networking models in flying ad-hoc networks (fanets): concepts and...

J Intell Robot SystDOI 10.1007/s10846-013-9959-7

Networking Models in Flying Ad-Hoc Networks(FANETs): Concepts and Challenges

Ozgur Koray Sahingoz

Received: 7 September 2013 / Accepted: 14 September 2013© Springer Science+Business Media Dordrecht 2013

Abstract In recent years, the capabilities androles of Unmanned Aerial Vehicles (UAVs) haverapidly evolved, and their usage in military andcivilian areas is extremely popular as a result ofthe advances in technology of robotic systemssuch as processors, sensors, communications, andnetworking technologies. While this technology isprogressing, development and maintenance costsof UAVs are decreasing relatively. The focus ischanging from use of one large UAV to use ofmultiple UAVs, which are integrated into teamsthat can coordinate to achieve high-level goals.This level of coordination requires new network-ing models that can be set up on highly mobilenodes such as UAVs in the fleet. Such network-ing models allow any two nodes to communicatedirectly if they are in the communication range,or indirectly through a number of relay nodessuch as UAVs. Setting up an ad-hoc networkbetween flying UAVs is a challenging issue, andrequirements can differ from traditional networks,Mobile Ad-hoc Networks (MANETs) and Ve-hicular Ad-hoc Networks (VANETs) in termsof node mobility, connectivity, message routing,service quality, application areas, etc. This paper

O. K. Sahingoz (B)Computer Engineering Department,Turkish Air Force Academy, Yesilyurt,Istanbul, 34149, Turkeye-mail: [email protected]

identifies the challenges with using UAVs as relaynodes in an ad-hoc manner, introduces networkmodels of UAVs, and depicts open research issueswith analyzing opportunities and future work.

Keywords Networking models · Multi-UAVs ·FANET · VANET · Mobile networking ·UAV networks

1 Introduction

In the last two decades, as a result of the rapidtechnological advancement in computation, sen-sor, communication and networking technologies,Unmanned Aerial Vehicles (UAVs) promise newapplication areas for military and civilian areassuch as, relay for ad-hoc networks [1–3], searchand rescue operations [4], electronic attacks inhostile areas [5], ground target detection andtracking [6, 7], automatic forest fire monitoringand measurement [8], wind estimation [9], disastermonitoring [10], remote sensing [11], airspeed cal-ibration [12], agricultural remote sensing [13], etc.

It is a relatively easy task to use UAVsin an Unmanned Aerial System (UAS) for in-creasing communication range and data aggrega-tion capability of nodes in a system. In case ofinfrastructure-less places, such as enemy territo-ries or natural disaster areas, there is an immedi-ate need to build a network between teams, and a

Downloaded from http://iranpaper.ir www.NoavaranGermi.ir

J Intell Robot Syst

fast-deployed UAV can be an acceptable solutionas a relay node.

In single-UAV applications, each ground nodecan easily communicate with the UAV, and thusestablish a star topology network structure wherethe UAV is at the center of the star. By usingthis topology, a ground node can indirectly com-municate with others over the UAV. However,single UAV systems have some challenging issuesin peer-to-peer communication such as increasingtransmission range, sending more data, and mini-mizing any interference. A solution to these prob-lems is the use of high gain directional antennasinstead of omnidirectional antennas. Undoubt-edly, this leads to a limited improvement in theperformance of UAS, and this is not satisfactory.

To provide a longer presence over the theatreof operation in order to carry powerful processors,sensors and communication hardware, a largeUAV is preferred in single vehicle UASs, in whichsetting up a communication environment is easy.However, they are not only heavy and large butalso pose a significant danger to human life andproperty in case of a failure. Moreover, they areexpensive, and failure of a UAV can have a highcost.

On the other hand, technological advancementof electronics and sensor technologies have de-creased the production cost of UAVs, and low-cost mini-UAVs are becoming more and morepopular in both academic areas and practical ap-plications. They are small in size and weight;therefore, they pose little or no threat to humanlife and buildings/properties. In addition, they canbe reused for various types of applications. It is

difficult to detect and track UAVs while they areflying, due to their size. They can fly at much loweraltitudes, can be easily transported to differentareas, and can be launched by an individual inany kind of terrain without a runway or a specificlaunching device. However, mini-UAVs have re-stricted capabilities such as power, sensing, com-munication, and computation; however, the useof a team with mini-UAVs, as multiple UAVs,improves the capability and capacity of UASsand provides a flexible platform for a variety ofapplications. Usage of multi-UAV systems hassignificant advantages over single UAV systems:

• Especially in search missions, the usage of anumber of UAVs can parallelize individualtasks thus decreasing the completion time ofa mission.

• In a single UAV system, if the UAV or asensor/hardware fails, the UAV should returnto the base. However, in multi-UAV systems,other UAVs can share tasks among them-selves and this increases the fault tolerance ofthe system.

• In a heterogeneous UAV team, it is possibleto use the capabilities of other UAVs.

On the other hand, these advantages and improve-ments have a challenging issue: ef f icient commu-nication and coordination of UAVs in the team.Therefore, a multi-UAV system needs some re-quired hardware and a well-defined networkingmodel whose communication/network layer is de-picted in Fig. 1.

To achieve effective management of multipleUAV systems, and to perform complex tasks,

Fig. 1 A multi-UAVsystem’s processingunits and communication/network layer

Network Layer

Database GUIMission ControlGround Control

Sensor UnitGPS Image

Radar IR

Task Manager

UAV Controller

Laser ***

Communication Interface

UAV-1

* * * *

Sensor UnitGPS Image

Radar IR

Task Manager

UAV Controller

Laser ***


UAV-2 Sensor UnitGPS Image

Radar IR

Task Manager

UAV Controller

Laser ***


UAV-N


J Intell Robot Syst

UAVs need to cooperate with others in theteam. To enable this cooperation, UAVs are con-strained to stay within the communication rangeof one another. UAVs must achieve a high degreeof coordination and a robust inter-vehicle commu-nication network in an ad-hoc manner. Tradition-ally, the communication technologies on fixed net-works, Mobile Ad-hoc NETworks (MANETs),and slowly moving Vehicular Ad-hoc NETworks(VANETs) do not address the unique character-istics of these networks, which use highly mo-bile nodes. Therefore, there is a requirement ofdefining a new ad-hoc networking model, whichdiffers from other kinds of ad-hoc networks alongthe connectivity, quality of services, sensor types,node movement features, data delivery, servicediscovery, etc.

