

HP (0) 42 (M) 35 (L) 63 (L) 67 (H) 28 (L) 79 NSHmod








α + β + c 1 (4)








aij σ 1113944n


⎛⎝ ⎞⎠ (7)


MSE 1113944n

j1aLj1113872 1113873 (8)


MSE (05 minus 00)2

+(02 minus 00)2

+(03 minus 00)2

+(04 minus 00)2

094(9)













α = 08β = 01 γ = 01

Number of nodes


Packet forward rate





α β c

1 0 00 0 101 0 0902 0 0803 0 0704 0 0605 0 0506 0 0407 0 0308 0 02











Case 7

Case 9

Case 13

Case 21

Case 4Case 3Case 2

1

Number of neighbors

Number of nodes

Mobility


Packet forward rate

1 1 1

Case 1

Case 5Case 6

Case 8

Case 11Case 10

Case 12

Case 19


Case 17Case 18

Case 20

Case 22

Case 36








AC_list Iiend






































HPMPLP

IPlayer

Classifier



flow admission

Routing service


Traffic regulator

Scheduler

MAC








a QLHP


b QLMP


c QLLP


(10)


wHPa wMPb wLPc (11)











HP (high)

LP

ClassifyMP (medium)

LP (low)

60

30

10

Incomingpackets

HP MP








0 10 20 30 40 50 60 70 80 90 1000

10

20

30

40

50

60

70

80

90

100

501

23

45

6

7

8

9

10

11

12

13

14

1516

17

1819

20

21

22

23

24

25

26

2728

29

30

31

32

3334

35

36

3738

39

40

4142

43

44

45

46

47

48

49








0 20 40 60 80 1000

10

20

30

40

50

60

70

80

90

100

50

1

2

3

4

5

6

7

8

9

10

11

12 13

14

1516

17

18

19

20

21

22

23

24

25

26

27

28

2930

3132

33

3435 36

37

38

39

40 41

42

43

44

45

46

4748

49




Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300




Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300









50 100 150 200 250 30065

70

75

80

85

90

95

100

Number of nodes

Deli

very

rate

()

SD-MACFuzzyQoSDACME

StACSWAN

(a)

50 100 150 200 250 300Number of nodes

60

65

70

75

80

85

90

95

100

Pack

et d

eliv

ery

ratio

()

SD-MACFuzzyQoSDACME

StACSWAN

(b)



03

04

05

06

07

08

09

Sess

ion

adm

issio

n ra

tio

SWANStACDACME

FuzzyQoSSD-MAC












SWANStACDACME

FuzzyQoSSD-MAC


03

04

05

06

07

08

09

Sess

ion

com

plet

ion

ratio


SWANStACDACME

FuzzyQoSSD-MAC


0

005

01

015

Pack

et lo

ss ra






5 Conclusions


Data Availability




References








0

005

01

015

02

025

03

Aver

age e

nd-to

-end

del

ay (s

)

SD-MACFuzzyQoSDACME

StACSWAN



0

50

100

150

200

250

300

350

400

Aver

age t

hrou

ghpu

t (kb

ps)

SD-MACFuzzyQoSDACME

StACSWAN








































α + β + c 1 (4)








aij σ 1113944n


⎛⎝ ⎞⎠ (7)


MSE 1113944n

j1aLj1113872 1113873 (8)


MSE (05 minus 00)2

+(02 minus 00)2

+(03 minus 00)2

+(04 minus 00)2

094(9)













α = 08β = 01 γ = 01

Number of nodes


Packet forward rate





α β c

1 0 00 0 101 0 0902 0 0803 0 0704 0 0605 0 0506 0 0407 0 0308 0 02











Case 7

Case 9

Case 13

Case 21

Case 4Case 3Case 2

1

Number of neighbors

Number of nodes

Mobility


Packet forward rate

1 1 1

Case 1

Case 5Case 6

Case 8

Case 11Case 10

Case 12

Case 19


Case 17Case 18

Case 20

Case 22

Case 36








AC_list Iiend






































HPMPLP

IPlayer

Classifier



flow admission

Routing service


Traffic regulator

Scheduler

MAC








a QLHP


b QLMP


c QLLP


(10)


wHPa wMPb wLPc (11)











HP (high)

LP

ClassifyMP (medium)

LP (low)

