mobile ad hoc networking: milestones, challenges, and new research directions

12
IEEE Communications Magazine • January 2014 85 0163-6804/14/$25.00 © 2014 IEEE 1 http://datatracker.ietf. org/wg/manet/charter/ INTRODUCTION The multihop (mobile) ad hoc networking paradigm emerged, in the civilian field, in the 1990s with the availability of off-the-shelf wireless technologies able to provide direct network con- nections among users devices: Bluetooth (IEEE 802.15.1) for personal area networks, and the 802.11 standards family for high-speed wireless LAN (see Chapters 2 and 3 in [1, 2]). Specifical- ly, these wireless standards allow direct commu- nications among network devices within the transmission range of their wireless interfaces, thus making the single-hop ad hoc network a reality, that is, infrastructureless WLAN/WPAN where devices communicate without the need for any network infrastructure (Fig. 1). The multihop paradigm was then conceived to extend the possibility to communicate with any couple of network nodes, without the need to develop any ubiquitous network infrastructure. In the ’90s, we assisted in the usage of the multihop paradigm in mobile ad hoc networks (MANETs), where nearby users directly communicate (by exploiting the wireless-network interfaces of their devices in ad hoc mode) not only to exchange their own data but also to relay the traffic of other network nodes that cannot directly commu- nicate, thus operating as routers do in the legacy Internet. For this reason, in a MANET, the users’ devices cooperatively provide the Internet services, usually provided by the network infra- structure (e.g., routers, switches, servers). At its birth, the MANET was seen as one of the most innovative and challenging wireless net- working paradigms [3], and was promising to become one of the major technologies, increasing- ly present in the everyday life of everybody. The potentialities of this networking paradigm made ad hoc networking an attractive option for build- ing fourth-generation (4G) wireless networks, and hence MANET immediately gained momentum, and this produced tremendous research efforts in the mobile network community [1, 2, 4]. The Internet model was central to the MANET Internet Engineering Task Force (IETF) working group, which, inheriting the TCP/IP protocols stack layering, assumed an IP- centric view of a MANET; see “Mobile Ad Hoc Networks (MANETs)” by J. P. Macker and M. S. Scott Corson in [1]. The MANET research community focused on what we call pure general- purpose MANETs, where pure indicates that no infrastructure is assumed to implement the net- work functions, and no authority is in charge of managing and controlling the network. General- purpose denotes that these networks are not designed with any specific application in mind, but rather to support any legacy TCP/IP applica- tion, as shown in Fig. 2. Following this view, the research focused on enhancing and extending the IP-layer routing and forwarding functionalities in order to sup- port the legacy Internet services in a network without any infrastructure. At the network layer, we observed a proliferation of routing protocol proposals as legacy Internet routing protocols developed for wired networks are clearly not suitable for the unpredictable and dynamic ABSTRACT In this article we discuss the state of the art of (mobile) multihop ad hoc networking. This paradigm has often been identified with the solutions developed inside the IETF MANET working group, and for this reason it is called the MANET paradigm. However, they do not coincide, and in the last decade they clearly diverged. In this article, we start from the rea- sons why the MANET paradigm did not have a major impact on computer communications, and we discuss the evolution of the multihop ad hoc networking paradigm by building on the lessons learned from the MANET research. Specifically, we analyze four successful networking paradigms, mesh, sensor, opportunistic, and vehicular networks, that emerged from the MANET world as a more pragmatic application of the multihop ad hoc networking paradigm. We also present the new research directions in the multihop ad hoc networking field: people- centric networking, triggered by the increasing penetration of the smartphones in everyday life, which is generating a people-centric revolution in computing and communications. AD HOC AND SENSOR NETWORKS Marco Conti, Italian National Research Council Silvia Giordano, University of Applied Technology of Southern Switzerland Mobile Ad Hoc Networking: Milestones, Challenges, and New Research Directions

Upload: silvia

Post on 23-Dec-2016

227 views

Category:

Documents


3 download

TRANSCRIPT

Page 1: Mobile ad hoc networking: milestones, challenges, and new research directions

IEEE Communications Magazine • January 2014 850163-6804/14/$25.00 © 2014 IEEE

1 http://datatracker.ietf.org/wg/manet/charter/

INTRODUCTION

The multihop (mobile) ad hoc networkingparadigm emerged, in the civilian field, in the1990s with the availability of off-the-shelf wirelesstechnologies able to provide direct network con-nections among users devices: Bluetooth (IEEE802.15.1) for personal area networks, and the802.11 standards family for high-speed wirelessLAN (see Chapters 2 and 3 in [1, 2]). Specifical-ly, these wireless standards allow direct commu-nications among network devices within thetransmission range of their wireless interfaces,thus making the single-hop ad hoc network areality, that is, infrastructureless WLAN/WPANwhere devices communicate without the need forany network infrastructure (Fig. 1).

The multihop paradigm was then conceived toextend the possibility to communicate with anycouple of network nodes, without the need todevelop any ubiquitous network infrastructure. In

the ’90s, we assisted in the usage of the multihopparadigm in mobile ad hoc networks (MANETs),where nearby users directly communicate (byexploiting the wireless-network interfaces of theirdevices in ad hoc mode) not only to exchangetheir own data but also to relay the traffic ofother network nodes that cannot directly commu-nicate, thus operating as routers do in the legacyInternet. For this reason, in a MANET, theusers’ devices cooperatively provide the Internetservices, usually provided by the network infra-structure (e.g., routers, switches, servers).

At its birth, the MANET was seen as one ofthe most innovative and challenging wireless net-working paradigms [3], and was promising tobecome one of the major technologies, increasing-ly present in the everyday life of everybody. Thepotentialities of this networking paradigm madead hoc networking an attractive option for build-ing fourth-generation (4G) wireless networks, andhence MANET immediately gained momentum,and this produced tremendous research efforts inthe mobile network community [1, 2, 4].

The Internet model was central to theMANET Internet Engineering Task Force(IETF) working group, which, inheriting theTCP/IP protocols stack layering, assumed an IP-centric view of a MANET; see “Mobile Ad HocNetworks (MANETs)” by J. P. Macker and M.S. Scott Corson in [1]. The MANET researchcommunity focused on what we call pure general-purpose MANETs, where pure indicates that noinfrastructure is assumed to implement the net-work functions, and no authority is in charge ofmanaging and controlling the network. General-purpose denotes that these networks are notdesigned with any specific application in mind,but rather to support any legacy TCP/IP applica-tion, as shown in Fig. 2.

Following this view, the research focused onenhancing and extending the IP-layer routingand forwarding functionalities in order to sup-port the legacy Internet services in a networkwithout any infrastructure. At the network layer,we observed a proliferation of routing protocolproposals as legacy Internet routing protocolsdeveloped for wired networks are clearly notsuitable for the unpredictable and dynamic

ABSTRACT

In this article we discuss the state of the artof (mobile) multihop ad hoc networking. Thisparadigm has often been identified with thesolutions developed inside the IETF MANETworking group, and for this reason it is calledthe MANET paradigm. However, they do notcoincide, and in the last decade they clearlydiverged. In this article, we start from the rea-sons why the MANET paradigm did not have amajor impact on computer communications, andwe discuss the evolution of the multihop ad hocnetworking paradigm by building on the lessonslearned from the MANET research. Specifically,we analyze four successful networkingparadigms, mesh, sensor, opportunistic, andvehicular networks, that emerged from theMANET world as a more pragmatic applicationof the multihop ad hoc networking paradigm.We also present the new research directions inthe multihop ad hoc networking field: people-centric networking, triggered by the increasingpenetration of the smartphones in everyday life,which is generating a people-centric revolutionin computing and communications.