In the literature, some academic researchershave defined the explained UAV networkingmodel with different names, as depicted in Table 1.

Although, the used names have littledifferences in their definitions, it is a clearlyseen fact that the new communication model is asubclass of VANET. However, VANET routingprotocols are not feasible or does not providesufficient throughput for networks with highlymobile nodes. Therefore, it is more appropriateto name this new networking model as a FlyingAd-Hoc Network (FANET) [14], a subclass of aVANET.

While considering the advancements in ad-hoc networking, it can be explicitly foreseen thatUAVs will play an attractive function in setting up

Table 1 Names of highly mobile networking model

Given name Reference

Airborne network [15, 16](Airborne Telemetry Network,Airborne Communication Network,Airborne Backbone Network, etc.)

Networked Aerial Robots [17]Unmanned Aeronautical Ad-Hoc [18, 19]

Network (UAANET)UAV Ad-Hoc Network [20]Aerial Communication Network [21]Networks of UAVs [22, 23]Distributed Aerial Sensor Network [24]Flying Ad-Hoc Network (FANET) [14]

a highly mobile networks for enlarging the theatreof the operation. Therefore, the intent of thispaper1 is to introduce the challenges and network-ing models in FANET, and to present some openissues and future opportunities for researchers onthis topic.

In the following, background informationabout UAS and MANET is given in Section 2.After that, the ad-hoc networking concepts withflying objects are introduced, and different net-working models for FANET Network Layer aregiven in Sections 3 and 4, respectively. Subse-quently, open issues and challenges are depictedin Section 5. Finally, the paper is concluded inSection 6.

2 Background

2.1 Unmanned Aerial System (UAS)

While UAV technology and markets develop;its cost, weight and size decrease and perfor-mance of sensors/processors increases. These re-sults lead to a considerable interest and researchon UAVs, and hence, UASs have shown excep-tional promises for lots of application areas. AUAS consists of five main components (as de-picted in Fig. 2):

• A UAV is a pilotless aircraft that does notrequire any direct human intervention forflying. It can navigate autonomously accord-ing to its pre-programmed software, or can becontrolled remotely. Apart from basic planecomponents, it also contains some comput-ing devices and sensors for determining itsposition and for gathering information fromthe mission area. A UAV can be categorizedas a “Fixed Wing Vehicle”, which needs alaunch system for take-off, or “Rotor WingVehicle” where vertical take-offs and landings(VTOL) are possible. UAVs are also clas-sified as “mini”, “small”, or “large”, accordingto their weights.

1A preliminary version [25] of this paper was presented atICUAS 2013, Atlanta.


J Intell Robot Syst

Fig. 2 Unmanned aerialsystem with components

UAV withExternal payload

• Payloads are essential systems in UAVs, andthey are mainly the equipment added to theUAV for carrying out various operational mis-sions. These include sensors, emitters, lethal/nonlethal weapons, stores, etc. Mainly, pay-loads are carried in an internal payload bayof the UAV; in some cases, they can also beattached to the UAV, but this can result in amild deterioration of the aerodynamic proper-ties of the UAV.

• Command and Control Center contains aGround Control Station (GCS) and providestechnological facilities for human interactionsto the UAVs in the air.

• A Launch System is generally a catapult sys-tem, which gives a UAV its flight speed inan extremely short time and distance. At thesame time, most of the mini-UAVs can belaunched by hands while large UAVs need anairfield for take-off and landing.

• Control and Data Transmission Linksare used to describe how information issent/received both to/from the GCS andUAVs. If UAVs are in the line-of-sight(LOS), direct radio links can be used;however, if they are beyond line-of-sight(BLOS), then there is a need of higher levelmediator such as satellites or another UAV.

Undoubtedly, UAV is the most crucial part ofthe above-mentioned components. Therefore, theperformance, capacity, and operation theatre ofthe UAS directly related to its abilities, such asendurance, range, altitude, payload, etc. To useUASs in different application areas, there is a

requirement of decreasing the personnel needsand increasing the autonomy of UAVs by whichthey can fly freely in the sky and carry out themission without any centralized controller fromthe ground station.

2.2 Mobile Ad-Hoc Networking

A Mobile Ad-Hoc Network (MANET) is a dy-namically self-organizing and infrastructure-lessnetwork of mobile devices such as laptop, palm-top, cellphones, walkie-talkies, etc. These devicescan communicate over bandwidth-constrainedwireless links such as IEEE 802.11 a/b/g/n, 802.16,etc. Each device in MANET can move randomlywith different speeds; therefore, its links withother devices change frequently. For transferringdata between two remote nodes, each mobile de-vice must forward some data, which are unrelatedto its own use; therefore, a MANET runs not onlyas a host but also as a router.

MANET nodes are traditionally small and havelimited processor and energy capacities. It isdifficult to build and maintain such a network in alonger time span. Therefore, some main function-ality of network layers, such as routing, should bedone by these mobile nodes dynamically.

With the rapid enhancement in technologiesand sensors, there has been an increase in uti-lization rates, and application areas of MANET.As a result, mobile nodes have begun to embedin moving vehicles, such as cars, ambulances, fireengines, tanks, etc. This networking concept iscalled a Vehicular Ad-Hoc Network (VANET),which is a part of an Intelligent Transport System


J Intell Robot Syst

of smart vehicles with additional properties suchas collision/distance warning, driver assistance,cooperative driving/cruise control, propagation ofroad information, etc.