60

30

10

Incomingpackets

HP MP








0 10 20 30 40 50 60 70 80 90 1000

10

20

30

40

50

60

70

80

90

100

501

23

45

6

7

8

9

10

11

12

13

14

1516

17

1819

20

21

22

23

24

25

26

2728

29

30

31

32

3334

35

36

3738

39

40

4142

43

44

45

46

47

48

49








0 20 40 60 80 1000

10

20

30

40

50

60

70

80

90

100

50

1

2

3

4

5

6

7

8

9

10

11

12 13

14

1516

17

18

19

20

21

22

23

24

25

26

27

28

2930

3132

33

3435 36

37

38

39

40 41

42

43

44

45

46

4748

49




Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300




Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300









50 100 150 200 250 30065

70

75

80

85

90

95

100

Number of nodes

Deli

very

rate

()

SD-MACFuzzyQoSDACME

StACSWAN

(a)

50 100 150 200 250 300Number of nodes

60

65

70

75

80

85

90

95

100

Pack

et d

eliv

ery

ratio

()

SD-MACFuzzyQoSDACME

StACSWAN

(b)



03

04

05

06

07

08

09

Sess

ion

adm

issio

n ra

tio

SWANStACDACME

FuzzyQoSSD-MAC












SWANStACDACME

FuzzyQoSSD-MAC


03

04

05

06

07

08

09

Sess

ion

com

plet

ion

ratio


SWANStACDACME

FuzzyQoSSD-MAC


0

005

01

015

Pack

et lo

ss ra






5 Conclusions


Data Availability




References








0

005

01

015

02

025

03

Aver

age e

nd-to

-end

del

ay (s

)

SD-MACFuzzyQoSDACME

StACSWAN



0

50

100

150

200

250

300

350

400

Aver

age t

hrou

ghpu

t (kb

ps)

SD-MACFuzzyQoSDACME

StACSWAN











































α = 08β = 01 γ = 01

Number of nodes


Packet forward rate





α β c

1 0 00 0 101 0 0902 0 0803 0 0704 0 0605 0 0506 0 0407 0 0308 0 02











Case 7

Case 9

Case 13

Case 21

Case 4Case 3Case 2

1

Number of neighbors

Number of nodes

Mobility


Packet forward rate

1 1 1

Case 1

Case 5Case 6

Case 8

Case 11Case 10

Case 12

Case 19


Case 17Case 18

Case 20

Case 22

Case 36








AC_list Iiend






































HPMPLP

IPlayer

Classifier



flow admission

Routing service


Traffic regulator

Scheduler

MAC








a QLHP


b QLMP


c QLLP


(10)


wHPa wMPb wLPc (11)











HP (high)

LP

ClassifyMP (medium)

LP (low)

60

30

10

Incomingpackets

HP MP








0 10 20 30 40 50 60 70 80 90 1000

10

20

30

40

50

60

70

80

90

100

501

23

45

6

7

8

9

10

11

12

13

14

1516

17

1819

20

21

22

23

24

25

26

2728

29

30

31

32

3334

35

36

3738

39

40

4142

43

44

45

46

47

48

49








0 20 40 60 80 1000

10

20

30

40

50

60

70

80

90

100

50

1

2

3

4

5

6

7

8

9

10

11

12 13

14

1516

17

18

19

20

21

22

23

24

25

26

27

28

2930

3132

33

3435 36

37

38

39

40 41

42

43

44

45

46

4748

49




Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300




Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300









50 100 150 200 250 30065

70

75

80

85

90

95

100

Number of nodes

Deli

very

rate

()

SD-MACFuzzyQoSDACME

StACSWAN

(a)

50 100 150 200 250 300Number of nodes

60

65

70

75

80

85

90

95

100

Pack

et d

eliv

ery

ratio

()

SD-MACFuzzyQoSDACME

StACSWAN

(b)



03

04

05

06

07

08

09

Sess

ion

adm

issio

n ra

tio

SWANStACDACME

FuzzyQoSSD-MAC












SWANStACDACME

FuzzyQoSSD-MAC


03

04

05

06

07

08

09

Sess

ion

com

plet

ion

ratio


SWANStACDACME

FuzzyQoSSD-MAC


0

005

01

015

Pack

et lo

ss ra






5 Conclusions


Data Availability




References








0

005

01

015

02

025

03

Aver

age e

nd-to

-end

del

ay (s

)

