Вы находитесь на странице: 1из 18

Routing in WSNs

Tanmay Bhola
June 4, 2010

1
Abstract
Wireless Sensor Networks differ from other networks like IEEE 802.11
WLANs in their requirements, functionality and constraints. These networks
mostly comprise of small, battery powered nodes that perform the operation of
sensing and sending the information to a gateway node in an ad-hoc fashion. It
has been found that use of smart routing techniques tailored to the demands of
these networks can result in huge system-wide gains. This report highlights the
key challenges posed by the inherent properties of WSNs on routing protocols
and some of the state-of-the art protocols that have been designed to overcome
these.

2
Contents
1 Introduction 4

2 Fundamentals 5
2.1 Routing Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Routing Challenges in WSNs . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.3 Classification of routing protocols . . . . . . . . . . . . . . . . . . . . . . 6

3 Protocols 8
3.1 Optimized Link State Routing Protocol (OLSR) . . . . . . . . . . . . . . 8
3.1.1 Optimisations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.1.2 Protocol Operation . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.1.3 Analysis for use in WSNs . . . . . . . . . . . . . . . . . . . . . . 10
3.2 Ad-Hoc on Demand Distance Vector (AODV) . . . . . . . . . . . . . . . 11
3.2.1 Path Discovery Process . . . . . . . . . . . . . . . . . . . . . . . . 11
3.2.2 Forward and Reverse Path Setup . . . . . . . . . . . . . . . . . . 12
3.2.3 Routing Table maintenance . . . . . . . . . . . . . . . . . . . . . 12
3.3 Collection Tree Protocol (CTP) . . . . . . . . . . . . . . . . . . . . . . . 13
3.3.1 The Concept of Collection Protocols . . . . . . . . . . . . . . . . 13
3.3.2 Key Features of CTP . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.3.3 CTP Noe : Improvements over Basic CTP . . . . . . . . . . . . . 15

4 Discussion and Conclusion 16

3
1 Introduction
Wireless sensor networks (WSNs) comprise of small nodes with a radio, microcontroller
and sensors. The operation of these networks is based on these nodes communicating
over a wireless link in an ad-hoc fashion. This builds a platform for numerous appli-
cations of these networks including weather monitoring, precision agriculture, intrusion
detection, disaster management and tactical surveillance.

However, due to a limited transmission range of the wireless unit and in order to
span a large application area, these nodes use multi-hop transmission of data to the
base-station. Hence individual nodes have to play the role of a router as well as end
station. These networks also suffer from energy, memory and resource constraints. Var-
ious techniques have evolved over time to achieve data delivery in these networks under
the above listed constraints. These specialised techniques, often combined with some
general well-known routing concepts have lead to development of state of the art routing
protocols for wireless sensor networks.

This report expects the reader to have the basic knowledge about WSNs and their
working. It details the various challenges posed by WSNs on protocol designers in Sec. 2.
and moves on to listing the classification of routing protocols using different schemes.
It then describes major protocols that have emerged under the various classes, their
underlying concepts and benefits in Sec. 3.

4
2 Fundamentals
This section covers basic concepts in routing protocols and explains how these can be
extended to meet the demands of WSNs. The various requirements and challenges of
WSN routing have been discussed here to underline the need for specialised routing
techniques for these networks. It then goes on to classify different routing techniques
for WSNs.

2.1 Routing Basics


This section provides a very basic refresher to some of the routing concepts that would
be required for the following sections. More exhaustive coverage of routing fundamen-
tals can be found in [15] and [12]. Routing is the act of moving information around in
a network from a source to a destination. This might involve passing the data through
several intermediate nodes. Routing includes mechanisms for two basic activities: de-
termining optimal routing paths and transporting all the data arranged in packets. The
first step involves coming up with metrics that assist in evaluating a potential link for
data transfer and an algorithm that correctly computes the path based on those met-
rics. Hence route information has to be correctly obtained and used when required.
There are a variety of routing algorithms used in networking and these vary depending
on the goals of the designer and application. Some of these goals could be optimality,
simplicity and minimal overhead, robustness, stability, speed and flexibility. Routing
algorithm designers weigh the relative importance of these features for a given network.
The process of packetisation involves coming up with the optimal data structure and is
based on a trade-off between the information sent and desired network load (protocol
overhead).