3 Ad-Hoc Networking with Flying Nodes

With the increasing usage of multi-UAVs in an ad-hoc manner, a need for a new networking layerhas emerged. The mobile nodes of the systemsare highly mobile vehicles (UAVs); therefore, thislayer is mainly a subclass of VANET, and this islocated in a sub layer of an Aerial Network Layer,as depicted in Fig. 3 of a network-centric archi-tecture [25]. A network layer is principally re-sponsible for the routing process, which selects anappropriate path between the sender nodes andreceiver nodes over different routers/nodes. Themain task of a FANET Network Layer is to act asan intermediary between these layers while per-forming the traditional routing on its own level.

Using multi-UAVs in an ad-hoc network man-ner brings some advantages [14]:

• Decreases the mission completion time: Themissions, such as reconnaissance, surveillance,search and rescue, can be carried out fasterwith proportional to the number of UAVs.

• Decreases total/maintenance cost: Instead ofusing a large and expensive UAV, the usage ofmulti-mini-UAVs costs are lowered in termsof acquisition and maintenance.

• Increases scalability: It increases the area ofoperation theatre by easily adding new UAVs.UAS dynamically reorganizes nodes’ routingtables by taking into account newly addedUAVs.

• Increases survivability: Multi-UAV systemsare more tolerant to faults of hardware/sensors. In the case of some sensors failing ora loss of control of a UAV, the mission cancontinue with any remaining UAVs.

• Decreases detectability (Low radar cross-section): A radar cross-section is extremelycrucial for military applications. Mini-UAVshave low radar cross-sections, low infraredsignatures, and low acoustic signatures due totheir sizes and composite structures. There-fore, they may not be easily detectable byradars (especially compared to airplanes andlarge UAVs).

Although multi-UAV systems have some sig-nificant advantages; due to their dynamic net-work topology, communication of two distantnodes over other UAVs is still a challenging is-sue. Fueled by this requirement, in the literature,

Fig. 3 A network-centricarchitecture in abattlefield


J Intell Robot Syst

some routing protocols have been proposed forFANETs. These protocols mainly depend on com-munication types. In a FANET, there are mainlytwo different communication types: UAV-to-UAVand UAV-to-Infrastructure communications.

In UAV-to-UAV communication, each UAVcan communicate with others in order to meet theneeds of different application areas, such as co-operative path planning and target tracking. TwoUAVs can either directly communicate with eachother, or a multi hop communication path can beconstructed over other UAVs. UAVs can haveshort range, and long range of communicationbetween them, selection of the range also dependson the needed data transfer rate.

In UAV-to-Infrastructure communication, UAVscommunicate with a fixed infrastructure, such asground stations, satellites, or warships near theoperation theatre to provide information servicesfor other users in the global networks.

Communication between UAVs and UAV-to-Infrastructure is also a challenging issue. To in-crease the data transfer rate and the performanceof the system, different type of antennas and sen-sors can be used. With the usage of GPS receiversand directed antennas, like phased array anten-nas, communication links can be effectively set inFANET.

Namudiri et al. divides design principles of anetwork with flying nodes in four different di-mensions, as depicted in Fig. 4 [26]. They claimedthat the synergy between these components sig-nificantly improves the capabilities of a system.With the same goal, U.S. Unmanned SystemsRoadmaps [27, 28] highlight some key enabling

Information Sharing (Secret Key Sharing,

Integrity protection and verification)

Control System (Mobility Models and Trajectory Planning)

Networks and Communications

(Connectivity, Coverage and Throughput)

Situational Awareness (Sense and Avoid, Conflict

Detection)

Fig. 4 A network-centric architecture in a battlefield

networking technologies, which are planned to bedeveloped as follows:

• Mobile ad-hoc networking protocols that en-able high reliability command and control ofautonomous UAVs.

• Better data compression, encryption andprocessing algorithms.

• Embedded network security with single-chipall-encapsulated encryption modules.

• High-gain, rugged, and lower cost multidirec-tional antennas.

• Highly efficient radios for low bandwidth andfull-time high-speed communications links.

• Support of various path diversity techniques,integrated networking, and data diversity.

While these UAV technologies are evolving at afast pace, the size of UAVs has been decreased,and the number of UAVs in UASs has been in-creasing. Hence, installing and maintenance of thenetworked communication between these UAVsand ground stations are emerged as crucial issuesto solve. Especially in some application fields,which use heterogeneous UAVs with differentpayload capacity, sensors, avionics, communica-tion range, and flight endurance; communicationand coordination of UAVs are challenging tasksin order to achieve such a networked commu-nication. In the following section, different rout-ing protocols in Network Layer of FANET aredescribed.

4 FANET Networking Models

In the literature [29, 30], many routing protocolsexists in wireless and ad-hoc networks such as pre-computed routing, dynamic source routing, on-demand routing, cluster based routing, flooding,etc. Due to a shortage of energy, to increase theFANET operation time, there are some needs todecrease transmitting power by sending a messageto closer nodes (UAVs) and by using multi-hoprouting between the sender and receiver nodesover highly mobile UAVs as relay nodes. FANETis a subclass of VANET and MANET; there-fore, firstly typical MANET routing protocols arepreferred and tested for FANET. Due to theUAV-specific issues, such as quick changes in link


J Intell Robot Syst

quality, most of these protocols are not directlyapplicable for FANET. Therefore, to adopt thisnew networking model, both some specific ad-hocnetworking protocols have been implemented andsome previous ones have been modified in theliterature. These protocols can be categorized infour main classes;

• Static protocols have static routing tablesthere is no need to refresh these tables.

• Proactive protocols, also known as tabledriven protocols, are periodically refreshedrouting tables.

• Reactive protocols, also called on-demand pro-tocols, discover paths for messages on demand.

• Hybrid protocols use both proactive and reac-tive protocols.

By using these routing protocols, a FANET candynamically discover new routes between commu-nicating nodes, and this network may allow addi-tion and subtraction of UAV nodes dynamically.

4.1 Static Routing Protocols

In static routing protocol, a routing table iscomputed and loaded to UAV nodes before amission, and cannot be updated during the oper-ation; therefore, it is static. In this type network-ing model, UAVs typically have a constant/fixedtopology [31]. Each node can communicate witha few numbers of UAVs or ground stations, andit only stores their information. In case of a fail-ure (of a UAV or ground station), for updatingthe tables, it is necessary to wait the end of themission. Therefore, they are not fault tolerant andappropriate for dynamic environments.