SD-MACFuzzyQoSDACME

StACSWAN



0

50

100

150

200

250

300

350

400

Aver

age t

hrou

ghpu

t (kb

ps)

SD-MACFuzzyQoSDACME

StACSWAN











































Case 7

Case 9

Case 13

Case 21

Case 4Case 3Case 2

1

Number of neighbors

Number of nodes

Mobility


Packet forward rate

1 1 1

Case 1

Case 5Case 6

Case 8

Case 11Case 10

Case 12

Case 19


Case 17Case 18

Case 20

Case 22

Case 36








AC_list Iiend






































HPMPLP

IPlayer

Classifier



flow admission

Routing service


Traffic regulator

Scheduler

MAC








a QLHP


b QLMP


c QLLP


(10)


wHPa wMPb wLPc (11)











HP (high)

LP

ClassifyMP (medium)

LP (low)

60

30

10

Incomingpackets

HP MP








0 10 20 30 40 50 60 70 80 90 1000

10

20

30

40

50

60

70

80

90

100

501

23

45

6

7

8

9

10

11

12

13

14

1516

17

1819

20

21

22

23

24

25

26

2728

29

30

31

32

3334

35

36

3738

39

40

4142

43

44

45

46

47

48

49








0 20 40 60 80 1000

10

20

30

40

50

60

70

80

90

100

50

1

2

3

4

5

6

7

8

9

10

11

12 13

14

1516

17

18

19

20

21

22

23

24

25

26

27

28

2930

3132

33

3435 36

37

38

39

40 41

42

43

44

45

46

4748

49




Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300




Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300









50 100 150 200 250 30065

70

75

80

85

90

95

100

Number of nodes

Deli

very

rate

()

SD-MACFuzzyQoSDACME

StACSWAN

(a)

50 100 150 200 250 300Number of nodes

60

65

70

75

80

85

90

95

100

Pack

et d

eliv

ery

ratio

()

SD-MACFuzzyQoSDACME

StACSWAN

(b)



03

04

05

06

07

08

09

Sess

ion

adm

issio

n ra

tio

SWANStACDACME

FuzzyQoSSD-MAC












SWANStACDACME

FuzzyQoSSD-MAC


03

04

05

06

07

08

09

Sess

ion

com

plet

ion

ratio


SWANStACDACME

FuzzyQoSSD-MAC


0

005

01

015

Pack

et lo

ss ra






5 Conclusions


Data Availability




References








0

005

01

015

02

025

03

Aver

age e

nd-to

-end

del

ay (s

)

SD-MACFuzzyQoSDACME

StACSWAN



0

50

100

150

200

250

300

350

400

Aver

age t

hrou

ghpu

t (kb

ps)

SD-MACFuzzyQoSDACME

StACSWAN





































AC_list Iiend






































HPMPLP

IPlayer

Classifier



flow admission

Routing service


Traffic regulator

Scheduler

MAC








a QLHP


b QLMP


c QLLP


(10)


wHPa wMPb wLPc (11)











HP (high)

LP

ClassifyMP (medium)

LP (low)

60

30

10

Incomingpackets

HP MP








0 10 20 30 40 50 60 70 80 90 1000

10

20

30

40

50

60

70

80

90

100

501

23

45

6

7

8

9

10

11

12

13

14

1516

17

1819

20

21

22

23

24

25

26

2728

29

30

31

32

3334

35

36

3738

39

40

4142

43

44

45

46

47

48

49








0 20 40 60 80 1000

10

20

30

40

50

60

70

80

90

100

50

1

2

3

4

5

6

7

8

9

10

11

12 13

14

1516

17

18

19

20

21

22

23

24

25

26

27

28

2930

3132

33

3435 36

37

38

39

40 41

42

43

44

45

46

4748

49




Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300




Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300









50 100 150 200 250 30065

70

75

80

85

90

95

100

Number of nodes

Deli

very

rate

()

SD-MACFuzzyQoSDACME

StACSWAN

(a)

50 100 150 200 250 300Number of nodes

60

65

70

75

80

85

90

95

100

Pack

et d

eliv

ery

ratio

()

SD-MACFuzzyQoSDACME

StACSWAN

(b)



