RESEARCH ON AD HOC NETWORKING: CURRENT

ACTIVITY AND FUTURE DIRECTIONS

Aleksi Penttinen
Networking Laboratory, Helsinki University of Technology, P.O.Box 3000
FIN-02015 HUT, Finland

Abstract is going on within the IETFs MANET working group [3]

for standards and the research is very active throughout
This article is a survey on the current status and direction the world.
of research on ad hoc networking. We categorize the ongo- Currently the most fundamental research issue in ad hoc
ing research and outline the major challenges which have networking, between the physical layer and the applica-
to be solved before widespread deployment of the tech- tion layer, is packet routing. In fixed infrastructure mo-
nology is possible. The views presented by Perkins in [1] bile networks routing is, for the most part, an engineering
are used as a premise, which is then complemented with problem (implementation of hand-overs etc.), whereas in
discussion and references to the latest publications. ad hoc networks it is essentially theoretical. The prob-
lems and their solutions considering packet routing are
closely related to those widely studied in the case of ordi-
1 Introduction nary fixed networks, but also completely new fundamen-
tal challenges have emerged due to the peculiar features
Ad hoc networks (AHNs) are wireless multi-hop packet of AHNs, such as:
networks without any fixed infrastructure. An AHN net-
work is formed solely by its terminals so that each termi- Dynamic network topology and structure
nal connected to the network provides also relaying ser-
vice for others, i.e. acts as a router. Advantages of such Nodes may join or leave the network
system are rapid deployment, robustness, flexibility and Some or all nodes may be mobile
inherent support for mobility.
AHN can work as a stand-alone autonomous network Limited bandwidth
providing internal connections for a group. Demand for
Constrained power
such networks could arise in the contexts of shared desk-
top meeting, disaster recovery, or in various military ap- Broadcast nature of transmission
plications. However, no commercial killer applications
are known for this technology yet. In this paper, we discuss the on-going research efforts
In the future, ad hoc networks probably form the out- to tackle the problems between the present day and the vi-
ermost region of the internetwork, where a wired back- sion of practical ad hoc networking solutions. We survey
bone connects both the fixed local area networks and the the current work in progress in the field and also anticipate
mobile (both the fixed infrastructure and the ad hoc) net- some of the next steps the research is taking. The leading
works. Whereas the base stations of a fixed infrastructure theme is the intellectual challenge posed by this new tech-
networks are directly connected to the core, an AHN is nology. Therefore, we select the important issues which
typically connected through a satellite link or a terres- are to be solved in order to enable widespread ad hoc net-
trial switch (fixed wired connection point, or mobile ra- work deployment and concentrate on very general level
dio link). This vision, however, requires still some further instead of protocol specific details. In other words, algo-
developments in ad hoc networking. rithms and ideas are preferred to implementations. Em-
Basic research and potential applications of ad hoc net- phasis is on the topics that are not discussed otherwise in
works are evolving together, spurring each other into fur- this seminar.
ther achievements. The need for an application can give The new world of AHN technology has been visited
directions for the research and, on the other hand, the re- by now, there is no such thing that completely unexplored
search enables new applications to be created. Although territories in this field. Active research has produced a
this network concept has been originally considered in the wide range of proposals, but so far not many problems are
context of packet radio networks [2] earlier, it has become solved. We start from the issues recognized in [1], with a
very popular again during the past few years. The work few additions, and describe their recent development with
the latest pointers to selected literature. Subjectivity can-
not be avoided in this sort of study, so the reader is pleaded
to be gentle in his or her criticism.
The area is developing at the pace of several hundreds 1 Routing 73
of publications yearly and the speed unavoidably out- 2 MAC, scheduling 22
dates any summaries, conclusions and predictions in very 3 Special AHNs 18
short time. It is clear that ad hoc networking is in its -Bluetooth (8)
infancy what comes to its research and further develop- 4 Applications 15
ment are being expected for years to come. Hence, pre- -Multimedia (6)
vious works considering the future research in this field 5 Clustering, organization 11
have been very few in number. In addition to [1] these is- 6 Technology, physical layer 9
sues are treated in [4], and in [5] the wireless networks 7 General overviews 9
research was considered in general. From the application 8 Internet Protocols on AHNs 8
point of view, another notable instance relevant the topic 9 Network management 7
is the Wireless World Research Forum (WWRF) [6], a 10 QoS, service differentiation 7
forum founded by several leading telecom corporations. 11 New network concepts 5
Their Book of Vision attempts to recognize future re- 12 Service Availability 5
search strategies in all wireless communications. 13 Positioning, situation awareness 5
14 Topology studies 4
The organization of the rest of this paper is as follows.
15 Practical studies 2
In Section 2 we analyze publishing activity in 2001 to get
16 Transport issues 2
a picture on the key topics. Section 3 goes down to the de-
17 Security 2
tails of open research issues, starting from the ideas in [1]
18 Mobility 1
and discussing the current development in each of the ar-
19 Cooperation 1
eas as well as the questions that are still lacking any kind
of resolution. In Section 4 we briefly go through two large
Power-awareness 22
scale projects that are going on as examples of compre-
hensive studies to bring AHNs into existence before con-
cluding with some futuristic visions in Section 5. Table 1: Publications on ad hoc networks within IEEE dur-
ing 2001, see Figure 1 for illustration