Load Carry and Deliver Routing (LCAD) [32,33] is one of the first routing models in FANET.In this model, a UAV loads data from a groundnode (or gets video image of its path); afterthat, it carries these valuable data to the desti-nation by flying; and finally it delivers the datato a destination ground node (such as a militaryteam or a ground control station), as depictedin Fig. 5. Although, in the initial definition ofLCAD, a single-source and a single-destinationnetworking scenario was considered, in practice,multiple-source multiple-destination networkingscenarios can also be implemented easily. Dueto the distant location of communicating nodes(the use of ground nodes and no other UAVsused for multi-hop communication), LCAD is freeof interference in the same networking system,and this increases the throughput of the system.This routing methodology is a feasible solutionespecially for bulk data transfer (such as latency-insensitive video images) and delay-tolerant ap-plications (such as secure transfer) with mini-mum hops [34]. LCAD routing aims to maxi-mize the throughput while increasing the secu-rity. However, as the distance of communicatingparts increases, the transmission delay becomesextremely large and intolerable. In this case, trans-mission time depends mainly on how fast a UAVcan fly this distance. At the same time, a UAV canget data from the source nodes/places, only whenit is over them. If there is only one UAV, it is notpossible to capture each event on a specific areaor load data from the source node when it is pro-duced. To decrease the transmission time; morethan one UAV can be used on the same path,speed of UAVs can be increased, and a LCADnetwork can be divided into smaller LCADsub-networks.

Fig. 5 Load carry anddeliver routing model


J Intell Robot Syst

Multi-Level Hierarchical Routing : Due to thestructure of the typical VANET environment, therouting protocols are especially organized as flatrouting in two dimensional space. However, inUAV-based network applications, the flying en-vironment is generally modeled in three dimen-sional space, and nodes have different attributessuch as size, flight height, energy usage capabil-ities, types of sensors, etc. Large-scale VANETapplications with hundreds/thousands of mobilenodes and typical FANET applications with clas-sical flat routing result in a certain performancedegradation. To solve this problem and to in-crease the network scalability, one appropriate so-lution is the use of hierarchical routing protocols[35, 36], which divide FANET into clusters ofUAVs.

Hierarchically organized UAV networks con-sist of a number of clusters to operate in differentmission areas, as shown in Fig. 6. Each cluster hasa cluster head (CH), which represents the wholecluster, and it is possible to assign different func-tionalities to each cluster. Each CH is in connec-tion with the upper/lower layers (ground stations,UAVs, satellites, etc.) directly or indirectly. Todisseminate data (by broadcasting) and controlinformation to other UAVs in the cluster, CHshould be in direct transmission range of otherUAVs in cluster. This model is better if UAVsare organized in different swarms, the missionarea is large, and many UAVs are used in thenetwork.

Data Centric Routing : Due to the nature of thewireless communication structure of UAVs, one-to-many data transmission can be preferred toone-to-one data transmission [37, 38]. This rout-ing is preferred when the data is requested by anumber of nodes, and it is distributed accordingto on-demand algorithms. Data-centric routingis a promising paradigm of routing mechanismand can be adapted for FANET [39, 40]. In thismodel, the data request and collection is doneaccording to data attributes rather than senderor receiver nodes’ IDs as depicted in Fig. 7. Thismodel is generally practiced with cluster struc-tures and publish-subscribe model, which depictsits efficiency on distributed computing models[41, 42].

In this model, the consumer node (can be aground node or a UAV) disseminates queries(such as “get photo images of area A if thereis a change of more than % 5”) as subscriptionmessage in order to collect specific data from aspecific area. The producer node decides whichinformation to publish and starts data dissemina-tion. When published data reach a UAV (as arelay node), it checks the subscription messageson it and forwards these data accordingly. Routingis done with respect to the content of data; and ifneeded, data aggregation algorithms can be usedfor energy-efficient data dissemination. Although,query dissemination and data collection processesadd some extra burden to the network load; dueto the elimination of redundant transmissions,

Fig. 6 Hierarchicalrouting model


J Intell Robot Syst

Fig. 7 Data centricrouting model

usage of this model results total efficiency in-crease. This type routing performs three dimen-sions of decoupling;

• Space decoupling means that communicatingparties do not need to know each other’s IDor location.

• Time decoupling means that communicatingparties do not need to be on-line at the sametime to communicate.

• Flow decoupling results in an asynchronouscommunication structure; therefore, messagesending process is not blocked by outsideinteractions.

A data centric routing model, especially pre-ferred for the cluster structured UAS, drives theUAV mission along a predetermined flight-planand does make high level cooperation betweenclusters.

4.2 Proactive Routing Protocols

Proactive routing protocols (PRP) use tables tostore all the routing information of each other’snode or nodes of a specific region in the net-work. Various table-driven protocols can be usedin FANET, and they differ in the way of updatemechanism of the routing table when the topologychanges. The main advantage of proactive routingis that it contains the latest information of theroutes; therefore, it is easy to select a path fromthe sender to the receiver, and there is no need to

wait. However, there are some explicit disadvan-tages. Firstly, due to the need of a lot of messageexchanges between nodes, PRPs cannot efficientlyuse bandwidth, which is a limited communicationresource of FANET; therefore, PRPs are not suit-able for highly mobile and/or larger networks.Secondly, it shows a slow reaction, when the topol-ogy is changed, or a failure is occurred. Two mainprotocols are widely used in VANETs: OptimizedLink State Routing (OLSR) and Destination-Sequenced Distance Vector (DSDV) protocols.

• Optimized Link State Routing (OLSR) is aproactive link-state routing protocol, whichuses two types of messages (hello and topol-ogy control messages) to discover neighbors[43]. Hello messages are used for detectingneighbor nodes in the direct communicationrange. This message contains the list of knownneighbors, and it is periodically broadcast toone-hop neighbors. On the other hand, topol-ogy control messages are used for maintain-ing topological information of the system.These messages are used periodically to re-fresh topology information; therefore, eachnode can re-calculate the routes to all nodesin the system. This periodic flooding natureof the protocol results a large amount ofoverhead. Therefore, to reduce this overheadMulti Point Relay (MPR) mechanism is used.In this mechanism, each node selects its ownMPRs from its neighbors and only those nodesare responsible for forwarding the routingmessages.