03

04

05

06

07

08

09

Sess

ion

adm

issio

n ra

tio

SWANStACDACME

FuzzyQoSSD-MAC












SWANStACDACME

FuzzyQoSSD-MAC


03

04

05

06

07

08

09

Sess

ion

com

plet

ion

ratio


SWANStACDACME

FuzzyQoSSD-MAC


0

005

01

015

Pack

et lo

ss ra






5 Conclusions


Data Availability




References








0

005

01

015

02

025

03

Aver

age e

nd-to

-end

del

ay (s

)

SD-MACFuzzyQoSDACME

StACSWAN



0

50

100

150

200

250

300

350

400

Aver

age t

hrou

ghpu

t (kb

ps)

SD-MACFuzzyQoSDACME

StACSWAN


























































HPMPLP

IPlayer

Classifier



flow admission

Routing service


Traffic regulator

Scheduler

MAC








a QLHP


b QLMP


c QLLP


(10)


wHPa wMPb wLPc (11)











HP (high)

LP

ClassifyMP (medium)

LP (low)

60

30

10

Incomingpackets

HP MP








0 10 20 30 40 50 60 70 80 90 1000

10

20

30

40

50

60

70

80

90

100

501

23

45

6

7

8

9

10

11

12

13

14

1516

17

1819

20

21

22

23

24

25

26

2728

29

30

31

32

3334

35

36

3738

39

40

4142

43

44

45

46

47

48

49








0 20 40 60 80 1000

10

20

30

40

50

60

70

80

90

100

50

1

2

3

4

5

6

7

8

9

10

11

12 13

14

1516

17

18

19

20

21

22

23

24

25

26

27

28

2930

3132

33

3435 36

37

38

39

40 41

42

43

44

45

46

4748

49




Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300




Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300









50 100 150 200 250 30065

70

75

80

85

90

95

100

Number of nodes

Deli

very

rate

()

SD-MACFuzzyQoSDACME

StACSWAN

(a)

50 100 150 200 250 300Number of nodes

60

65

70

75

80

85

90

95

100

Pack

et d

eliv

ery

ratio

()

SD-MACFuzzyQoSDACME

StACSWAN

(b)



03

04

05

06

07

08

09

Sess

ion

adm

issio

n ra

tio

SWANStACDACME

FuzzyQoSSD-MAC












SWANStACDACME

FuzzyQoSSD-MAC


03

04

05

06

07

08

09

Sess

ion

com

plet

ion

ratio


SWANStACDACME

FuzzyQoSSD-MAC


0

005

01

015

Pack

et lo

ss ra






5 Conclusions


Data Availability




References








0

005

01

015

02

025

03

Aver

age e

nd-to

-end

del

ay (s

)

SD-MACFuzzyQoSDACME

StACSWAN



0

50

100

150

200

250

300

350

400

Aver

age t

hrou

ghpu

t (kb

ps)

SD-MACFuzzyQoSDACME

StACSWAN













































HPMPLP

IPlayer

Classifier



flow admission

Routing service


Traffic regulator

Scheduler

MAC








a QLHP


b QLMP


c QLLP


(10)


wHPa wMPb wLPc (11)











HP (high)

LP

ClassifyMP (medium)

LP (low)

60

30

10

Incomingpackets

HP MP








0 10 20 30 40 50 60 70 80 90 1000

10

20

30

40

50

60

70

80

90

100

501

23

45

6

7

8

9

10

11

12

13

14

1516

17

1819

20

21

22

23

24

25

26

2728

29

30

31

32

3334

35

36

3738

39

40

4142

43

44

45

46

47

48

49








0 20 40 60 80 1000

10

20

30

40

50

60

70

80

90

100

50

1

2

3

4

5

6

7

8

9

10

11

12 13

14

1516

17

18

19

20

21

22

23

24

25

26

27

28

2930

3132

33

3435 36

37

38

39

40 41

42

43

44

45

46

4748

49




Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300




Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300









50 100 150 200 250 30065

70

75

80

85

90

95

100

Number of nodes

Deli

very

rate

()

SD-MACFuzzyQoSDACME

StACSWAN

(a)

50 100 150 200 250 300Number of nodes

60

65

70

75

80

85

90

95

100

Pack

et d

eliv

ery

ratio

()

SD-MACFuzzyQoSDACME

StACSWAN

(b)