AD HOC AND SENSOR NETWORKS

Marco Conti, Italian National Research Council

Silvia Giordano, University of Applied Technology of Southern Switzerland

Mobile Ad Hoc Networking: Milestones, Challenges, and New Research Directions

CONTI_LAYOUT.qxp_Layout 1/6/14 2:07 PM Page 85

Page 2: Mobile ad hoc networking: milestones, challenges, and new research directions

IEEE Communications Magazine • January 201486

nature of MANET topology [4]. Extensive effortshave been dedicated to building a set of stan-dard protocols. However, the released standardsprotocols, Ad Hoc On-Demand Distance Vector(AODV), Optimized Link State Routing(OLSR), Dynamic Source Routing (DSR), andTopology Broadcast Based on Reverse Path For-warding (TBRPF) (see the IETF MANET webpage1) have their pros and cons, and none ofthem is superior to the others in all contexts.Therefore, they are still under discussion asexperimental RFCs. Currently, the group is pur-suing a reactive (DYMO, i.e., AODV version 2)and a proactive protocol (OLSRv2).

The research interest rapidly spread fromrouting to all layers of the Internet protocolstack; [1, 2] present a complete view of MANETresearch from the physical up to the applicationlayer. On top of the IP, MANET generallyassumes the use of the UDP and TCP transportprotocols. Unfortunately, TCP does not workproperly in this scenario, as extensively discussedin the literature [2]. To improve the perfor-mance of TCP in a MANET, several proposalshave been presented. Most of these proposalsare modified versions of the legacy TCP used inthe Internet. However, TCP-based solutionsmight not be the best approach when operatingin MANET environments; hence, several authorshave proposed novel transport protocols tailoredto the MANET features. On top of that, middle-ware and applications constitute the less investi-gated areas in the MANET field. Indeed, in thedesign of general-purpose MANETs there wasnot a clear understanding of the applications forwhich multihop ad hoc networks are an opportu-nity. Lack of attention to the applications proba-bly constitutes one of the major causes for thenegligible MANET impact in the wireless net-working field. Lack of attention to the applica-tions also limited the interest to developmiddleware solutions tailored on MANETs [2].

In addition to an in-depth re-analysis of alllayers of the protocol stack, MANET researchalso focused on cross-layer research topics withspecial attention to energy efficiency, security,and cooperation [1].

After more than one decade of intenseresearch efforts, the MANET research field pro-duced profound theoretical results (e.g., perfor-mance bounds on MANET performance [5, 6]),or innovative protocol and architectural solu-tions (e.g., innovative cross-layer architecturesand protocols as discussed in [7, Chapter 1]), butin terms of real world implementations andindustrial deployments, the pure general-pur-pose MANET paradigm suffers from scarceexploitation and low interest in the industry andamong users [8]. Why has this happened? Aninitial answer to this question was provided in2007, in two companion articles that made a crit-ical analysis of the MANET research activities[8, 9] pointing out that the main reasons forMANET’s missed expectations are due to thelack of:• Implementation, integration, and experimenta-

tion• Simulation credibility• Socio-economic motivations

In addition, that analysis also highlightedthat the mesh network, vehicular network,opportunistic network, and sensor networkparadigms were or ig inat ing f rom theMANET research field. These multihop adhoc networking paradigms, by learning fromthe MANET experience, emerged with thepromise to avo id MANET’s mis takes .Indeed, these new “MANET-born”paradigms distanced themselves from themain weaknesses of the MANET by follow-ing a more pragmatic development approach.Currently, six years after that analysis, mesh,sensor , opportunist ic , and vehicular net-works are a reality in the mobile ad hoc net-work ing f ie ld , and the i r success can besummarized by the following pragmatic devel-opment strategy: 1 Application oriented development, which (as

opposed to the general-purpose design ofMANET) first identifies the applicationscenarios to be addressed before startingthe development of the technical solutions.

2 Complexity reduction. Depending on the spe-cific application scenario, some MANETconstraints have been relaxed; for example,the assumption that the network is com-posed only of users’ devices, and/or that thecommunication model has to comply withthe Internet one.

3 The focused research approach addressesonly the research topics relevant to buildingrobust and effective networks for support-ing the specific application scenario(s), andnot pretending to replace the Internet.

4 The use of realistic simulation models inorder to base the protocol development oncredible simulation studies.

5 The development of real network testbedswith the users’ involvement, in the earlystages of the design of these new paradigms,in order to put the users in the loop of thenetwork design and experimentation.

Figure 1. Single-hop ad hoc network.

Figure 2. The pure general-purpose MANET approach.

TCP/IP traffic

TCP/IP traffic

IP protocol

IP p

roto

col

IP protocol

IP protocol

IP protocol

IP p

roto

col

CONTI_LAYOUT.qxp_Layout 1/6/14 2:07 PM Page 86

Page 3: Mobile ad hoc networking: milestones, challenges, and new research directions

IEEE Communications Magazine • January 2014 87

In this article, we review the current status ofthe (mobile) multihop ad hoc networkingresearch. Specifically, we discuss the milestonesand challenges in mesh, sensor, opportunistic,and vehicular networks. We also discuss what isforeseen in the future of multihop ad hoc net-working. In particular, we discuss the people-centric revolution. Thanks to the increasingdiffusion of smartphones, the people-centricparadigm combines wireless communicationsand sensor networks with the daily life andbehaviors of people to build computing andcommunication solutions that are tightly coupledwith people.

We present and discuss mesh, sensor, oppor-tunistic, and vehicular networks, respectively.For each paradigm, we show how the pure andgeneral-purpose MANET paradigm (and therelated research results) has been turned into anetworking paradigm that has gained users’and market acceptance by exploiting thelessons learned from MANETs. For this rea-son, the presentation of each paradigm is orga-nized to highlight the main elements of thepragmatic development strategy: application-ori-ented development, problem complexity reduc-tion, focused research, use of realisticsimulation models, and development of realnetwork testbeds with the users’ involvement.Finally, we conclude by summarizing theparadigms, followed by the introduction of themulti-paradigm vision.

MESH NETWORKING: AN EFFECTIVELOW-COST EXTENSION OF THE

INTERNET

APPLICATION-ORIENTED PARADIGM

The design of the wireless mesh network (WMN)paradigm started from a well defined set ofapplication scenarios that is summarized in [10]:“providing a flexible and low cost extension ofthe Internet.” The initial WMN prototypes weremainly driven by the initiatives of communitiesof users with individual volunteers setting upIEEE 802-11-based long-distance point-to-pointlinks among their houses (Fig. 3) to build a com-munity network and offer a variety of services totheir participants, ranging from file sharing andcommunity-wide voice over IP (VoIP) to Inter-net access through community owned WMN-to-Internet gateways. Nowadays, metropolitan-scaleWMNs are a reality in many modern urban areassupported by municipalities and governmentorganizations2 offering a wide range of servicesranging from security surveillance to intelligenttransportation services.

COMPLEXITY REDUCTIONStarting from the well defined application sce-nario, it was immediately possible to reduce theMANET’s complexities. Specifically, the WMNparadigm introduces an architectural shift, withrespect to MANETs, by adopting a two-tier net-work architecture based on multihop communi-cations. A multihop wireless backbone is formedby dedicated (and often fixed) wireless meshrouters, which run a multihop routing strategy to

interconnect with each other, as illustrated inFig. 4. Some of the mesh routers act as gate-ways, providing the WMN with a direct connec-tion to the Internet and other wired/wirelessnetworks. Finally, to support seamless datatransport services for users’ devices (also calledmesh clients), mesh access points are connectedto the mesh routers to offer connectivity to meshclients.

Consequently, topology changes due to mobil-ity or energy issues do not influence traffic for-warding in WMNs as in MANETs, and theimpact of the mobility is restricted to the lasthop (i.e., the connection between the users andthe mesh access points).

FOCUSED RESEARCHFollowing the above architectural organization,the main challenges in WMNs have been subdi-vided into three main areas:• Exploiting the wireless mesh routers to

build a robust network backbone intercon-necting all mesh routers, and possibly somegateways to/from the Internet

• Defining a set of routing protocols that, byexploiting the network backbone, can iden-tify the best path(s) for traffic forwardinginside the WMN and to/from the Internet

• Supporting the users’ mobility among meshaccess pointsAll areas have been the subjects of intense

research activities providing effective solutionsto these problems. In particular, a large body ofresearch focused on building a robust wirelessmesh backbone by exploiting advanced multi-radio, multi-channel, and multi-rate capabili-ties, using heterogeneous wireless technologiesand types of antennas [11]. This means that aWMN environment is characterized by a richlink and path diversity, which provides anunprecedented opportunity to find paths thatcan satisfy the application requirements. Effec-tive channel assignment solutions have beendeveloped to construct robust and efficientwireless mesh backbones ([12, 13]). However,the multifaceted characteristics of a WMNcomplicate the path selection process. Forexample, the interference in WMNs is a verycomplex phenomenon, and the routing process

Figure 3. Community mesh networks.