J Intell Robot Syst

• Destination Sequenced Distance Vector(DSDV) is a table-driven proactive routingprotocol, which mainly uses the Bellman–Fordalgorithm with small adjustments to be moreappropriate for ad-hoc networks. In DSDV,each node maintains a routing table (withsequence number) for all other nodes, not justfor the neighbor nodes [44]. Whenever thetopology of the network changes, these changesare disseminated by the protocol’s updatemechanisms (periodic and/or triggered updates).These updates can result to routing-loopswithin the network. To solve this problem andto determine the freshness of a route, DSDVuses sequence numbers, which are assignedby destination nodes. A route with highersequence number is preferred instead of aroute with lower sequence number. The mainadvantages of DSDV are both the simplicity ofthe algorithm and the usage of these sequencenumbers, which guarantees the protocol to beloop free. However, it has some drawbacks.To enable up-to-date routing table, eachnode periodically broadcast routing tableupdates, and this brings an overhead to thenetwork.

4.3 Reactive Routing Protocols

Reactive Routing Protocol (RRP) is known as on-demand routing protocol, which means if thereis no communication between two nodes, thereis no need to store (or to try to store) a routebetween them. RRP is designed to overcome theoverhead problem of PRP. In RRP, a route be-tween communicating nodes is determined ac-cording to the demand from the source node.There are two different messages in this routingmodel: RouteRequest messages and RouteReplymessages. RouteRequest messages are producedand dispatched by flooding to the network bythe source node, and the destination node repliesto this message with a RouteReply message. Byreceiving a RouteReply message the communi-cation begins. As a result, each node maintainsonly the routes that are currently in use. There isno periodic messaging in this protocol; therefore,RRP is bandwidth-efficient. On the other hand,the procedure of finding routes can take a long

time; therefore, high latency may appear duringthe route finding process.

• Dynamic Source Routing (DSR) is a simpleand effective RRP, which is designed mainlyfor multi-hops for wireless mesh networks[45]. In DSR, the source node broadcasts aroute request message to its neighbor nodes,which are in the wireless transmission range.In the whole communication process, therecan be many route request messages. There-fore, to avoid confusion, the source node addsa unique request-id number to the producedmessage. DSR is a source demanding rout-ing protocol and the source node stores theentire hop-by-hop route of the destinationnode. If the source node is unable to use itscurrent route, due to changes in the networktopology, then the route maintenance mech-anism is activated. In such case, the sourcenode has to use another route to the des-tination; if there is none, a new route dis-covery phase is started. This routing pro-tocol was implemented by Brown et al. in[46] and they reported that finding a newroute in UAV network with DSR can beirritating.

• Ad-hoc On-demand Distance Vector (AODV)is a reactive protocol, which has sameon-demand characteristics with DSR withdifferent maintaining mechanisms of routingtable [47]. In AODV, each node stores a rout-ing table, which contains a single record foreach destination; while in DSR, each nodecan store multiple entries in its routing ta-ble for each destination. Another differencewith DSR stems out from the fact that DSRdata packets carry the complete path betweensource and the destination nodes. In AODV,the source node (and also other relay nodes)stores the next-hop information correspond-ing to each data transmission. AODV routingprotocol consists of three phases: route discov-ery, packet transmitting and route maintain-ing. If the source node has packets to send, itfirst starts a route discovery process to locatethe destination node and then dispatches thesepackets over a determined route. Discoveryprocess enables determined routes without a


J Intell Robot Syst

loop, and it uses a sequence number to deter-mine an up-to-date route of the destination.An expiration time is used to keep a route’sfreshness. In this process, intermediate nodesalso update their routing tables. After a route-id constructed, packets can be transferred overit. As a result of mobile nodes, some linkfailures may occur, and this connection losstriggers a repairing process to maintain theroutes.

4.4 Hybrid Routing Protocols

Hybrid routing protocol (HRP) is a combinationof previous protocols, and is presented to over-come their shortcomings. By using HRP, the largelatency of the initial route discovery process inreactive routing protocols can be decreased andthe overhead of control messages in proactiverouting protocols can be reduced. It is especiallysuitable for large networks, and a network isdivided into a number of zones where intra-zonerouting is performed with the proactive approachwhile inner-zone routing is done using the reactiveapproach.

• Zone Routing Protocol (ZRP) is based on theconcept of zones [48]. In this protocol, eachnode has a different zone, which is definedas the set of nodes whose minimum distance(in terms of number of hops) to this nodewhich is not greater than predefined radiusρ. Therefore, the zones of neighboring nodesoverlap. In MANET, the largest part of thewhole traffic is directed to nearby nodes. Therouting inside the zone is called as intra-zonerouting, and it uses proactive approach tomaintain routes. If the source and destinationnodes are in the same zone, the source nodecan start data transmission immediately. Theinter-zone routing is responsible for sendingdata packets to outside of the zone. It usesreactive approach to maintain routes. The de-lay caused by the route discovery is mini-mized by using bordercasting. Bordercasting isused instead of traditional broadcasting, andreply messages are only produced by bordernodes of a zone. These border nodes then

repeat either inter- or intra-zone routing asneeded.

• Temporarily Ordered Routing Algorithm(TORA) is a hybrid distributed routing pro-tocol for multi-hop networks, in which routersonly maintain information about adjacentrouters (i.e., one-hop knowledge) [49]. Its aimis to limit the propagation of control mes-sage in the highly dynamic mobile computingenvironment, by minimizing the reactions totopological changes. Although, it mainly usesa reactive routing protocol, it is also enhancedwith some proactive approaches. It builds andmaintains a Directed Acyclic Graph (DAG)from the source node to the destination. Thereare multiple routes between these nodes inDAG. It is preferred for quickly finding newroutes in case of broken links and for increas-ing adaptability. TORA does not use a short-est path solution, and longer routes are oftenused to reduce network overhead. Each nodehas a parameter value termed as “height” inDAG, and no two nodes have the same heightvalue. Data flow as a fluid from the highernodes to lowers. It is structurally loop-free be-cause data cannot flow to higher height nodes.In the route discovery phase, this height pa-rameter is returned to the querying node, andin this process all intermediate nodes updatestheir routing tables (TORA table) with theincoming route and heights.