2 Research status
This section attempts to summarize the current research
that is being conducted on ad hoc networks. Due to the
vast amount of the material about the topic, we chose to
take a representative sample of the most recent research
results and to categorize them to get a rough overview on
the situation. For this purpose we will look into the pub-
lishing activity of the IEEE (publications are available at 1
[7]).
2

2.1 Overview 3 19
18
17
16
15
The classification, shown in Table (1) and Figure (1), 14
13
was done for all the ad hoc networking related publica- 4 12
11
tions, conference and journal articles without distinguish- 5 10
ing them, published within the IEEE organization in 2001.
6 7 8 9

Categories were selected so that they would describe their

contents unambiguously, but still provide detailed and in-
formative knowledge on large scale (i.e the number of
classes was tried to keep as small as possible). Power- Figure 1: Publications on ad hoc networks within IEEE
aware protocols were counted twice, once into the Power- during 2001, see Table 1 for key
awareness category and once into the corresponding cat-
egory in where the protocol or algorithm was considered.
For example, a publication with the topic A power-aware
routing protocol would be counted into both Routing and
Power-awareness categories. Support for different routing protocols
Four clear facts can be immediately extracted from the
data: Interoperation with other wireless networks

Routing protocols are being studied extensively Aggregation

Overall volume of the research effort on AHNs is This survey here summarizes and complements the ap-
high (>200 publications a year just in this organi- proach presented by Perkins [1], with a few additions
zation!) and several updates. The discussion attempts to sketch the
following aspects for the topics: motivation, novel ideas
The spectrum of the topics is wide; ad hoc net- since the publication of [1], and the still remaining prob-
working has brought together protocol engineers and lems and their relative importance.
mathematicians