03

04

05

06

07

08

09

Sess

ion

adm

issio

n ra

tio

SWANStACDACME

FuzzyQoSSD-MAC












SWANStACDACME

FuzzyQoSSD-MAC


03

04

05

06

07

08

09

Sess

ion

com

plet

ion

ratio


SWANStACDACME

FuzzyQoSSD-MAC


0

005

01

015

Pack

et lo

ss ra






5 Conclusions


Data Availability




References








0

005

01

015

02

025

03

Aver

age e

nd-to

-end

del

ay (s

)

SD-MACFuzzyQoSDACME

StACSWAN



0

50

100

150

200

250

300

350

400

Aver

age t

hrou

ghpu

t (kb

ps)

SD-MACFuzzyQoSDACME

StACSWAN





































a QLHP


b QLMP


c QLLP


(10)


wHPa wMPb wLPc (11)











HP (high)

LP

ClassifyMP (medium)

LP (low)

60

30

10

Incomingpackets

HP MP








0 10 20 30 40 50 60 70 80 90 1000

10

20

30

40

50

60

70

80

90

100

501

23

45

6

7

8

9

10

11

12

13

14

1516

17

1819

20

21

22

23

24

25

26

2728

29

30

31

32

3334

35

36

3738

39

40

4142

43

44

45

46

47

48

49








0 20 40 60 80 1000

10

20

30

40

50

60

70

80

90

100

50

1

2

3

4

5

6

7

8

9

10

11

12 13

14

1516

17

18

19

20

21

22

23

24

25

26

27

28

2930

3132

33

3435 36

37

38

39

40 41

42

43

44

45

46

4748

49




Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300




Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300









50 100 150 200 250 30065

70

75

80

85

90

95

100

Number of nodes

Deli

very

rate

()

SD-MACFuzzyQoSDACME

StACSWAN

(a)

50 100 150 200 250 300Number of nodes

60

65

70

75

80

85

90

95

100

Pack

et d

eliv

ery

ratio

()

SD-MACFuzzyQoSDACME

StACSWAN

(b)



03

04

05

06

07

08

09

Sess

ion

adm

issio

n ra

tio

SWANStACDACME

FuzzyQoSSD-MAC












SWANStACDACME

FuzzyQoSSD-MAC


03

04

05

06

07

08

09

Sess

ion

com

plet

ion

ratio


SWANStACDACME

FuzzyQoSSD-MAC


0

005

01

015

Pack

et lo

ss ra






5 Conclusions


Data Availability




References








0

005

01

015

02

025

03

Aver

age e

nd-to

-end

del

ay (s

)

SD-MACFuzzyQoSDACME

StACSWAN



0

50

100

150

200

250

300

350

400

Aver

age t

hrou

ghpu

t (kb

ps)

SD-MACFuzzyQoSDACME

StACSWAN







































0 10 20 30 40 50 60 70 80 90 1000

10

20

30

40

50

60

70

80

90

100

501

23

45

6

7

8

9

10

11

12

13

14

1516

17

1819

20

21

22

23

24

25

26

2728

29

30

31

32

3334

35

36

3738

39

40

4142

43

44

45

46

47

48

49








0 20 40 60 80 1000

10

20

30

40

50

60

70

80

90

100

50

1

2

3

4

5

6

7

8

9

10

11

12 13

14

1516

17

18

19

20

21

22

23

24

25

26

27

28

2930

3132

33

3435 36

37

38

39

40 41

42

43

44

45

46

4748

49




Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300




Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300









50 100 150 200 250 30065

70

75

80

85

90

95

100

Number of nodes

Deli

very

rate

()

SD-MACFuzzyQoSDACME

StACSWAN

(a)

50 100 150 200 250 300Number of nodes

60

65

70

75

80

85

90

95

100

Pack

et d

eliv

ery

ratio

()

SD-MACFuzzyQoSDACME

StACSWAN

(b)



03

04

05

06

07

08

09

Sess

ion

adm

issio

n ra

tio

SWANStACDACME

FuzzyQoSSD-MAC












SWANStACDACME

FuzzyQoSSD-MAC


03

04

05

06

07

08

09

Sess

ion

com

plet

ion

ratio


SWANStACDACME

FuzzyQoSSD-MAC


0

005

01

015

Pack

et lo

ss ra






5 Conclusions


Data Availability




References








0

005

01

015

02

025

03

Aver

age e

nd-to

-end

del

ay (s

)