5 Open Issues and Challenges

A FANET is somewhat different from traditionalMANETs and VANETs; however, the funda-mental idea is the same: having mobile nodesand networking in an ad-hoc manner. Hence,in a FANET, some challenges are valid as in aVANET while facing with additional challenges.Although, many researches have been performedto increase the efficiency of network with flyingnodes, there are still many unsolved problems,which should be explored in future works:

(1) National Regulations: UAVs are increas-ingly used in many application areas, and


J Intell Robot Syst

they get their places in the modern informa-tion age. While UAVs increasingly becomea part of each country’s national airspacesystem, most of countries’ current air regu-lations do not allow controlled UAV opera-tions in civil airspace. This can be seen as thebiggest current barrier to the development ofUASs in civilian areas. Therefore, there is aserious need to define distinctive rules andregulations to integrate UAV flights into thenational airspace.

(2) Routing: In a FANET, due to the fastmovement of UAVs, network topology canchange quickly. Data routing between UAVsfaces a serious challenge, which is differentfrom low mobility environment. The rout-ing protocols should be able to update rout-ing tables dynamically according to topologychanges. Most of previous routing algorithmsin MANET are partly fail to provide a reli-able communication between UAVs. There-fore, there is a need of developing new rout-ing algorithms and networking model forconstructing a flexible and responsive inte-gration model.

(3) Path Planning: In a large-scale mission areaand multi-UAV operation, cooperation andcoordination between UAVs are not onlydesirable but also crucial feature to increaseefficiency. In the operation theatre, therecan be some dynamic changes like addition/removal of UAVs, physical static obstacles,dynamic threats (such as mobile radars), etc.In such cases, each UAV has to changeits previous path, and new ones should bere-calculated dynamically. Thus, new algo-rithms/methods in dynamic path planningare required to coordinate the fleets ofUAVs.

(4) Quality of Service (QoS): A FANET can beused for many types of applications, and ittransports different types of data, which in-clude GPS locations, streaming video/voice,images, simple text messages, etc. FANETneed to support some service qualities tosatisfy a set of predetermined service per-formance constraints like delay, bandwidth,

jitter, packet loss, etc. Defining a compre-hensive framework for QoS-enabled middle-ware is a crucial challenge that should beovercome due to the highly mobile and dy-namic structure of FANET.

(5) Integration with a Global Information Grid(GIG): GIG is a worldwide surveillance net-work and computer system intended to pro-vide Internet-like capability that allows any-one connected to the system to collaboratewith other users and to get process and trans-mit information anytime and anywhere inthe world. A FANET should connect to fu-ture Information Grids as one of the maininformation platforms to increase efficiencyof a UAS by using a UAV’s communicationpackages, equipment suites, sensors, etc.

(6) Coordination of UAVs and manned air-crafts: It is inevitable that, in the future,flights of UAVs with other manned aircraftare likely to increase. This coordination willenable the destruction of enemy aircraft withminimal losses. At the same time, theseUAVs can be used as electronic jammersand for real time video reconnaissance inenemy areas. Therefore, the collaboration ofUAVs and manned aircraft should be in anetworked environment.

(7) Standardize FANETs: A FANET uses var-ious wireless communication bands suchas VHF, UHF, L-Band, C-band, Ku-Band,etc. These bands also used in different ap-plication areas like GSM networks, satel-lite communication, etc. To reduce the fre-quency congestion problem, there is a needto standardize these communications bands,signal modulation and multiplexing models.

(8) UAV mobility and placement: Mini-UAVsare smaller in size and can carry limitedpayloads, like a single radar, infrared cam-era, thermal camera, image sensor, etc. Ifthere is a need to use different sensors, theyshould be loaded on different UAVs, e.g.,one UAV can be loaded with an infraredcamera, while another UAV is equippedwith a high-resolution camera. This allowsmultiple images to be taken from the same


J Intell Robot Syst

area, which can be hundreds and thousandsof square meters. There is an open issue inthis topic to optimize the UAV placement toreduce energy consumption while increasingthe taken information.

6 Conclusion

The capabilities and roles of Unmanned AerialVehicles (UAVs) are promising, and they willplay an increasingly prominent role in a largeoperation area, in a broader range of applicationsand complicated missions. Progressively, UAVsneed to cooperate with each other in order toperform complex tasks especially in areas that arerelatively inaccessible from the ground, and thereis a need to quickly and easily deploy a networkedsystem. These cooperating UAVs form a multi-UAV system, which also aims to decrease themission completion time and increase reliabilityof the system for aerial missions compared with asingle large-UAV system. To increase the scalabil-ity of the system, there is a need of new network-ing standards concepts in multi-UAV systems.Networking of multi-UAVs is not only desirablebut also a crucial feature to increase the efficiencyof the system by ensuring connectivity of thesystems in non-LOS, urban, hostile, and/or noisyenvironmental management systems. Because ofthe highly mobile nodes, the networking struc-ture should be constructed in ad-hoc manner, andis called as Flying Ad-Hoc Network (FANET),which requires scalable, reliable, real-time andpeer-to-peer mobile ad-hoc networking betweenUAVs and ground stations.

Networking between UAVs is significantlydifferent from traditional ad-hoc networking as-sumptions Mobile Ad-Hoc Networks (MANET)and Vehicular Ad-Hoc Networks (VANET) interms of connectivity, data delivery, latency, ser-vice and etc. In this article, various networkingprotocols in FANET and its different applicationareas are surveyed. At the same time, it is aimedto motivate researchers to find solutions for theopen research issues, which have been detailed inthe paper.