2 See http://www.muni-wireless.com

CONTI_LAYOUT.qxp_Layout 1/6/14 2:07 PM Page 87

Page 4: Mobile ad hoc networking: milestones, challenges, and new research directions

IEEE Communications Magazine • January 201488

must be aware of the interference existingbetween links and traffic flows to take advan-tage of the multi-channel and multi-radio capa-bilities [14]. In addition, in most applicationscenarios the portion of traffic a mesh routerdelivers to other routers in the network (i.e.,intra-mesh traffic) will be minimal with respectto the traffic conveyed over connections estab-lished with external hosts (i.e., inter-mesh traf-fic). As a result, most WMN traffic is usuallybetween mesh clients and mesh gateway(s).This implies that traffic gets aggregated at themesh gateways, which can have a negative effecton network performance. For instance, depend-ing on the network topology and the routingstrategy, many mesh routers may select thesame gateway, and congestion levels can buildup excessively on the shared wireless channelaround the gateway. This may also lead touneven utilization of the gateways’ resources.As a consequence, routing protocols must care-fully take into consideration the distribution oftraffic loads [7, “Quality of Service in MeshNetworks”].

REALISTIC SIMULATION ANDEXPERIMENTATIONS WITH USERS

The development of the WMN paradigm hasexploited a combined and synergic use of simu-lation and experimentation [15]. The firstphase of this paradigm was dominated by theexperimentation of community networks main-ly based on the direct use of MANET proto-cols. The initial success of the WMN paradigmamong users stimulated in-depth researchactivities to develop protocols tuned to thespecific features of the WMN. In this secondphase, simulation and experimentation wereused in a joint way. Experiments were used toinvestigate the feasibility of the developedsolutions in real settings, while simulation stud-ies were used both in the early stages of theprotocol design (to compare and contrast alter-nate solutions) and to study the scalability ofthe proposed solutions on large networks, thusovercoming the limitations of the real testbedsin terms of network size and scenario diversity.Real testbeds were also useful to set up realis-tic simulation models and to validate the simu-lation results [7, “Experimental Work versusSimulation in the Study of Mobile Ad HocNetworks”].

CURRENT RESEARCH

Following the pragmatic development approach,currently WMN is a well-established network-ing paradigm for which several companies areoffering commercial solutions, and WMNs areused to extend Internet connectivity in severalmetropolitan areas. From a research stand-point, the fundamental aspects of WMNs havebeen investigated in depth, providing robustsolutions for supporting legacy Internet applica-tions. However, some research aspects of thistechnology are still under investigation to makeit more robust and able to support moreadvanced services requiring quality of service(QoS) support (e.g., video streaming applica-tions [16]), and operating in an energy-efficientway [17].

SENSOR NETWORKS: A MARKET SOLUTION TO A

LARGE RANGE OF PROBLEMS

APPLICATION-ORIENTED PARADIGM

Wireless sensor networks (WSNs) represent aspecial class of multihop ad hoc networks thatare developed to control and monitor a widerange of events and phenomena. Similar toWMNs, WSNs are successful in both academiaand industry. WSNs are deployed for specificapplication scenarios (e.g., precision agriculture[18] and structural monitoring [19]). Thus, thedesign of these networks highly depends on thespecific application scenario and the require-ments of the application in terms of reliability,timeliness, and so on. In a WSN, a number ofwireless nodes are deployed within the monitor-ing area, as illustrated in Fig. 5. These wirelessnodes can communicate directly among them-selves and typically follow the multihopparadigm (through the other sensor nodes) totransmit the collected information toward a sinknode. From the sink node, they eventually reachthe Internet.

COMPLEXITY REDUCTIONThe application driven view and the need tosolve a concrete problem was reflected in a natu-ral shift toward a more pragmatic approach. Infact, WSN research always has in mind the realdevelopment. Thus, in several cases, WSNs aredeveloped without mobility or with the supportof some higher-capacity nodes. For example,when the sensor node density is low, and hencethe sensor network is disconnected, mobile ele-ments (also referred to as data mules or messageferries) are used to collect the sensed data anddeliver them to the sink [20].

FOCUSED RESEARCHIn the last 10 years, WSNs have triggered inten-sive scientific activities, which has produced alarge body of literature to address the WSNresearch challenges, mostly centered on fourmain areas [21]:• Application network scenarios: with the cov-

erage of a wide range of application targetsand scenarios [22]

Figure 4. MESH network two-tier architecture.

TCP/IP traffic

Internet

CONTI_LAYOUT.qxp_Layout 1/6/14 2:07 PM Page 88

Page 5: Mobile ad hoc networking: milestones, challenges, and new research directions

IEEE Communications Magazine • January 2014 89

• Communication protocols for WSN: mediumaccess control (MAC) protocols, routing pro-tocols, clustering algorithms, security, net-works with mobile nodes, and so on [23, 24]

• Hardware and software sensor system: charac-teristics and requirements for a sensor node[25], system platforms and operating system(OS), testbeds, diagnostics, and debuggingsupport

• Network services and WSN optimization:from hardware advances to the understand-ing of physical-related phenomena with thedesign of algorithms and schemes for ener-gy efficiency, time and clock synchroniza-tion, security, coverage and connectivity,security and localization, network perfor-mances, and so on [26]

REALISTIC SIMULATION ANDEXPERIMENTATION WITH USERS

The need for real development produced anorthogonal approach to MANET validation.While initially simulation was largely used [27],it was quite soon replaced by large testbedsdeployed at several universities,3 which are openfor access to the research community. This allowsvalidation with real sensor nodes, and the finalresults, which are finally given to the end user,are necessarily collected in a real deployment.One of the largest sensor network deploymentsin the world (with about 5000 sensors connectedto same base station) is the GreenOrb4 WSN,which uses several types of sensors (for measur-ing temperature, humidity, illumination, carbondioxide, etc.) for forest surveillance.

CURRENT RESEARCHWe can argue that legacy WSN is a very wellconsolidated area, and mainly specialized net-work scenarios are still researched. However,some additional complementary issues are stillmatters of research, such as:• Problems related to privacy, security and

trust [28, 29]• Network performance [30]• User-friendly development [31]Furthermore, new WSN paradigms are emerg-ing, such as:• Energy harvesting WSNs• WSNs with robots• Underwater WSNs• The use of mobile phones as a human-cen-

tric sensing toolHereafter, we discuss the first three, while thefourth is discussed later.

Energy harvesting WSN: In [7, “Wireless Sen-sor Networks with Energy Harvesting”], theauthors explore the opportunities and challengesof energy-harvesting-based WSNs (EHWSNs),which result from endowing WSN nodes with thecapability of extracting energy from the surround-ing environment. Energy harvesting can exploitdifferent sources of energy, such as solar power,wind, mechanical vibrations, temperature varia-tions, magnetic fields, and so on. By continuouslyproviding energy and storing it for future use,energy harvesting subsystems enable WSN nodesto last potentially forever. This benefit openednew directions of investigation like efficient ener-

gy storing, optimal techniques for energy harvest-ing, and energy prediction models for predictablepower sources (mainly solar and wind). On top ofEHWSNs, task allocation protocols constitute ahighly studied area given that the design objec-tives of communication protocols have movedfrom energy conservation to opportunistic opti-mization of the use of the harvested energy.

WSNs with robots: Despite the plethora ofsuccessful application in several domains, WSNsare still prone to operational events that ariseduring deployment, maintenance, or replace-ment. Wireless sensor and robotic networks(WSRNs) entrust those tasks to resource-richmobile robots tending resource-constrained sta-tionary sensors [32]. Actions are collaborativelyexecuted by the robotic agents on the basis ofthe information received by deployed standalonesensing units, as surveyed in [7, “Robot-AssistedWireless Sensor Networks: Recent Applicationsand Future Challenges”]. Thanks to the latestadvances in multi-robot systems, the WSN’soperational tasks can be simplified and are nowstudied from the robot intervention standpoint,and distribution of task and multi-robot coordi-nation are now the major challenges.