Based on the status of the research projects conduct- 3.1 Scalability

ing this work, further rapid development is to be ex-
Most of the visionaries depicting applications which are
pected at all fronts during the next few years.
anticipated to benefit from the ad hoc networking tech-
nology take scalability as granted. Imagine, for example,
2.2 Remarks on the categorization the vision of ubiquitous computing where the networks
can grow to thousands of nodes. How can be the swarm
Care should be taken when interpreting the categoriza-
of control messages carried out in this dynamic environ-
tion in detail. Although the major directions are visible,
ment? It is unclear how large an ad hoc network can actu-
the sample is by no means exhaustive or unbiased, spe-
ally grow.
cial issues of magazines may have highlighted some topics
Ad hoc networks suffer, by nature, from the scalability
more than deserved. Furthermore, the classification of pa-
problems in capacity. For a rough idea about this, we may
pers has some inaccuracies. Compromises were unavoid-
look into simple interference studies. In a non-cooperative
able because each paper was counted only in one category
network, where omni-directional antennas are being used,
(with the exception of power-awareness) despite the fact
the throughput per node decreases at a rate 1/ N , where
that its contents may have been suitable for another class
N is the number of nodes [8]. That is, in a network with
as well.
100 nodes, a single device gets approximately one tenth
of the theoretical data rate of the network interface card at
3 Open questions maximum. This problem, however, cannot be fixed except
by physical layer improvements, such as smart antennas.
As we saw in the previous section, ad hoc networking has If the available capacity sets some limits for commu-
been a popular field of study during the last few years. nications, so do the protocols as well. Route acquisition,
Almost every aspect of the network has been explored in service location and encryption key exchange are just ex-
some level of detail. Yet, no ultimate resolution to any of amples of tasks that will require considerable overhead,
the problems is found or, at least, agreed on. On the con- which will grow rapidly with the network size. If the
trary, more questions have arisen than been answered. scarce resources are wasted with profuse control traffic,
This section outlines the major problems remaining to it is clear that ad hoc networks will see never dawn in
be solved. The protocol dependent development possibil- practice. Scalability is an important research topic for the
ities are mostly omitted and the focus is on the big pic- future, not only because of its necessity for ad hoc net-
ture, on the problems that stand in a way of having peer- works, but also because of the applicability of same ideas
to-peer connectivity everywhere in the future. The topics in the Internet.
are: In the protocol design itself, several issues have to
be considered with the potential applications in mind.
Scalability Whereas proactive routing is not scalable in a dynamic en-
vironment as such, on-demand protocols allow deploying
Quality of Service
large networks in the expense of increased route acqui-
Client server model shift sition latency. The minimum route acquisition latency is
the product of maximum network diameter and minimum
Security node traversal time for route requests. Correspondingly,
demands for short latencies for route acquisition limit the
Interoperation with the Internet network size drastically. Can this be accepted by the ap-
Power control plications? If not, what can be done?
Traditional way of scaling the network has been hi-
Node cooperation erarchical routing, the running of routing and other net-
works functions on a several hierarchical levels. Hierar-
chies can be constructed by clustering algorithms which
collect nodes near each other into groups. While hier-
archy may not be natural for all ad hoc networks, it is
one of the very few methods capable of providing any
relief to the scalability problem. Mobility and dynamic
hierarchy, however, need to be carefully taken into ac-
count in order to achieve any practical solutions. Several
clustering solutions have already been brought forward
recently, e.g. [9, 10]. Some other approaches have pro-
posed also wide-area routing protocols [11] to comple-
ment broadcast-intensive local routing. Development of
simple rules to cluster nodes and share routing informa-
tion will remain actively researched.
Routing simulations discussed in the literature have
been very small in comparison of the futuristic idea of Figure 2: Services can be difficult to locate in ad hoc net-
ubiquitous computing or sensor dust of tens of thousands works
of nodes. Therefore, large scale simulation studies and
also trial deployment are essential to study in the future.
Future research will probably develop scenarios, isolate
demand protocols can be configured to return only com-
useful applications, to optimize the trade-off between ca- munications paths that comply with the desired parame-
pacity and scalability in each case separately. For a gen- ters.
eral solution much remains to be done. Without develop-
QoS in AHNs is still largely unexplored area, a good in-
ment in intelligent antennas and multiuser detection, scal-
troduction to the field of study can be found from [12]. Is-
ing AHNs to thousands of nodes seems a daunting task
sues of QoS robustness, QoS routing policies, algorithms
indeed.
and protocols with multiple, including preemptive, priori-
ties are to be researched in the future. It seems that in ad
3.2 Quality of Service hoc networks end user may have to haggle with the net-
work on the QoS parameters as high quality is frequently
The heterogeneity of existing applications in the Internet not available.
have challenged the network which is able to provide only Quality of service cannot be guaranteed for a long time
best-effort service. Voice, live video and file transfer, to because of the link quality variations due to the inter-
mention the common examples, all have very differing re- ferences etc. Methods to detect and report changes in
quirements what comes to delay, jitter, bandwidth, packet the connection quality should be investigated in the fu-
loss probability etc. Quality of Service (QoS) is being de- ture. Perkins suggest an addition of a new ICMP message
veloped to meet the emerging requirements. QoS is a guar- (QOS LOST) to be defined to inform the endpoints that a
antee by the network to provide certain performance for new route discovery should be initiated.
a flow in terms of the quantities mentioned above. QoS
routing attempts to locate routes that satisfy given perfor-
mance constraints and then reserve enough capacity for 3.3 Client-server model shift and service lo-
the flow. cation
If we consider ad hoc networks as a natural extension
of the Internet at the borders where direct connections In the Internet, a network client is typically configured to
to fixed infrastructure are unavailable, these applications use a server as its partner for network transactions. These
(voice, live video, file transfer etc.) will exist also in ad servers can be found automatically or by static configu-
hoc networks and so there will also be naturally a demand ration. In ad hoc networks, however, the network struc-
for QoS. However, the lack of fixed infrastructure in ad ture cannot be defined by collecting IP-addresses into sub-
hoc networks makes the QoS appear even more challeng- nets. There may not be servers, but the demand for basic
ing problem than ever before. Bandwidth is seriously lim- services still exists. Address allocation, name resolution,
ited, routes are using links with differing quality and sta- authentication and the service location itself are just ex-
bility. Links are often asymmetrical so that, for example, amples of the very basic services which are needed but
QoS for telephony (2-way traffic with QoS-demands) may their location in the network is unknown and possibly even