2.2 Routing Challenges in WSNs


Routing in WSNs can be challenging due to the inherent characteristics of these networks
that differentiate them from other wireless networks [1]. A few of these challanges have
been listed below.

(i) Energy Consumption without losing accuracy: The battery powered oper-
ation of sensor nodes puts a direct constraint on the design of routing protocols for
WSNs. Protocols have to aid in the energy-efficient operation of these networks.

(ii) Data Reporting Method: In WSNs, data delivery can be event-driven, time-
driven, query-driven or hybrid [1]. The time-driven method is suitable for weather
monitoring applications whereas even-driven approach applies to security and
surveillance applications. Routing protocols are hence highly influenced by the
data reporting method.

5
(iii) Node Heterogeneity: Depending on the application, the nodes can have different
roles and capabilities (in terms of computation, communication and power) in a
network. This should be given some consideration when designing WSN routing
protocols.

(iv) Fault Tolerance : Sensor nodes might fail due to lack of power, damage, in-
terference and hardware failure. In such cases, the MAC and Routing protocols
must aid in formation of new routes so that information can still be relayed to the
gateway node . The low price hardware used in the sensor nodes to keep system
costs down increases the responsibility on routing protocols for achieving this.

(v) Data Aggregation: In dense WSNs, many sensor nodes might be generating
packets containing redundant data, due to their proximity. Using a lower number
of sensor nodes is not the optimal way to solve this problem, as nodes can often fail,
due to reasons listed above, leading to missing results at the base station. However,
the amount of traffic flowing in the network can be reduced while maintaining a
optimal level of resolution by data aggregation. This involves the data from nearby
sensors being combined according to some aggregation function before being sent
to the base station. The aggregation function can make a decision on the final
value to be sent from a set of values based on a function like average, minima,
maxima or duplicate supression.

(vi) Scalability: WSNs can have a large number of nodes spreading over a vast area.
More nodes can also be added dynamically after the system is already in operation.
This means that the routing protocols need to be able to handle changes in topology
and growth of the network. They should be able to work with tens to thousands
of sensor nodes.

2.3 Classification of routing protocols


Due to the huge variety of protocols and their features, there is no unique or standard
classification scheme followed for all routing protocols. However, two ways of categoris-
ing routing protocols have been listed below.

(1) Based on time of route computation: Routing protocols can be classified based
on time the routes are computed. This scheme is common in Mobile Ad-Hoc net-
works.
– Proactive (Table-Based)
In a proactive protocol, every node maintains one or more tables representing the
topology of the network. The tables are updated at regular intervals to make sure up-
to-date information is present in the system at all times. Whenever a packet needs
to be sent, this information is used by the node. Optimised Link State Routing

6
Protocol (OLSR) is a popular Ad-Hoc proactive protocol. This protocol has been
discussed in detail in Sec. 3
– Reactive (On-Demand)
Reactive protocols do not demand the route-discovery process to be started until a
route to the destination is required by a packet. This helps in reducing the overall
network overhead at the cost of routing delay, as routes are not stored in memory
all the time. Ad-Hoc on-demand Distance Vector Routing protocol (AODV) is a
new protocol that uses reactive routing techniques for ad-hoc networks. This has
been discussed further in Sec. 3

(2) Based on Network Structure:The routing protocols in WSNs need to address


various demands of the network they are being used in. This is needed to meet
the energy requirements of the system. Based on the network structure, routing
protocols can be divided into 3 major categories [1].
– Flat
In a flat network, all nodes essentially play the same role. This means that the
nodes have similar features in terms of hardware capabilities and duties (sensing
and routing) in the network. Heinzelman et al. came up with a number of adaptive
protocols under this category called SPIN (Sensor Protocols for Information via
Negotiation)[6].
– Hierarchical (Cluster-Based)
This set of protocols has emerged to address the data aggregation and scalability
requirements of WSNs. In a hierarchical setup, higher-energy nodes can be used to
perform special tasks like data aggregation, processing and sending data to the base
station, whereas low-energy nodes can be used simply to sense data near the target
and send it only to the cluster head. Such protocols help in reducing energy-wastage,
redundant data transmission and improve overall system scalability. LEACH Pro-
tocol suggested in [5] leads to overall network lifetime gains using clustering.
– Location-Based
In these protocols, nodes are addressed and used for routing based on their location
in the network. This location can be estimated using received signal strengths or a
low power GPS receiver. This helps in formation of routes and improves efficiency,
as only those nodes need to be considered that are on the route to the base station
from the point of detection. GPSR (Greedy Perimeter Stateless Routing for Wireless
Networks) [8] suggests a geographic location based routing scheme.