References

1. Jiang, F., Swindlehurst, A.L.: Dynamic UAV Relay Po-sitioning for the Ground-to-Air Uplink. IEEE Globe-com Workshops (2010)

2. Sahingoz, O.K.: Multi-level dynamic key managementfor scalable wireless sensor networks with UAV. In:The 7th International Conference on Ubiquitous In-formation Technologies & Applications (CUTE 2012).Ubiquitous Information Technologies and Applica-tions. Lecture Notes in Electrical Engineering, vol. 214,pp. 11–19 (2013)

3. Sahingoz, O.K.: Large scale wireless sensor networkswith multi-level dynamic key management scheme. J.Syst. Archit. Available online 5 June 2013. ISSN 1383-7621. doi:10.1016/j.sysarc.2013.05.022 (2013)

4. Manathara, J., Sujit, P.B., Beard, R.: Multiple UAVcoalitions for a search and prosecute mission. J. Intell.Robot. Syst. 62(1), 125–158 (2011)

5. Cevik, P., Kocaman, I., Akgul, A., Akca, B.: The smalland silent force multiplier: a swarm UAV—electronicattack. J. Intell. Robot. Syst. 70(1–4), 595–608 (2013)

6. York, G., Pack, D.: Ground target detection using co-operative unmanned aerial systems. J. Intell. Robot.Syst. 65(1–4), 473–478 (2012)

7. Zhu, S., Wang, D., Low, C.: Ground target tracking us-ing UAV with input constraints. J. Intell. Robot. Syst.69(1–4), 417–429 (2013)

8. Merino, L., Caballero, F., Martínez-de-Dios, J.R.,Maza, I., Ollero, A.: An unmanned aircraft system forautomatic forest fire monitoring and measurement. J.Intell. Robot. Syst. 65(1–4), 533–548 (2012)

9. Cho, A., Kim, J., Lee, S., Kee, C.: Wind estimation andairspeed calibration using a UAV with a single-antennaGPS receiver and pitot tube. IEEE Trans. Aerosp.Electron. Syst. 47, 109–117 (2011)

10. Maza, I., Caballero, F., Capitan, J., Martinez-De-Dios,J.R., Ollero, A.: Experimental results in multi-UAVcoordination for disaster management and civil secu-rity applications. J. Intell. Robot. Syst. 61(1–4), 563–585(2011)

11. Xiang, H., Tian, L.: Development of a low-cost agricul-tural remote sensing system based on an autonomousunmanned aerial vehicle. Biosyst. Eng. 108(2), 174–190(2011)

12. Cho, A., Kim, J., Lee, S., Kee, C.: Wind estimation andairspeed calibration using a UAV with a single-antennaGPS receiver and pitot tube. IEEE Trans. Aerosp.Electron. Syst. 47(1), 109–117 (2011)

13. Xiang, H., Tian, L.: Development of a low-cost agricul-tural remote sensing system based on an autonomousunmanned aerial vehicle (UAV). Biosyst. Eng. 108(2),174–190 (2011)

14. Bekmezci, I., Sahingoz, O.K., Temel, S.: Flying Ad-HocNetworks (FANETs): a survey. Ad Hoc Netw. 11(3),1254–1270 (2013)

15. Cheng, B.N., Moore, S.: A comparison of MANETrouting protocols on airborne tactical networks.


http://dx.doi.org/10.1016/j.sysarc.2013.05.022

J Intell Robot Syst

In: Military Communications Conference—MILCOM2012, pp. 1–6 (2012)

16. Frew, E.W., Brown, T.X.: Airborne communicationnetworks for small unmanned aircraft systems. Proc.IEEE 96(12), 2008–2027 (2008)

17. Aloul, F.A., Kandasamy, N.: Sensor deployment forfailure diagnosis in networked aerial robots: a sat-isfiability-based approach, theory and applications ofsatisfiability testing, SAT 2007. Lect. Notes Comput.Sci. 4501, 369–376 (2007)

18. Shirani, R., St-Hilaire, M., Kunz, T., Zhou, Y., Li,J., Lamont, L.: Combined reactive-geographic routingfor unmanned aeronautical ad-hoc networks. In: 8thInternational Wireless Communications and MobileComputing Conference (IWCMC-2012), pp. 820–826(2012)

19. Li, Y., St-Hilaire, M., Kunz, T.: Enhancements toreduce the overhead of the reactive-greedy-reactiverouting protocol for unmanned aeronautical ad-hocnetworks. In: 8th International Conference on WirelessCommunications, Networking and Mobile Computing(WiCOM), pp. 1–4 (2012)

20. Cai, Y., Yu, F.R., Li, J., Zhou, Y., Lamont, L.: Mediumaccess control for Unmanned Aerial Vehicle (UAV)ad-hoc networks with full-duplex radios and multi-packet reception capability. IEEE Trans. Veh. Tech-nol. 62(1), 390–394 (2013)

21. Reynaud, L., Rasheed, T.: Deployable aerial commu-nication networks: challenges for futuristic applica-tions. In: Proceedings of the 9th ACM Symposium onPerformance Evaluation of Wireless Ad-Hoc, Sensor,and Ubiquitous Networks (PE-WASUN ‘12), pp. 9–16(2012)

22. Bök, P.-B., Tuchelmann, Y.: Context-aware QoS con-trol for wireless mesh networks of UAVs. In: Inter-national Conference Computer Communications andNetworks (ICCCN), pp. 1–6 (2011)

23. Li, Y., St-Hilaire, M., Kunz, T.: Improving routing innetworks of UAVs via scoped flooding and mobil-ity prediction. In: IFIP Wireless Days (WD), pp. 1–6(2012)

24. Rohde, S., Goddemeier, N., Daniel, K., Wietfeld, C.:Link quality dependent mobility strategies for distrib-uted aerial sensor networks. In: GLOBECOM Work-shops, pp. 1783–1787 (2010)

25. Sahingoz, O.K.: Mobile networking with UAVs: op-portunities and challenges. In: International Confer-ence on Unmanned Aircraft Systems (ICUAS-2013),pp. 933–941 (2013)

26. Namuduri, K., Wan, Y., Gomathisankaran, M., Pendse,R.: Airborne network: a cyber-physical system per-spective. In: Proceedings of the First ACM MobiHocWorkshop on Airborne Networks and Communica-tions (Airborne ‘12), pp. 55–60 (2012)