not be achieved by just one route. These facts are often changing over time. Where do services reside (see Figure
completely ignored by the routing protocols, although the (2)? Who is administering or maintaining the these ser-
link-state algorithms could be used to find suitable routes vices?
if the links are given suitable QoS costs. Alternatively, on- In ad hoc some recent proposals have considered in-
tegrating route discovery and service location tasks by In [16] the authors discussed the security problems in
allowing only particular kind of services to react to the general and proposed a self-organized public-key infras-
broadcast requests. This approach, however, can be seen tructure for ad hoc network cryptography. Key exchang-
to have the following deficiencies: ing, however, raise again the scalability issues. Further-
more, defining keys for multicast transmission seems even
Inserting application service discovery into a net- tougher challenge.
work layer protocol violates the modular protocol de- Secure routing was considered in [17], which had an
sign. appealing idea of dividing the data on N pieces which are
send along separate routes and, at the destination, the orig-
The client may not be able to specify the required
inal message is reconstructed out of any (M out of
service in a way that the request can be carried on
N ) pieces of the message.
the network layer.
Security is indeed one of the most difficult problems to
Authorization can be difficult at the network layer. be solved, but it has received only modest attention so far.
The golden age of this research field can be expected to
Other possibilities are, e.g., using well-known multicast dawn only after the functional problems on the underlying
addresses for very basic features, such as DNS. Also pro- layers have been agreed on.
tocols for service location have been proposed. Some re-
cent works on this field include [13, 14].
3.5 Interoperation with the Internet
An intellectual challenge related to the service avail-
ability problems is the design of distributed network func- It seems very likely that one of the most common applica-
tions. It could be investigated whether and which services tions of ad hoc networks require a connection to the Inter-
(or their locations) could be shared or circulated among net. By ad hoc network technology the coverage of wire-
nodes? Still the question of who is administering and ulti- less LAN systems can be expanded and complemented.
mately responsible for the services remain unanswered. However, the issue of defining the interface between the
two very different networks is not straightforward.
If a node in ad hoc network has an Internet connection,
3.4 Security
it could offer Internet connectivity to the other nodes. The
Ad hoc networks are particularly prone to malicious be- node could defined itself as a default router and the whole
havior. Lack of any centralized network management or ad hoc network could be considered to be single-hop
certification authority makes these dynamically changing from the Internet perspective although the connections are
wireless structures very vulnerable to infiltration, eaves- physically over several hop links. Recently a practical so-
dropping, interference etc. Security is often considered to lution for this problem was suggested in [18]. The idea
be the major roadblock in commercial application of ad was to combine the Mobile-IP technology with ad hoc
hoc network technology [15]. routing [18] so that the gateway node can be considered
Security requirements depend naturally on the applica- to be foreign agent for Mobile IP.
tion where they are needed. In cases where all the termi-
nals are on the same side, such as military or emer- 3.6 Power control
gency rescue applications, it is enough to get protection
against outside interference. In civilian, especially com- Power-aware networks are currently being extremely pop-
mercial, applications even mere lack of cooperation may ular within the ad hoc networking research. The motiva-
be enough to bring the network on its knees. The nodes tion for power-aware thinking for wireless communica-
enter and leave the networks as they wish and links may tions is obvious, as summarized in [19]:
be using nodes that should not have access to data. How
to define membership in ad hoc networks, how to classify Functional utility New features and functionality
nodes to the trusted and the not-trusted ones? usually costs additional energy. By increasing energy
Traditional methods of protecting the data with cryp- efficiency, devices may meet new user demands with-
tographic methods face a challenging task of key distribu- out reduced useful lifetime.
tion and refresh. Accordingly, the research efforts on secu-
Size and weight Increased power efficiency can al-
rity have mostly concentrated on secure data forwarding.
low smaller and lighter power source.
However, many security risks are related to the peculiar
features of ad hoc networks. The most serious problem Maintenance Power sources will always need to be
is probably the risk of a node being captured and com- replaced or recharged at some point, and the cost for
promised. This node would then have access to structural this can vary from inconvenient to prohibitive.
information on the network, relayed data, but it can also
send false routing information which would paralyze the Environmental Battery designs contain acids and
entire network very quickly. heavy metals, which must be disposed of properly.
 There are two research topics which are partially sim- Protocol Layers
ilar: the maximization of lifetime of a single battery and - retransmission policies
the maximization of the lifetime of the whole network. Transport layer - congestion control
The former is related to commercial applications and node
cooperation issues whereas the latter is especially of mil- - avoiding transmission
itary etc. interest, where the node cooperation is already - avoiding weak nodes
Network layer
assumed. - sleep patterns; reduced topology
The goals can be achieved either by developing bet-
- avoiding overhearing
ter batteries, or by making the network terminals oper- MAC layer
ation more power efficient. The first approach is likely to
give a 40% increase in battery life in near future (with Li-
Polymer batteries) [19]. As to the device power consump-
tion, the primary aspect are achieving energy savings is Tx Receive Idle Sleep
through the low power hardware development using tech-
niques such as variable clock speed CPUs, flash memory, Network Interface State
and disk spindown [20]. However, from the networking
point of view our interest naturally focuses on the devices Figure 3: Energy saving methods on protocol layers and
network interface, which is often the single largest con- between transmitter states
sumer of power.
Energy efficiency at the network interface can be im- On transport level, collisions and retransmissions
proved by developing transmission/reception technologies should be avoided at all cost. Traditional TCP is badly
on the physical layer and by sensing inactivity on the ap- suited for wireless interfaces as it cannot separate packet
plication layer, but especially with certain networking al- losses due to congestion and due to transmission errors,
gorithms; MAC, routing and handling of end-to-end con- which can be common. An experimental transport proto-
nections. In all these approaches, savings are based on in- col for wireless networks is the Wave and Wait protocol
telligently turning off the interface when it is not needed. [26].
Medium Access Control - protocols can be made Radio interface of a network device can have four dif-
power-aware by simple rules: when the node has nothing ferent states (while powered on): transmit, receive, idle,
to send or receive, or it overhears a transmission (i.e. the and sleep. The energy consumption of these is highest for
radio channel is busy) it can power off the network inter- transmission and lowest for sleep mode. Figure (3) shows
face and wake again after a while to see is there anything a summary of the energy saving methods as well as the