7
3 Protocols
This section describes some of the routing protocols that have come out of research
in this area. It details the problems they address, their key features and most suited
applications.

3.1 Optimized Link State Routing Protocol (OLSR)


OLSR Protocol suggested in [7] provides an optimisation over Link State routing pro-
tocols (like OSPF [9]) to cater to needs of ad-hoc networks. This is an example of a
pro-active protocol, as it is based on periodic exchange of control messages to aid in
maintaining routing tables. Due to this, OLSR is able to readily provide routes to a
destination when required, at the cost of bandwidth used for sending topology updates
periodically. OLSR tries to reduce this cost using an optimisation mechanism. The fol-
lowing sections detail the optimisation schemes used, functioning of the OLSR protocol
and its advantages and disadvantages.

3.1.1 Optimisations
In link-state protocols, the main bandwidth inefficiency comes from the control overhead
used to maintain and constantly update the routing table. All the links with neighbour-
ing nodes are listed and flooded in the entire network to help create a dynamic topology
map.OLSR tries to reduce this wastage by using the concept of Multipoint Relays, which
has been explained below.

– Multipoint Relays (MPRs) aim to minimize the flooding of broadcast packets


in the network by reducing duplicate transmissions in the same region. For each node,
its MPRs are the nodes in the network that retransmit its packets further. The other
neighbours that are not in this set, simply read and process the packets but do not for-
ward them to other nodes. This mechanism is shown in Fig. 1. For each node, its MPRs
are the one-hop neighbours that together provide routes to all two-hop neighbours. It is
important to note that the protocol does start up with the entire set of one-hop neigh-
bours selected as MPRs and tries to iteratively remove neighbours with redundant links
in order to optimise the system as more information is collected regarding the links.
– MPR Selector Set for a node contains the neighbouring nodes that have chosen
it as their MPR. This is maintained to ensure data flow using MPRs as intermediate
nodes. This set can keep changing with time and uses constant exchange of messages
(TC messages, discussed below) in order to ensure smooth operation of the protocol.

There are 2 main improvements added through this protocol.


(i) It tries to reduce the amount of control data being sent around the network. In-
stead of declaring all links in the messages, it only sends information regarding its

8
Figure 1: Comparing flooding to use of MPRs (right). The node in orange (in the
centre) is shown after having selected a set of MPRs in blue that collectively provide
connectivity to all other nodes in the networks.

Multipoint relay selectors.

(ii) It tries to minimise the flooding of the data being sent by using only MPRs to
diffuse the messages in the network.

3.1.2 Protocol Operation


OLSR relies on MPRs to find routes to different network destinations. The protocol
uses 2 types of messages, called HELLO and TC (Topology Control) in order to use
MPRs for routing. The underlying operation of the protocol is however quite similar to
that of a link state protocol like OSPF [9].

• Neighbour Sensing using HELLO Messages : Each node has to learn about
its neighbours in order to construct routes. For this purpose, each node broadcasts
HELLO messages that contain information about its neighbours and their link
status. These are heard by all one-hop neighbours but are not relayed any further.
It verifies that the link is bi-directinal before placing the entry in a Neighbour
table maintained by that node. This information helps a node in selecting its
MPRs. Various algorithms for selecting MPRs have been presented in [13]. These
HELLO messages also contain a field called Link Code, which can be set to ’MPR’
in order to notify the neighbours that have been selected as MPRs. Using this
information, those nodes build up the MPR Selector Set (containing nodes that
have selected it as their MPR).