Underwater WSN: Reference [7, “Underwa-ter Networks with Limited Mobility: Algorithms,Systems, and Experiments”] discusses how WSNscan be employed in the underwater environ-ment, where lack of feasible recharging solutionslimits power, and nodes leverage their limitedmobility for increased functionality and extendedlifetime. In this case, WSNs combine a fixedlocation with some mobility similar to that ofwater column profilers. However, to improvesensing quality while optimizing system lifetime,the system needs automated intelligence to alloweach node to optimally place itself, such that theoverall system best captures the variable of inter-est (which usually changes over both time andspace). Any solution must be robust to commu-nication failures in the network, as well as powerefficient. Reference [7, “Advances in Underwa-ter Acoustic Networking”] complements the pre-vious work with a comprehensive account of

Figure 5. WSNs are often pure, as are traditional MANETs, but they are appli-cation specific, as shown in those figures: wild animal monitoring, construc-tion monitoring, and fire prevention.

3 See, for example, Mote-labhttp://motelab.eecs.har-vard.edu/ , Twisthttp://www.tkn.tu-berlin.de/menue/telecom-munication_networks_group/ and Indriyahttp://indriya.comp.nus.edu.sg/

4

http://www.greenorbs.org/

CONTI_LAYOUT.qxp_Layout 1/6/14 2:07 PM Page 89

Page 6: Mobile ad hoc networking: milestones, challenges, and new research directions

IEEE Communications Magazine • January 201490

recent advances in underwater acoustic commu-nications and networking. Transmission tech-niques, architectures, protocols, and algorithmsfor underwater networking are being activelyresearched. In underwater WSNs, acoustic com-munications are a more valid solution than tradi-tional radio frequency (RF) communications,due to the latter’s transmission requirements insuch environment. Still, due to the physicalproperties of the propagation medium, underwa-ter acoustic signals suffer from severe transmis-sion loss, time-varying multipath propagation,Doppler spread, limited and distance-dependentbandwidth, and high propagation delay, generat-ing links unreliability and severe limits in theavailable bandwidth.

OPPORTUNISTIC NETWORKING: A FIRST STEP IN

PEOPLE-CENTRIC NETWORKING

APPLICATION ORIENTED PARADIGM

Opportunistic networking is one of the mostinteresting evolutions of the multihop network-ing paradigm. Indeed, while MANET representsan engineering approach to hide node mobilityby constructing “stable” end-to-end paths as inthe Internet, opportunistic networks do not con-sider node mobility a problem but as an oppor-tunity to exploit. In opportunistic networks themobility of nodes creates contact opportunitiesamong nodes, which can be used to connectparts of the network that are otherwise discon-nected. Specifically, according to this paradigm(also referred to as delay-tolerant or challengednetworks), nodes can physically carry buffereddata while they move around the network areauntil they get in contact with a suitable next-hopnode (i.e., until a forwarding opportunity exists).Thus, as opposed to a MANET, a node keeps onstoring data when no good next hop exists. Thisimplies that, with the opportunistic paradigm, adata can be delivered from a source toward adestination, even if an end-to-end path betweenthem never exits, by exploiting the sequence ofconnectivity graphs generated by nodes’ move-ment [33, 34]. This is a relatively youngparadigm, and opportunistic network research isstill ongoing. Therefore, one may argue that itsimpact is still to be proved. However, given thatwe can consider vehicular ad hoc networking(VANET) one of the most advanced and con-crete developments of the opportunistic net-working paradigm, we can claim that theopportunistic network paradigm already has asignificant role in the computer networking field.

In addition to VANET, other scenarios, moti-vating opportunistic network use, are discussedin [7, “Applications in Delay-Tolerant andOpportunistic Networks”]. Opportunistic net-working looks very suitable for communicationsin pervasive environments where the environ-ment is saturated by devices (with short-rangewireless technologies) that can self-organize in anetwork for local interactions among users. Inthese scenarios, the network is generally parti-tioned in disconnected islands, which might beinterconnected by exploiting the nodes’ mobility.

This implies a shift from legacy packet-basedcommunication, toward message-based commu-nication, bringing along new opportunities forapplication protocol design.

COMPLEXITY REDUCTIONOpportunistic networking departs from the Inter-net-oriented approach used in MANETs as itdoes not impose an end-to-end path from sourceto destination. This eliminates the huge effort ofhiding node mobility, which caused high com-plexity in MANETs. In opportunistic networkingthe multihop networking paradigm becomes besteffort (without the need to maintain paths ortables) and user-centric, and (human) mobilitybecomes an opportunity for communicating.

FOCUSED RESEARCHThree main directions have characterized theopportunistic networking research: mobility mod-els, routing protocols, and data dissemination.

Mobility models: Understanding and model-ing the properties of the human mobility is cru-cial for characterizing the constraints ofopportunistic communications, and designingpractical and effective forwarding strategies [35,36]. Major elements of human mobility charac-terization are the contact time, the distributionof the contact duration between two devices,and the inter-contact time (ICT), the distributionof the time between two consecutive contactsbetween devices. In particular, the characteriza-tion of the ICT distribution has generated agreat debate in the scientific community wheredifferent research groups have claimed com-pletely different results ranging from heavy-tailed distribution functions — with [37] orwithout [38] an exponential cutoff — to anexponential distribution [39]. In [40], the authorshave shown a fundamental result that helpsexplain the differences among the different ICTdistributions observed in the literature. Under-standing the properties of ICT distribution is acritical issue due to the fact that on this distri-bution depends the effectiveness of several rout-ing protocols for opportunistic networks. Forexample, in [38] the authors have shown that fora simple forwarding scheme, like the two-hopscheme, the expected delay for message for-warding might be infinite depending on theproperties of the ICT distribution. Reference [7,“Mobility Models in Opportunistic Networks”]introduces recent measurement studies onhuman mobility and realistic mobility models.

Routing: Opportunistic routing protocolshave to face three main problems:

•The uncertainty of future connectivity•The characteristics of human (or vehicu-lar) mobility•The heterogeneity of node resources andmobilityReplication techniques are used to cope with

the first obstacle, while utility-based forwardingtechniques are used for the second and third ones.As indicated in [7, “Opportunistic Routing”],opportunistic routing schemes are mainly built onone or more of the following basic mechanisms:• Mobility-assisted mechanism: Store-and-

carry a message until there is a communica-tion opportunity;

While a MANET

represents an

engineering

approach to hiding

node mobility by

constructing “stable”

end-to-end paths as

in the Internet,

opportunistic net-

works do not

consider node mobil-

ity as a problem but

as an opportunity to

exploit.

CONTI_LAYOUT.qxp_Layout 1/6/14 2:08 PM Page 90

Page 7: Mobile ad hoc networking: milestones, challenges, and new research directions

IEEE Communications Magazine • January 2014 91

• Local forwarding mechanism: Decisions aremade to bring the message probabilisticallycloser to its destination.

• Replication mechanism: Propagate multiplecopies of the same message in parallel, toincrease the probability of at least onebeing delivered.

• Coding mechanism: Source coding and net-work coding techniques are used to reduceperformance variability and improveresource usage.However, due to the large variety of scenarios

and characteristics, no single routing scheme isoptimal for every application scenario.

Data dissemination: This represents a con-crete application scenario in opportunistic net-working in which nodes usually play both rolesof “publisher” and “subscriber” at the sametime. Several dissemination strategies — classi-fied in six main categories in [7, “Data Dissemi-nation in Opportunistic Networks”] — havebeen proposed, based on the specific problemtargeted and the type of solution proposed:• Popularity-based strategies: drive the selec-

tion of content items to be cached based onthe popularity of the content itself

• Social-behavior-based strategies: selectionperformed based on users’ social dimension

• Publish/subscribe strategies: selection basedon content-centric overlays

• Global-optimality-based strategies: a globaloptimization solution approximated by alocal one

• Heterogeneous-technologies-based strate-gies: selection realized by combining thebroadband wireless infrastructure with anopportunistic network of user devices

• Peer-to-peer (p2p) strategies: selection per-formed using approaches similar to unstruc-tured p2p systems.