to do. This has significant advantages as receiving unnec- transmitter states between which the corresponding sav-
essary data is surprisingly expensive in terms of energy ings are obtained.
consumption. One such protocol, PAMAS, is introduced Power-aware techniques will remain an important re-
in [21], for which the authors reported up to 70% energy search area in near future on all aspects described above.
savings. How these can be combined with the conflicting QoS de-
Just above the MAC-layer reside different topology re- mands? How can we make sure that a node cannot give
duction algorithms. Their premise is that if the network false information on its power level just to avoid relaying?
is dense enough, only a subset of nodes is required to be Heuristic rerouting to maximize the network lifetime?
relaying nodes to maintain full connectivity. This means
that some of the nodes can be put to a sleep state (such
as provided by IEEE 802.11 [22]) only to wake up peri-
3.7 Node cooperation
odically to see whether there are incoming traffic directly Closely related to the security issues, the node cooperation
to them. Active nodes form a forwarding backbone in the stands in the way of commercial application of the tech-
network, which can be found distributedly as discussed in nology. The fundamental question is: Why should anyone
[23]. This problem is closely related to (minimum) domi- relay others data? The answer is simple: to receive the
nating set problem in graph theory. corresponding service from the others. However, when
In routing, one usually tries to maximize the network differences in amount and priority of the data are exist-
lifetime. In other words, routes are selected by their trans- ing, the situation is more complex. Surely, a critical fire
mission energy cost giving the priority to the nodes with alarm box should not waste its batteries for relaying gam-
full batteries. This way the time to network partition can ing data, nor should it denied access because of this be-
be maximized distributedly [24]. Furthermore, unicast and havior.
multicast routing should be considered separately when Encouraging nodes to cooperate may lead to the intro-
considering energy-efficiency due to the broadcast nature duction of billing in ad hoc networks, with a similar idea
of the transmission [25]. which was suggested for Internet congestion control in
[27]. Well-behaving network members could be rewarded A similar problem arises when a device is powered on
for the relaying and selfish or malicious users could be at a border of several networks and it has to choose which
charged. Implementation of any kind of billing mecha- one to join.
nism, however, is very challenging. These issues are still
wide open. 3.9.2 Anytime, anywhere
These questions have been recently introduced in e.g.
[28], where a simple mechanism based on a counter in One of the most important aims of the research on all
each node is studied. wireless networks is to provide seamless integration of all
types of networks. This issue raises questions how can the
ad hoc networks designed so that they are compatible with
3.8 Support for different routing protocols e.g. WLAN or 3rd Generation cellular networks? Could
the other networks extended from last-hop to multi-hop
If energy costs or other dynamic quantities are to be wireless connections using these techniques?
tracked, there may be significant differences between rout-
ing algorithm performances. For certain sensor networks,
static node-state based algorithms enable route optimiza-
3.10 Aggregation
tion especially in multicast case, whereas such algorithm Finally, there is the question of rationalizing and collect-
would be too cumbersome for networks with mobility. ing the research results. Research has been extremely ac-
Dozens of routing protocols have bee introduced, all tive during the past few years. The pace has been so fast
of which typically perform well in some situations while that the big picture is somewhat blurred. That is why there
having significant weaknesses in other cases. Question is, is a need for summarizing research efforts to combine,
can the heterogeneity of the ad hoc networks be covered not just compare, different approaches. The trend is to-
by any single routing algorithm? wards more complete ad hoc networking solutions instead
If no all-round routing protocol can be found or agreed of specific protocols in the near future. The first works on
on, the networks have to be capable of supporting several this field has been conducted for energy conserving pur-
protocols. This can happen so that as a certain number poses because of its inherent multilayer-structure that
of network nodes detect that their routing is not optimal provides a natural environment for combining different
the network switches to another protocol. When a proto- ideas.
col ceases to be optimal and how can the change be im- There is work to be done to find best possible combina-
plemented? tions of MAC, topology reduction, and routing protocols.
Another option is that the network is able to simultane- There is also work to be done in combining preferable
ously support several routing methods. How to define the properties of different protocols. This will naturally lead
interfaces and self-awareness? to discussion on specific networks, application tailored so-
lutions, as the ultimate ad hoc networking solution is still
far away, if it even can be found.
3.9 Interoperation with other wireless net-
works
4 Major ongoing efforts
3.9.1 Ad hoc networks
In this section we discuss two very different but compre-
The self-organization of ad hoc networks is a challenge hensive examples of recent ongoing long term research
when two independently emerged networks collide. This projects aiming at producing functional ad hoc networks.
is an unexplored research topic that has implications on
all levels on the system design.
What happens when two autonomous ad hoc networks
4.1 The Terminodes project
move into same area? Surely they are unable to avoid in- The Terminodes project [29] is a 10-year joint effort of
terfering each other. Ideally, the networks would recog- seven Swiss research institutes started in 2000. Their aim
nize the situation and be merged. However, the issue of is to study and prototype a large-scale ad hoc network
joining two networks is not trivial; the networks may be with the emphasis on the self-organization feature. The
using different synchronization, or even medium access or network has a two-layer routing system which is based
routing protocols. Security becomes also a major concern on location information at long ranges while routing tra-
in these cases. Can the networks adapt to the situation? ditionally over few hops. Other issues considered so far
A common example; a military unit moving into an area in this project have been mobility management, GPS-free
covered by a sensor network could be such a situation; positioning, cooperation and security.
moving unit would probably be using different routing The project had goals also on higher level of abstraction
protocol with location information support, while the sen- than just the network design. It is meant to be an intellec-
sor network would have a simple static routing protocol. tual stimulus to find new research challenges, e.g. how
 public authority ad hoc networks, which can be used for
a very specific purposes. For the most part the networks
will be quite small except for the sensor applications, in-
cluding the millimeter sized sensing networks.
In the second phase, the future cellular infrastructure
extensions could be implemented so that ad hoc network-
ing would act as a basis of the whole 4th generation
wireless technology. An image of completely unrestricted
anytime, anywhere communications using this technol-
ogy seems, however, to belong to the more distant fu-
ture. Problems with security, authorization and manage-
ment are daunting indeed in large scale networks. Hence,
it is more likely that the technology will be used to aug-
ment wireless LAN technology with the limited network
size or hops in connections.
Figure 4: Small sensors using ad hoc radio [30]