SD-MACFuzzyQoSDACME

StACSWAN



0

50

100

150

200

250

300

350

400

Aver

age t

hrou

ghpu

t (kb

ps)

SD-MACFuzzyQoSDACME

StACSWAN





































0 20 40 60 80 1000

10

20

30

40

50

60

70

80

90

100

50

1

2

3

4

5

6

7

8

9

10

11

12 13

14

1516

17

18

19

20

21

22

23

24

25

26

27

28

2930

3132

33

3435 36

37

38

39

40 41

42

43

44

45

46

4748

49




Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300




Nodes75

Nodes100

Nodes125

Nodes150

Nodes175

Nodes200

Nodes225

Nodes250

Nodes275

Nodes300









50 100 150 200 250 30065

70

75

80

85

90

95

100

Number of nodes

Deli

very

rate

()

SD-MACFuzzyQoSDACME

StACSWAN

(a)

50 100 150 200 250 300Number of nodes

60

65

70

75

80

85

90

95

100

Pack

et d

eliv

ery

ratio

()

SD-MACFuzzyQoSDACME

StACSWAN

(b)



03

04

05

06

07

08

09

Sess

ion

adm

issio

n ra

tio

SWANStACDACME

FuzzyQoSSD-MAC












SWANStACDACME

FuzzyQoSSD-MAC


03

04

05

06

07

08

09

Sess

ion

com

plet

ion

ratio


SWANStACDACME

FuzzyQoSSD-MAC


0

005

01

015

Pack

et lo

ss ra






5 Conclusions


Data Availability




References








0

005

01

015

02

025

03

Aver

age e

nd-to

-end

del

ay (s

)

SD-MACFuzzyQoSDACME

StACSWAN



0

50

100

150

200

250

300

350

400

Aver

age t

hrou

ghpu

t (kb

ps)

SD-MACFuzzyQoSDACME

StACSWAN








































50 100 150 200 250 30065

70

75

80

85

90

95

100

Number of nodes

Deli

very

rate

()

SD-MACFuzzyQoSDACME

StACSWAN

(a)

50 100 150 200 250 300Number of nodes

60

65

70

75

80

85

90

95

100

Pack

et d

eliv

ery

ratio

()

SD-MACFuzzyQoSDACME

StACSWAN

(b)



03

04

05

06

07

08

09

Sess

ion

adm

issio

n ra

tio

SWANStACDACME

FuzzyQoSSD-MAC












SWANStACDACME

FuzzyQoSSD-MAC


03

04

05

06

07

08

09

Sess

ion

com

plet

ion

ratio


SWANStACDACME

FuzzyQoSSD-MAC


0

005

01

015

Pack

et lo

ss ra






5 Conclusions


Data Availability




References








0

005

01

015

02

025

03

Aver

age e

nd-to

-end

del

ay (s

)

SD-MACFuzzyQoSDACME

StACSWAN



0

50

100

150

200

250

300

350

400

Aver

age t

hrou

ghpu

t (kb

ps)

SD-MACFuzzyQoSDACME

StACSWAN











































SWANStACDACME

FuzzyQoSSD-MAC


03

04

05

06

07

08

09

Sess

ion

com

plet

ion

ratio


SWANStACDACME

FuzzyQoSSD-MAC


0

005

01

015

Pack

et lo

ss ra






5 Conclusions


Data Availability




References








0

005

01

015

02

025

03

Aver

age e

nd-to

-end

del

ay (s

)

SD-MACFuzzyQoSDACME

StACSWAN



0

50

100

150

200

250

300

350

400

Aver

age t

hrou

ghpu

t (kb

ps)

SD-MACFuzzyQoSDACME

StACSWAN





































5 Conclusions


Data Availability




References








0

005

01

015

02

025

03

Aver

age e

nd-to

-end

del

ay (s

)

SD-MACFuzzyQoSDACME

StACSWAN



0

50

100

150

200

250

300

350

400

Aver

age t

hrou

ghpu

t (kb

ps)

SD-MACFuzzyQoSDACME

StACSWAN

































































Documents

Software-DefinedMultilayeredAdmissionControlforQualityof …downloads.hindawi.com/journals/wcmc/2020/2989751.pdf · 2020-02-23 · 1.Introduction MobileAd-hocNetworks(MANET)formsaself-organized,