REALISTIC SIMULATION ANDEXPERIMENTATIONS WITH USERS

Learning from MANET experience, opportunis-tic-network researchers have spent lot of effortsto develop realistic simulators (e.g., ONE5) andexperimental testbeds [41]. As humans typicallycarry the devices, it is the human mobility thatgenerates the communication opportunities, asillustrated in Fig. 6. Thus, opportunistic-networkexperiments, necessarily, involves real users.Studying human mobility traces is the startingpoint to understand the properties of the humanmobility, with the aim of providing a characteri-zation of the temporal properties ofdevices/humans mobility. Therefore, these exper-iments are also used to derive mobility models,which are the basis for realistic opportunistic-network simulations.

CURRENT RESEARCHIn addition to the activity in three main areas dis-cussed above, the research community is current-ly investigating the interchange betweenopportunistic networking and the social charac-teristics of people. Indeed, mobile devices act asproxies of the humans in the cyber world andthus inherit the social links of their owners [42].In this sense, an emerging research area in oppor-

tunistic networking is related to the characteriza-tion of the social behaviors. In [43] there is a firstattempt to link the similarity of social character-istics (homophily) of humans with the probabilityof contacts, thus connecting the opportunisticnetworking with a large research area in humansciences. Another new relevant area that isemerging from opportunistic networking isopportunistic computing [41]: the services pre-sent on users devices can get opportunisticallydistributed and combined accordingly to usersinterests and needs. Projects as EU FP7-FIRESCAMPI (“Service platform for social AwareMobile and Pervasive computIng”), US NSFDoC (“Distributed Opportunistic Computing”),and EIT-ICT Lab MONC (Mobile OpportunisticNetworking and Computing) are opening theresearch path in opportunistic computing.

VEHICULAR NETWORKING

APPLICATION ORIENTED PARADIGMA vehicular ad hoc network (VANET) is a multi-hop ad hoc network made up of vehicles thatcommunicate among them by exploiting wirelesstechnologies (typically) belonging to the 802.11family. This is a specialization of multi-hop adhoc network paradigm well motivated by thesocio-economic value of advanced IntelligentTransportation Systems (ITS) aimed at reducingthe traffic congestions, the high number of traf-fic road accidents, etc. Indeed VANET can sup-port a large plethora of applications includingsafety traffic applications (e.g., collision avoid-ance, road obstacle warning, safety message dis-seminations, etc.), traffic information andinfotainment services (e.g., games, multimediastreaming, etc.). For example, a car involved inan accident can exploit the possibility to directlycommunicate with other vehicles to inform near-by vehicles of the dangerous situation. An exten-sive survey of the vehicular applications ispresented in [44].

COMPLEXITY REDUCTIONAdvanced ITS systems require both vehicle-to-roadside (V2R) and vehicle-to-vehicle (V2V)communications. In V2R communications a vehi-cle typically exploits infrastructure-based wirelesstechnologies, such as cellular networks, WiMAXand WiFi, to communicate with a roadside base

Figure 6. Opportunistic networking is based on human mobility.

5

http://www.netlab.tkk.fi/tutkimus/dtn/theone/

CONTI_LAYOUT.qxp_Layout 1/6/14 2:08 PM Page 91

Page 8: Mobile ad hoc networking: milestones, challenges, and new research directions

IEEE Communications Magazine • January 201492

station/access point. However the roadside unitsare not dense enough to guarantee the networkcoverage required by ITS applications and henceV2V communications are adopted to extend thenetwork connectivity/coverage and to guaranteebetter network performance.

V2V communications are based on the“pure” multi-hop ad hoc network paradigm asMANET. Specifically, according to thisparadigm, the vehicles on the road dynamicallyself-organize in a VANET by exploiting theirwireless communication interfaces, see Fig. 7.However the high level of vehicles’ mobility andthe possibility of sparse networking scenarios,which occur when the traffic intensity is low,make inefficient the legacy store-and-forwardcommunication paradigm used in MANET andpush toward the adoption of the more flexible,pragmatic and robust store-carry-and-forwardparadigm adopted by the opportunistic networks(see the “Opportunistic Networking” section). Inaddition, whenever possible, V2V communica-tions exploit V2R communications to make theircommunications more robust reducing someweaknesses and vulnerabilities of pure infra-structure-less communications.

The high socio-economic value of vehicularapplications pushed the international standard-ization bodies to develop technical specificationsto be adopted by vehicle industry. Among theseit is worth reminding the IEEE 1609 standardfamily for Wireless Access in Vehicular Environ-ment (WAVE) that has been developed uponthe IEEE 802.11p standard. Typically, powerconsumption is not an issue for this network asvehicle batteries are continuously recharged.

FOCUSED RESEARCHV2R communications exploit well-establishedtechnologies (WiFi, WiMAX, Zigbee, etc.) oper-ating in infrastructure-based mode (e.g., see“Enabling Technologies and Standards for MobileMultihop Wireless Networking” [7]). Their adop-tion in vehicular networks requires a careful anal-ysis of their performance when operating in highlydynamic environments where the connection timebetween the vehicle and the roadside unit may beshort and/or several vehicles are connected to thesame roadside unit (scalability issue).

The V2V research field inherited MANETresults related to multi-hop ad hoc routing/for-warding protocols [45], which have been tunedand modified for adapting them to the peculiarfeatures of the vehicular field [46]. The routingtask is challenging in VANET due to the highmobility of vehicles which are intermittently con-nected. However, in VANET, the mobility of thenetwork nodes (i.e., vehicles) is constrained bythe road characteristics and the other vehiclesmoving along the road. Special attention hasbeen reserved to the development of optimizedone-to-all (in a specific region) routing protocolsas several applications developed for VANETuse broadcasting (geocasting) communicationservices [47] to distribute a message from a vehi-cle to all other vehicles (in a given area). Broad-casting or geocasting are also the basiccommunication services for content dissemina-tion in a VANET, i.e., the dissemination toother vehicles of a file containing relevant infor-mation (e.g., a city map or infotainment infor-mation such as a music mp3 file). Due to theintermittent connectivity conditions, the oppor-tunistic paradigm applied to vehicular networkshas recently generated a large body of literaturemainly on routing protocols and data dissemina-tion in vehicular networks (e.g., [48, 49]).

REALISTIC SIMULATION ANDEXPERIMENTATIONS WITH USERS

Both simulation and experimentation have amajor role in designing and evaluating the solu-tions developed for vehicular networks. In thesimulation of vehicular networks a lot of atten-tion has been dedicated to develop realisticmodels of roads and of the vehicle mobility byexploiting the extensive literature developed inthe field of transportation systems, e.g., modelsof how cars move along a road taking into con-sideration their speeds, the distance amongthem, traffic signals, the road layout, etc [50].

Several well-known network simulator likens-2, ns-3, SWANS, OMNET++, OPNET, etc.,are able to take as input trace files, generated byspecialized vehicle mobility models (e.g., Vanet-MobiSim and SUMO), which provide realisticvehicle mobility behaviours. However, in somescenarios, the interdependences among vehiclecommunications and their mobility patternsmake the problem more complex. For example,when a congestion/accident occurs the communi-cation among vehicles influences their mobilityby triggering a road change and/or a speedreduction. For this kind of simulation the aboveapproach is not suitable and therefore theVANET community is currently developing sim-ulators which are able to take into account theinterdependencies between the vehicles’ mobilityand their communications. For example, TraNS(Traffic and Network Simulation environment)integrates ns-2 and SUMO by providing feed-backs from the network simulator (ns-2) toaffect the mobility traces produced by theSUMO mobility simulator.

Another important aspect to consider (toproduce realistic VANET simulations) is relatedto the simulation of the wireless channel amongvehicles and to/from roadside units taking into

Figure 7. VANET.

CONTI_LAYOUT.qxp_Layout 1/6/14 2:08 PM Page 92

Page 9: Mobile ad hoc networking: milestones, challenges, and new research directions

IEEE Communications Magazine • January 2014 93

account how the radio signal propagates in thisenvironment.