do we define a formal model for fair exchange? [11]. Ul- 5.2 The Revolution?
timately the project is a societal vision in which this kind
of communication is considered. Assume that most of the problems discussed in this pa-
per are solved and there is a possibility to deploy secure
4.2 Smart Dust broadband self-organizing ad hoc networks with hundreds
or thousands of nodes. What will happen?
Maybe the most concrete project related to ad hoc net- The general trend is towards low-level infrastructure
works is DARPA-funded Smart Dust [30] at the Uni- and increased end-user responsibility. Being able to freely
versity of California, Berkeley. Smart Dust goals are to communicate transfer information with close-by people
design, build and test networks consisting of small sen- is, of course, convenient, but it may even have larger soci-
sor nodes (see Figure (4)). The applications include vari- etal effects.
ous military purposes, such as battlefield sensor networks,
traffic mapping, sensor mine fields, etc. In civilian use On a larger scale, ad hoc networking can very well be
these devices can be used e.g. building sensors in earth- the next revolution in the world of communications. It en-
quake zones. The sensors communicate with RF or opti- ables local communities to manages their own need for
cal transmitters and the communications issues (which are connectivity using their own local resources. The control
not the main theme in the project) raise some interesting of the local network will be hence again where it belongs
questions, see [31]. and the ISP era might be coming to an turning point. Why
to a call a taxi if you are visiting your next door neigh-
bor or the shop around the corner? The telecommunica-
5 Future view tions business would experience shift from the operators
towards both device manufacturers and end-users them-
It is sometimes useful to try to predict the future to get selves. Are the operators going to allow this development?
new ideas and see the present day in a more appropriate Can it be done without their support?
context on larger scale. Future is unknown, but it is, after In this local community networking view can be the
all, the result of the actions we take now. In this Section we seed of a completely new approach to communicating
look into the crystal ball and give scenarios on the future with people and henceforth to understanding what it
development. How do ad hoc networks evolve? What are means to be a part of a community. In these local networks
the enabling technologies? What kind of applications we many services, such as local web pages, e-mail and tele-
are going to see in the near future? phony, would be free from charge and jurisdiction by re-
mote administrations. Naturally, there would be some pri-
vacy concerns in the beginning, but trust inside the com-
5.1 Going Ad Hoc munity should provide the necessary umbrella for the pri-
There are many open questions related to ad hoc networks vacy.
applications. Before a public demand for any set of ap- Ad hoc networks have indeed the potential to change
plications can be found, these networks will be deployed how we see the communications world today. For alter-
in various specialized cases. In the first phase, which can native scenarios of the wireless future, where the services
be already foreseen we will have autonomous military and stay centralized, interested reader should consult e.g. [32].
5.3 Conclusion Le Boudec. Self-organization in mobile ad-hoc net-
works: the approach of terminodes. IEEE Commu-
Whereas ad hoc networks will become widely used in mil- nications Magazine, pages 166174, June 2001.
itary contexts in near future, the corporate world has to
continue the daunting search for profitable commercial [12] Satyabrata Chakrabarti and Amitabh Mishra. Qos
applications and possibilities of the technology. Mean- issues in ad hoc wireless networks. IEEE Communi-
while, the academic community has adopted the new field cations Magazine, 39(2):142148, February 2001.
as a playground to apply their ideas to create something
[13] L.M. Feeney, B. Ahlgren, and A. Westerlund. Spon-
completely new. In all, although the widespread deploy-
taneous networking: an application oriented ap-
ment of ad hoc networks is still years away, the research
proach to ad hoc networking. IEEE Communications
in this field will continue being very active and imagina-
Magazine, 39(6):176181, June 2001.
tive.
[14] H. Koubaa and E. Fleury. A fully distributed me-
diator based service location protocol in ad hoc
References networks. In Proc. of Global Telecommunications
Conference (GLOBECOM), 2001, volume 5, pages
[1] Charles E. Perkins (Ed.). Ad Hoc Networking. 29492953. IEEE, 2001.
Addison-Wesley, December 2000.
[15] Zygmunt J. Haas, Jing Deng, Ben Liang, Panagio-
[2] John Jubin and Janet D. Tornow. The DARPA packet tis Papadimitatos, and S. Sajama. Wireless ad hoc
radio network protocol. Proc. of IEEE, 75(1):2132, networks. In John Proakis, editor, Encyclopedia of
1987. Telecommunications (to appear). John Wiley, De-
cember 2002.
[3] IETF. Mobile ad hoc networks working
group. http://www.ietf.org/html.charters/manet- [16] Jean-Pierre Hubaux, Levente Buttyan, and Srdjan
charter.html. Capkun. The quest for security in mobile ad hoc
networks. Proc. of MobiHOC 2001, pages 146155,
[4] C.-K. Toh. Ad Hoc Mobile Wireless Networks: Pro- October 2001.
tocols and Systems. Prentice Hall, December 2001.
[17] P. Papadimitratos and Z.J. Haas. Secure routing
[5] Magnus Frodigh, Stefan Parkvall, Christian Roobol, for mobile ad hoc networks. In SCS Communi-
Per Johansson, and Peter Larsson. Future-generation cation Networks and Distributed Systems Modeling
wireless networks. IEEE Personal Communications, and Simulation Conference (CNDS 2002), San An-
pages 1017, October 2001. tonio, TX, January 2002.