27. Clapper, J., Young, J., Cartwright, J., Grimes, J.: Un-manned systems roadmap 2007–2032. Technical Re-port, Department of Defense (2007)

28. Winnefeld, J.A., Kendall, F.: Unmanned systems inte-grated roadmap FY 2011-2036. Technical Report, De-partment of Defense (2011)

29. Hyland, M.T.: Performance evaluation of ad hoc rout-ing protocols in a swarm of autonomous unmannedaerial vehicles. PhD Thesis, Air Force Institute ofTechnology (2007)

30. Gu, D.L., Pei, G., Ly, H., Gerla, M., Zhang, B.,Hong, X.: UAV aided intelligent routing for ad-hocwireless network in single-area theater. In: IEEEWireless Communications and Networking Conference-(WCNC 2000), vol. 3, pp. 1220–1225 (2000)

31. Franchi, A., Secchi, C., Ryll, M., Bulthoff, H.H.,Giordano, P.R.: Shared control: balancing auton-omy and human assistance with a group of Quadro-tor UAVs. IEEE Robot. Autom. Mag. 19(3), 57–68(2012)

32. Cheng, C.M., Hsiao, P.H., Kung, H.T., Vlah, D.: Maxi-mizing throughput of UAV-relaying networks with theload-carry-and-deliver paradigm. In: IEEE WirelessCommunications and Networking Conference (WCNC2007) (2007)

33. Le, M., Park, J.S., Gerla, M.: UAV Assisted disruptiontolerant routing. In: Military Communications Confer-ence, MILCOM 2006, pp. 1–5 (2006)

34. Jonson, T., Pezeshki, J., Chao, V., Smith, K., Fazio, J.:Application of Delay Tolerant Networking (DTN) inairborne networks. In: IEEE Military CommunicationsConference-(MILCOM 2008), pp. 1–7 (2008)

35. Lamont, G.B., Slear, J.N., Melendez, K.: UAV swarmmission planning and routing using multi-objectiveevolutionary algorithms. In: IEEE Symposium onComputational Intelligence in Multicriteria DecisionMaking, pp. 10–20 (2007)

36. Sun, Z., Wang, P., Vuran, M.C., Al-Rodhaan, M.A.,Al-Dhelaan, A.M., Akyildiz, I.F.: BorderSense: borderpatrol through advanced wireless sensor networks. AdHoc Netw. 9(3), 468–477 (2011)

37. Ko, J., Mahajan, A., Sengupta, R.: A network-centricUAV organization for search and pursuit opera-tions. In: IEEE Aerospace Conference, pp. 2697–2713(2002)

38. López, J., Royo, P., Pastor, E., Barrado, C., Santa-maria, E.: A middleware architecture for unmannedaircraft avionics. In: ACM/IFIP/ USENIX Interna-tional Conference on Middleware companion (MC ‘07)(2007)

39. de Jong, E.: Flexible data-centric UAV platform easesmission adaptation. White paper, Available online.http://www.rti.com/whitepapers/RTI_Data-Driven_Approach_to_UAV.pdf (2011). Accessed 3 Aug 2013

40. Koller, A.A., Johnson, E.N.: Design, implementa-tion, and integration of a publish/subscribe-like Multi-UAV communication architecture. In: AIAA Mod-eling and Simulation Technologies Conference andExhibit, pp. 1–17 (2005)

41. Sahingoz, O.K., Sonmez, A.C.: Fault tolerance mech-anism of agent-based distributed event system. Lect.Notes Comput. Sci. 3993, 192–199 (2006)

42. Sahingoz, O.K., Erdogan, N.: MAPSEC: mobile-agentbased publish/subscribe platform for electronic com-merce. Lect. Notes Comput. Sci. 2869/2003, 348–355(2003)


http://www.rti.com/whitepapers/RTI_Data-Driven_Approach_to_UAV.pdf

http://www.rti.com/whitepapers/RTI_Data-Driven_Approach_to_UAV.pdf

J Intell Robot Syst

43. Clausen, T., Jacquet, P.: Optimized Link State Rout-ing Protocol (OLSR) RFC 3626. http://www.ietf.org/rfc/rfc3626 (2003). Accessed 3 Aug 2013

44. Perkins, C.E., Bhagwat, P.: Highly dynamic Destination-Sequenced Distance-Vector Routing (DSDV) for mo-bile computers. In: Proceedings of the Conference onCommunications Architectures, Protocols and Appli-cations (SIGCOMM ‘94), pp. 234–244 (1994)

45. Johnson, D.B., Maltz, D.A.: Dynamic source routingin ad hoc wireless networks. In: Mobile Computing,Chapter 5, pp. 153–181. Kluwer Academic Publishers(1996)

46. Brown, T.X., Argrow, B., Dixon, C., Doshi, S.,Thekkekunel, R.G., Henkel, D.: Ad hoc UAV ground

network (AUGNet). In: 3rd AIAA Unmanned Unlim-ited Technical Conference, pp. 29–39 (2004)

47. Murthy, S., Garcia-Luna-Aceves, J.J.: An efficientrouting protocol for wireless networks. In: ACMMobile Networks and Applications, pp. 183–197(1996)

48. Haas, Z.J., Pearlman, M.R.: Zone Routing Protocol(ZRP) a hybrid framework for routing in ad hoc net-works, 2nd edn. Ad hoc Networking, vol. 1, chapter 7,pp. 221–253. Addison-Wesley (2001)

49. Park, V., Corson, S.: Temporarily-Ordered Rout-ing Algorithm (TORA) Version 1. Internet draft,IETF MANET working group. http://tools.ietf.org/html/draft-ietf-manet-tora-spec-04. Accessed 3 Aug 2013


http://www.ietf.org/rfc/rfc3626

http://www.ietf.org/rfc/rfc3626

http://tools.ietf.org/html/draft-ietf-manet-tora-spec-04

http://tools.ietf.org/html/draft-ietf-manet-tora-spec-04

networking models in flying ad-hoc networks (fanets): concepts and...

Documents