An updated and in-depth discussion ofVANET simulation models and tools can befound in “Mobility Models, Topology and Simu-lations in Vanet” [7].VANET prototypes havebeen extensively used to verify the feasibility ofVANET solutions and the effectiveness ofVANET applications. For these reasons severaltestbeds have been developed worldwide. Amongthese, it is worth remembering the CarTel projectat MIT (which developed a 27-car testbed fortesting V2V, V2R, road surface monitoring, etc.),the DiselNet project at UMass (which is basedon 35 buses which implement a DTN-based com-munication paradigm), the ShanghaiGrid project(which involves thousands of taxies and buses forvehicle and traffic monitoring and environmentsensing) and the pioneering FleetNet project(involving several companies and universities inGermany to demonstrate their platform for inter-vehicle communications). An updated survey ofVANET experimental activities can be found in“Experimental work on Vanet” [7].

CURRENT RESEARCHSeveral interesting and challenging issues arestill to be addressed in the VANET field; a spe-cial attention should be reserved to develop real-istic models to characterize the mobility of theVANET nodes [50], and to analytically study theperformance of 802.11 technology in VANET[51]. Current research is also intensive on datacommunication. A survey is presented in thechapter by Awwad, Yi, and Stojmenovic in [7]. Afsew other chapters in the same book survey theongoing research activities on experimentaltestbeds and simulation tools.

THE PEOPLE-CENTRIC REVOLUTION

APPLICATION-ORIENTED PARADIGMIn mobile multihop networking, the technologycharacteristics can be naturally combined withpeople’s behaviors and daily life events, allowingthe definition of a first example of a networkingparadigm that puts people in the center, as thenetwork is built with the people’s devices. Thishas become more and more true and real withthe recent emergence of smartphones. In fact,the major innovation introduced by mobilephones is that they do move, and move with peo-ple (the owners). For this reason, the networkingparadigms based on those devices (e.g., mobilephone sensing, opportunistic networking/comput-ing, or mobile cloud computing) are shifting froma device-to-device paradigm toward a person-to-person paradigm: the people-centric vision [52].This vision, as clarified below, considers theknowledge and experience in MANETs. As foropportunistic networking, being centered aroundpeople’s mobile devices makes this paradigm veryapplication oriented, while opening up newresearch challenges related to human mobility.

COMPLEXITY REDUCTIONThe people-centric paradigm privileges the top-down approach, taking into account people’sneeds and constraints first. The network, interms of connections, data, and resources, builds

with people, leaving no room for traditional sys-tem-oriented approaches. This generates a cyber-physical convergence as the mobile phone actsas a proxy of human behavior in the cyber world.While people really move into the physicaldimension, their mobile phones project theirmovements in the cyber world.

REALISTIC SIMULATION ANDEXPERIMENTATIONS WITH USERS

Clearly, evaluation of people-centric solutionsmust be done by users. Even in the case of simu-lation, some real human traces are necessary.

FOCUSED CURRENT RESEARCHAs discussed below, the main directions of peo-ple-centric networking are: • Mobile phone sensing: where billions of

users’ mobile devices/phones are used aslocation-aware data collection instrumentsfor real world observations

• Mobile phone cloud: where people’s devicesare used to offer a cloud computing service

MOBILE PHONE SENSINGIn mobile phone sensing (also known as crowdsensing) the physical world is sensed withoutdeploying a sensor network: it creates a participa-tory sensor network where people take an activerole in the decision stages of the sensing system[53]. A participatory system provides tools toshare, publish, search, interpret, and verify infor-mation collected using a custodian device [54]. Inopportunistic sensing the sensing activities are per-formed by the (opportunistic) exploitation of allthe sensing devices available in the environmentwhenever there is a match with the applicationrequirements. In addition, multi-modal sensorsspread in the environment can be exploitedopportunistically to infer precise informationabout the social behavior of the users and thesocial environment around them. Indeed, partici-patory and opportunistic sensing offers unprece-dented opportunities for pervasive urban sensing[55]: to effectively collect and process the digitalfootprints generated by humans when interactingwith the surrounding physical world and the socialactivities therein. A major goal of these sensingactivities is to investigate the hybrid city (i.e., acity that operates simultaneously in the cyber/digi-

Figure 8. Mobile phone sensors used for participatory sensing.

CONTI_LAYOUT.qxp_Layout 1/6/14 2:08 PM Page 93

Page 10: Mobile ad hoc networking: milestones, challenges, and new research directions

IEEE Communications Magazine • January 201494

tal and physical realms) to investigate humanbehavior and socio-economic relationships ([56,57]). This is a highly challenging and innovativeresearch objective that can bring the developmentof novel urban applications that benefit citizens,urban planners, and policy makers. Preserving theprivacy of individuals while contributing theirsensed data is a major challenge for progressingtoward pervasive urban sensing [58].

MOBILE PHONE CLOUDCloud computing has gained momentum in thelast years as a response to the high increase ofservice requests derived from new high resource-consuming applications: the computing resources(hardware and software) are delivered as a ser-vice over the Internet. While this new servicemodel has shown to be very powerful, it is asso-ciated with a high energy cost in accessing theservices themselves. As shown by several notablestudies in the smartphone community [59, 60],accessing remote computing resources requiresend devices to invest a significant amount oftheir energy resources.

Mobile smart devices are becoming more andmore capable, and will be increasingly able tooffer more and more services, effectively servingas a mobile cloud. By pooling their resources andencouraging their opportunistic exploitations byresource-poor devices, mobile smart devices canstreamline the sustainable scalability of theFuture Internet by bringing services andresources closer to where they are needed, there-by lowering the energy premium paid byresource-constrained systems [52].

The importance of mobile cloud computinggrows with the proliferation of mobile phonedevices and ubiquitous computing resources.Rather than seeking cloud services from a far-away data center, we envision that tomorrow’sembedded systems will seek cloud services fromsmart mobile devices in their surroundings [52].This will enable the exploitation of the vast arrayof resources available on user devices and, at thesame time, permit significant energy savings forembedded systems, which will not have to pay theenergy premium required to access a distantcloud. However, there is clearly a trade-off: the(exclusive) local execution of computationally

intensive tasks is necessarily a suboptimalapproach. At the opposite end, cloud computingoffers high-end resources at a non-negligible ener-gy cost. If we stop looking at mobile communica-tion technologies simply as a means to connect amobile device to the infrastructure, and abolishthe strict consumer–operator service model, wecan begin to see the true potential of today’smobile devices and the additional communicationopportunities arising from direct device-to-devicecontacts [52]. This is clearly centered aroundusers: they play both the roles of consumer andproducer, sometimes eager to share their resourceswith other people. Also, some more popularresources are likely to exist on devices close to theuser, thus allowing energy and time savings.

This naturally leads to the opportunistic com-puting vision, where services can be combinedacross multiple nodes, are offered by any node,and can be offloaded to any node (and not just aspecial subset of infrastructure nodes). Thus,opportunistic computing offers a distributedmobile cloud, that is, a computing system that har-nesses the CPU, memory, energy, and sensingresources of multiple nodes of heterogeneouscapabilities that collectively form a cloud that isdistributed in both space and time [52]. Each nodein the distributed cloud corresponds to a pairwiseencounter, and may therefore provide useful localcontext information, which is something that a dis-tant computing cloud would not be able to give.

Specifically, as indicated in Fig. 9, individualdevices may combine and exploit each other’sresources to boost their computing power andovercome the limitations of their own resources[41], without the communication energy footprintand extreme centralization of cloud computing.Therefore, the distributed mobile cloud modeloffered by opportunistic computing can be seenas the means to provide energy-efficient comput-ing access to resource-constrained devices.

SUMMARY AND CONCLUSIONS:TOWARD A MULTI-PARADIGM ERA

While the MANET stimulated some decades ofresearch efforts, it did not significantly impact thewireless networking market, due to several weak-nesses in its original design and in the researchapproach, as extensively discussed in previousworks [7, Chapter 1; 8]). However, the lessonslearned from MANET experience provide the basisfor a pragmatic development strategy that has beenapplied in the development of several “MANET-born” communication network paradigms— mesh,sensor, opportunistic, vehicular, and people-centricnetworking – which are, or soon will be, successfuland widespread. In this article we have extensivelydiscussed these emerging networking paradigms,showing that their design and deployment can bedescribed according to the five points of the prag-matic development strategy: application orienteddevelopment, complexity reduction, focusedresearch, realistic simulation models, and develop-ment of real network testbeds.