[6] Wireless World Research Forum. The book [18] Yuan Sun, Elizabeth M. Belding-Royer, and
of visions 2001. http://www.wireless-world- Charles E. Perkins. Internet connectivity for ad hoc
research.org/, December 2001. mobile networks. International Journal of Wireless
Information Networks, special issue on Mobile Ad
[7] IEEE. IEEE/IEE electronic library. hoc Networks (to appear), 2002.
http://ieeexplore.ieee.org. [19] Chiara Petrioli, Ramesh R. Rao, and Jason Redi.
[8] Piyush Gupta and P. R. Kumar. The capacity of wire- Guest editorial: Energy conserving protocols. ACM
less networks. IEEE Transactions on Information Mobile Networks and Applications, 6(3):207209,
Theory, 46(2):388404, March 2000. June 2001.
[20] Christine E. Jones, Krishna M. Sivalingam, Prathima
[9] S. Basagni. Distributed clustering for ad hoc net-
Agrawal, and Jyh Cheng Chen. A survey of energy
works. In Proceedings of the 1999 International efficient network protocols for wireless networks.
Symposium on Parallel Architectures, Algorithms, Wireless Networks, 7(4):343358, September 2001.
and Networks (I-SPAN99), pages 310315, June
1999. [21] Suresh Singh and C. Raghavendra. Pamas: Power
aware multi-access protocol with signalling for ad
[10] P. Basu, N. Khan, and T.D.C. Little. A mobility hoc networks. ACM SIGCOMM Computer Commu-
based metric for clustering in mobile ad hoc net- nication Review, 28(3):526, July 1998.
works. In Proceedings of Distributed Computing
Systems Workshop 2001, pages 413418, 2001. [22] IEEE Computer Society LAN MAN Standards
Committee. Wireless LAN medium access control
[11] Ljubica Blazevic, Levente Buttyan, Srdjan Capkun, and physical layer specifications, IEEE 802.11 stan-
Silvia Giordano, Jean-Pierre Hubaux, and Jean-Yves dard, August 1999.
[23] Benjie Chen, Kyle Jamieson, Hari Balakrishnan, and
Robert Morris. Span: an energy-efficient coordina-
tion algorithm for topology maintenance in ad hoc
wireless networks. ACM Wireless Networks Journal
(to appear), 8(5), September 2002.
[24] Jae-Hwan Chang and Leandros Tassiulas. Energy
conserving routing in wireless ad-hoc networks. In
INFOCOM, volume 1, pages 2231, 2000.