For each paradigm, we have summarized themain features, research challenges and solutions,and current status of deployment. While mesh,sensor, and vehicular networks already have a

Figure 9. Mobile phone cloud — users can get computing facilities from otherusers.

CONTI_LAYOUT.qxp_Layout 1/6/14 2:08 PM Page 94

Page 11: Mobile ad hoc networking: milestones, challenges, and new research directions

IEEE Communications Magazine • January 2014 95

major role in the pervasive networking market,the opportunistic networking and people-centricparadigms are attracting a growing intereststrongly coupled with the increasing use ofsmartphones, and are expected to become cen-tral in the coming years. Table 1 presents a sum-mary of the current status of research andmarket trends in these paradigms, and comparesthem with MANET.

It can be expected that in the near future theabove paradigms will be mixed together and/orintegrated with infrastructure-based networks,thus generating novel networking paradigms,merging the self-* feature properties of multihopad hoc networks and the reliability and robust-ness features of infrastructure-based networks.Major examples of this trend are hybrid cellular-opportunistic networks, which aim to mitigatethe cellular data traffic problems (by offloadingsome traffic by opportunistically exploiting theusers’ devices), and vehicular networks, whereopportunistic networking techniques are exploit-ed to overcome the coverage limitations of road-side infrastructure-based networks. Moregenerally, we can envisage complex applicationscenarios where several networking paradigmsare mixed together, like the one illustrated inFig. 10, where each paradigm does not act in iso-lation, but there is sharing among them, withmutual benefit: the multi-paradigm era.

REFERENCES[1] S. Basagni et al., Eds., Mobile Ad Hoc Networking, IEEE

Press and Wiley, 2004.[2] I. Chlamtac, M. Conti, and J. Liu, “Mobile Ad Hoc Net-

working: Imperatives and Challenges,” Ad Hoc NetworkJ., vol. 1, no. 1 Jan.–Feb.–Mar., 2003.

[3] S. Giordano and W. W. Lu, “Challenges in Mobile AdHoc Networking,” IEEE Commun. Mag., vol. 39, no. 6,2001.

[4] S. Giordano. “Mobile Ad Hoc Networks,” Handbook ofWireless Networks and Mobile Computing, Wiley, pp.325–46.

[5] M. Grossglauser and D. N. C. Tse, “Mobility Increasesthe Capacity of Ad Hoc Wireless Networks,” IEEE/ACMTrans. Net., vol. 10, no. 4, 2002, pp. 477–86.

[6] P. Gupta and P. R. Kumar, “The Capacity of WirelessNetworks,” IEEE Trans. Info. Theory IT, vol. 46, no. 2,2000, pp. 388–404.

[7] S. Basagni et al., Eds., Mobile Ad Hoc Networking —Cutting Edge Directions, IEEE Press and Wiley, 2013.

[8] M. Conti and S. Giordano, “Multihop Ad Hoc Network-ing: The Theory,” IEEE Commun. Mag. Feature Topic onAd Hoc and Sensor Networks, vol. 45, no. 4, Apr. 2007,pp. 78–86.

[9] M. Conti and S. Giordano, “Multi-Hop Ad Hoc Network-ing: The Reality,” IEEE Commun. Mag., Feature Topic onAd Hoc and Sensor Networks, vol. 45, no. 4, Apr. 2007,pp. 88–95.

[10] R. Bruno, M. Conti, and E. Gregori, “Mesh Networks:Commodity Multihop Ad Hoc Networks,” IEEE Com-mun. Mag., Mar. 2005, pp. 123–31.

[11] B. Bakhshi, S. Khorsandi, and A. Capone, “On-LineJoint QoS Routing and Channel Assignment in Multi-Channel Multi-Radio Wireless Mesh Networks,”Comp. Commun., vol. 34, issue 11, July 2011, pp.1342–60.

[12] H. Skalli et al., “Channel Assignment Strategies for Multi-Radio Wireless Mesh Networks: Issues and Solutions,” IEEECommun. Mag., vol. 45, no. 11, 2007, pp. 86–95.

[13] A. A. Franklin, A. Balachandran, and C. Siva Ram Murthy, “Online Reconfiguration of ChannelAssignment in Multi-Channel Multi-Radio Wireless MeshNetworks,” Comp. Commun., vol. 35, issue 16, 2012,pp. 2004–13.

Table 1. Summary of MANET-born paradigms.

MANET

Market relevance Application scenarios Research interest Open research issues

Decreasing Niche (military and safety) Low and decreasing Standardization

WMN High Smart cities Medium butdecreasing QoS, performance optimization

WSN HighAmbient intelligence,

pervasiveness, monitoring, andcontrolling

HighUser-friendly development,

performance, mobility, WSRN, under-water, and acoustic sensor networks

OppNet Medium butincreasing

Mobile social networkingData dissemination, offloading Increasing

Privacy, performance,communication approaches, and

architectures

VANET High Road safety, traffic efficiency,infotainment High QoS support, reliable broadcasting

protocols, and standardization

People-centric

Medium butincreasing

Social-based scenarios,mobile phone scenarios Increasing Privacy, performance, service

approaches, and architectures

Figure 10. The multi-paradigm era.

CONTI_LAYOUT.qxp_Layout 1/6/14 2:08 PM Page 95

Page 12: Mobile ad hoc networking: milestones, challenges, and new research directions

IEEE Communications Magazine • January 201496

[14] K. R. Chowdhury, M. Di Felice, and L. Bononi,“XCHARM: A Routing Protocol for Multi-Channel Wire-less Mesh Networks,” Comp. Commun., vol. 36, issue14, 2013, pp. 1485–97.

[15] K. Tan et al., “Comparing Simulation Tools and Experi-mental Testbeds for Wireless Mesh Networks,” Perva-sive and Mobile Computing, vol. 7, issue 4, Aug. 2011,pp. 434–48.

[16] R. Bruno, M. Conti, and A. Pinizzotto, “Routing Inter-net Traffic in Heterogeneous Mesh Networks: Analysisand Algorithms,” Performance Evaluation J., vol. 68,no. 9, 2011, pp. 841–58.

[17] A. de la Oliva, A. Banchs, and P. Serrano, “Throughputand Energy-Aware Routing for 802.11 based Mesh Net-works,” Comp. Commun., vol. 35, issue 12, 2012, pp.1433–46.

[18] S. Li et al., “NCOME: Practical Land Monitoring in Pre-cision Agriculture with Sensor Networks,” Comp. Com-mun., vol. 36, issue 4 2013, pp. 459–67.

[19] G. Hackmann et al., “A Holistic Approach to Decentral-ized Structural Damage Localization Using Wireless Sen-sor Networks,” Comp. Commun., vol. 36, issue 1 2013,pp. 29–41.

[20] M. Di Francesco, S. K. Das, G. Anastasi, “Data Collec-tion in Wireless Sensor Networks with Mobile Elements:A Survey,” TOSN, vol. 81, no. 7, 2011.

[21] J. Yick, B. Mukherjee, and D. Ghosal, “Wireless SensorNetwork Survey,” Comp. Net., 5212, 2008, pp.2292–330.

[22] T. Arampatzis, J. Lygeros, and S. Maesis, “A Survey ofApplications of Wireless Sensors and Wireless SensorNetworks,” IEEE Intelligent Control, 2005, pp. 719–24.

[23] I. F. Akyildiz et al., “Wireless Sensor Networks: A Sur-vey,” IEEE Commun. Mag., 2002, vol. 40, pp. 102–14.

[24] C. Konstantopoulos et al., “Special Issue: ReactiveWireless Sensor Networks,” Comp. Commun., vol. 36,issue 9 2013, pp. 963–64.

[25] M. Augusto et al., “Survey on Wireless Sensor NetworkDevices,” Proc. ETFA’03, IEEE, 2003, vol. 1, pp. 537–44.

[26] A. Munir and A. Gordon-Ross, 2010, “OptimizationApproaches in Wireless Sensor Networks,” SustainableWireless Sensor Networks, Y. Kheng Tan, Ed.

[27] I. Stojmenovic, “Simulations in Wireless Sensor and AdHoc Networks: Matching and Advancing Models, Met-rics and Solutions,” IEEE Commun. Mag., vol. 46, no.12, Dec. 2008, pp. 102–07.

[28] N. Li et al., “Privacy Preservation in Wireless SensorNetworks: A State-of-the-Art Survey,” Ad Hoc Networks— Elsevier, Apr. 2007, pp. 1501–14.

[29] X. Chen, “Sensor Network Security: A Survey,” IEEECommun. Surveys & Tutorials, vol. 112, 2009, pp.52–73.

[30] D Puccinelli et al., “The Impact of Network Topologyon Collection Performance,” Wireless Sensor Networks,2011, pp. 17–32.

[31] A. Foerster et al., “FLEXOR: User Friendly Wireless Sen-sor Network Development and Deployment,” 13th IEEEInt’l. Symp. a World of Wireless, Mobile and MultimediaNetworks WoWMoM ‘12, San Francisco, CA, June 2012.

[32] J. Chen, H. Frey, and X. Li, Special Issue: Wireless Sen-sor and Robot Networks: Algorithms and Experiments,Comp. Commun., vol. 35, issue 9, 2012, pp. iii–iv.

[33] U. G. Acer, P. Drineas, and A. A. Abouzeid, “Connec-tivity in Time-Graphs,” Pervasive and Mobile Comput-ing, vol. 7, issue 2 2011, pp. 160–71.

[34] S. Ferretti, “Shaping Opportunistic Networks,” Comp.Commun., vol. 365, 2013, pp. 481–503.

[35] D. Karamshuk et al., “Human Mobility Models forOpportunistic Networks,” IEEE Commun. Mag., vol.4912, Dec. 2011, pp. 157–65.

[36] T. Hossmann et al., “Collection and Analysis of Multi-Dimensional Network Data for Opportunistic Network-ing Research,” Comp. Commun., vol. 35, issue 13,2012, pp. 1613–25.

[37] T. Karagiannis et al., “Power Law and ExponentialDecay of Intercontact Times between Mobile Devices,”IEEE Trans. Mobile Computing, vol. 9, 2010, pp.1377–90.

[38] A. Chaintreau et al., “Impact of Human Mobility onOpportunistic Forwarding Algorithms,” IEEE Trans.Mobile Computing, vol. 6, no. 6, June 2007, pp. 606–20.

[39] W. Gao et al., “Multicasting in Delay Tolerant Net-works: A Social Network Perspective,” Proc. ACM Mobi-Hoc 2009.

[40] A. Passarella and M. Conti, “Characterising AggregateInter-Contact Times in Heterogeneous OpportunisticNetworks,” Proc. IFIP TC6 Networking 2011, Valencia,Spain, May 2011, pp. 301–13.

[41] M. Conti et al., “From Opportunistic Networks toOpportunistic Computing,” IEEE Commun. Mag., Sept.2010, vol. 489, pp. 126–39.

[42] A. Passarella et al., “Ego Network Models for FutureInternet Social Networking Environments,” Comp. Com-mun., vol. 3518, 2012, pp. 2201–17.

[43] A. Förster et al., “On Context Awareness and SocialDistance in Human Mobility Traces,” ACM MobiOpp2012, Zurich, Mar. 2012.

[44] E. Hossain et al., “Vehicular Telematics over Heteroge-neous Wireless Networks: A Survey,” Comp. Commun.,vol. 33, issue 7, 2010, pp. 775–93.

[45] F. Li and Y. Wang, “Routing in Vehicular Ad Hoc Net-works: A Survey,” IEEE Vehic. Tech. Mag., June 2007,pp. 12–22.

[46] H. Hartenstein and K. P. Laberteaux, “A Tutorial Surveyon Vehicular Ad Hoc Networks,” IEEE Commun. Mag.,June 2008, pp. 164–71.

[47] N. Wisitpongphan et al., “Broadcast Storm MitigationTechniques in Vehicular Ad Hoc Networks,” IEEE Wire-less Commun., vol. 14, no. 6, Dec. 2007, pp. 84–94.

[48] J. Burgess et al., MaxProp: Routing for Vehicle-BasedDisruption-Tolerant Networks,” Proc. IEEE INFOCOM,2006.

[49] S. Panichpapiboon and W. Pattara-Atikom, “A Reviewof Information Dissemination Protocols for VehicularAd Hoc Networks,” IEEE Commun. Surveys & Tutorials,vol. 14, issue 3, 2012, pp. 784–98.

[50] J. Harri, F. Filali, and C. Bonnet, “Mobility Models forVehicular Ad Hoc Networks: A Survey and Taxonomy,”IEEE Commun. Surveys & Tutorials, vol. 11, no. 4, 2009,pp. 19–41.

[51] R. Bruno and M. Conti, “Throughput and FairnessAnalysis of 802.11-Based Vehicle-to-Infrastructure DataTransfers,” Proc. IEEE MASS 2011, Valencia, Spain, Oct.2011.

[52] S. Giordano and D. Puccinelli, “The Human Element asthe Key Enabler of Pervasivness,” Proc. 10th IEEE IFIPAnnual Mediterranean Ad Hoc Networking Wksp. Med-HocNet 2011, June 2011, Italy.

[53] N. D. Lane et al., “Urban Sensing Systems: Opportunis-tic or Participatory?,” Proc. 9th Wksp. Mobile Comput-ing Systems and Applications, 2008, pp. 11–16.

[54] N. D. Lane et al., “Urban Sensing: Opportunistic orParticipatory,” Proc. 1st Wksp. Sensing on EverydayMobile Phones in Support of Participatory Research,Sydney, Australia, Nov. 6, 2007.

[55] D. Cuff, M. H. Hansen, and J. Kang, “Urban Sensing: Outof the Woods,” Commun. ACM, 513, 2008, pp. 24–33.

[56] E. Miluzzo et al., “Tapping into the Vibe of the CityUsing VibN, a Continuous Sensing Application forSmartphones,” Proc. 1st Int’l. Symp. from Digital Foot-prints to Social and Community Intelligence, pp. 11–18.

[57] F. Calabrese et al., “Pervasive Urban Applications,”Pervasive and Mobile Computing, vol. 9, issue 5, 2013.

[58] J. Shi et al., “PriSense: Privacy-Preserving Data Aggre-gation in People-Centric Urban Sensing Systems,” Proc.INFOCOM 2010, San Diego, CA, Mar. 2010.

[59] E. Cuervo et al., “MAUI: Making Smartphones LastLonger with Code Offload,” MobiSys ’10.

[60] J. Baliga et al., “Green Cloud Computing: BalancingEnergy in Processing, Storage, and Transport,” Proc.IEEE, Jan 2011.

BIOGRAPHIESMARCO CONTI ([email protected]) is a research directorof the Italian National Research Council. He has publishedmore than 300 research papers in journals and conferenceproceedings and four books related to design, modeling,and performance evaluation of computer networks. Hereceived several Best Paper Awards, including at IFIP TC6Networking 2011 and IEEE WoWMoM 2013. He has servedas General and Program (Co)Chair for several conferences,including Networking 2002, IEEE WoWMoM 2005 and2006, IEEE PerCom 2006 and 2010, ACM MobiHoc 2006,and IEEE MASS 2007.

Silvia Giordano ([email protected]) has a Ph.D. fromEPFL, is a professor at University of Applied Technology ofSouthern Switzerland (SUPSI), and associate researcher atCNR. She directs the Networking Laboratory. She has pub-lished extensively in the areas of QoS, traffic control, andwireless and mobile networks. She is an Editor of severaljournals, and on the steering, organizing and executive com-mittees of several international conferences. She is an ACMDistinguished Speaker and on the Board of ACM N2Women.

We can envisage

complex application

scenarios, where sev-

eral networking

paradigms are mixed

together where each

paradigm does not

act in isolation, but

there is a sharing

among them, with

mutual benefit: the

multi-paradigm era.

CONTI_LAYOUT.qxp_Layout 1/6/14 2:08 PM Page 96