[25] J.E. Wieselthier, G.D. Nguyen, and A. Ephremides.

Algorithms for energy-efficient multicasting in ad
hoc wireless networks. In Military Communica-
tions Conference Proceedings, 1999. IEEE MIL-
COM 1999., volume 2, pages 14141418, 1999.
[26] V. Tsaoussidis, A. Lahanas, and H. Badr. Wave &
wait protocol (wwp): low energy, high throughput
for mobile ip-devices. In Proc. of ICON 2000, 2000.
[27] Jeffrey K. MacKie-Mason and Hal R. Varian. Pric-
ing the internet. In B. Kahin and J. Keller, editors,
Public Access to the Internet. Prentice-Hall, New
Jersey, 1994.
[28] L. Buttyan and J. P. Hubaux. Stimulating co-
operation in self-organizing mobile ad hoc net-
works. Technical Report DSC/2001/046, EPFL-DI-
ICA, August 2001.
[29] http://www.terminodes.org.
[30] http://robotics.eecs.berkeley.edu/ pister/SmartDust/.
[31] J. Kahn, R.H. Katz, and K. Pister. Emerging chal-
lenges: Mobile networking for smart dust. J. Comm.
Networks, pages 188196, September 2000.
[32] Aurelian Bria, Fredrik Gessler, Olav Queseth,
Rickard Stridh, Matthias Unbehaun, Jiang Wu, and
Jens Zander. 4th-Generation wireless infrastruc-
tures: Scenarios and research challenges. IEEE
Personal Communications, pages 2531, December
2001.