• Topology Information passing using TC messages: Topology control mes-


sages are broadcasted by nodes elected as MPRs and contain the information
regarding their MPR Selector Set. There is a sequence number associated with
each MPR selector set that helps in keeping this information up to date, in a table
called the Topology Table. The greater the sequence number, the more current

9
is the topology information contained in the message. Topology control messages
may be sent earlier than the scheduled time when a change has occured in the
MPR selector set.

• Route Calculation using Neighbour and Topology Tables: The tables


maintained in the nodes using mechanisms explained above are used to calculate
optimal routes to all destinations in the network. These routes are then stored
in the node’s Routing table. This table is built from the available information by
tracing the connected pairs in reverse order (from destination to the source). It
contains destination address, next hop and distance fields. Changes in the neigh-
bour and topology tables lead to changes in the routing table. For example, if a
neighbour node goes down or updated TC information with higher sequence num-
ber is received, routes need to be updated. The detailed procedure for calculating
the routing table has been shown in [7] which is not explained here to maintain
the level of complexity.

3.1.3 Analysis for use in WSNs


This protocol has various advantages and disadvantes over other routing protocols.Choice
of this protocol for a WSN would require analysis of the application and its requirements.
Overall, it is more suitable for dense deployment of nodes, which donot suffer from ma-
jor hardware constraints and require routes to different destinations quite frequently.
Its advantages and disadvantages have been stated below.

Advantages

• Offers low latency as paths are ready to be used when needed. This can be highly
useful in applications where information is useful only if delivered as soon as the
event takes place.

• Highly useful in dense and large networks. Node density can be quite high in
WSNs, and use of Multipoint relays would be useful in reducing bandwidth inef-
ficiencies of simple link state routing protocols.

Disadvantages

• Inefficient in sparse networks, where it switches to simple Link State protocol.

• Higher control overhead as compared to on-demand routing protocols

• Higher computation and storage required by this protocol might not be suitable
for simple, cheap WSN nodes.

10
3.2 Ad-Hoc on Demand Distance Vector (AODV)
AODV protocol suggested in [11] was designed keeping in mind the goals of minimis-
ing broadcasts of route advertisements (commonly found in pro-active protocols) and
transmission latency when using new routes. There are 3 main packets used by AODV
for performing control operations. These have been listed below.

• RREQ: A route request packet is used by the source node to discover a path to
some destination, when needed. This is used in the broadcast mode and uses a
unique sequence number for different destinations.

• RREP: The route reply packet sent (in unicast mode) by the destination to the
source on receiving the RREQ. This follows the reverse path setup by the travelling
request packet.

• RERR: Packet containing the notification regarding a broken link that is sent to
other nodes in the network. These nodes react to this notification by removing
the entry from their routing tables.

Using these packets, the source node is able to find a route to the destination only when
required (on demand) using the distance vector algorithm [14]. The working of AODV
is shown in figs. 2 and 3 and is explained in the following section.

Figure 2: Figure showing use of RREP


Figure 3: The source using the forward
message by destination node being sent
path setup by destination to route a
using reverse path to source. Direct
data packet. Direct from [10]
from [10]

3.2.1 Path Discovery Process


The first phase of path discovery invovles the RREQ packet sent in broadcast mode by
the source. This is further forwarded by the nodes till a node is found to have a route
to the destination (or its the destination itself).

11
3.2.2 Forward and Reverse Path Setup
As the RREQ packet travels through the network in search of the destination, it sets up
a route on the way called the ’Reverse Path’. This is the path that has to be used by the
destination to send a route reply packet. Once the destination (or a route containing
a path to it) is discovered, it sends the RREP packet, setting up a ’Forward Path’ for
communication. Data can then be transmitted along this path to the destination.

3.2.3 Routing Table maintenance


Based on the routes that are discovered using the processes explained above, a routing
table is maintained by the nodes. A few timers are used by AODV protocol to maintain
the freshness of the information in the table.

• Route Request Expiry timer : As the RREQ packet is broadcasted in the network,
many different neighbours relay this packet in attempt to find a route to the
destination. This process also includes setup of a reverse path along the way,
in case the source has to be informed regarding the route discovered. The main
purpose of this router is to remove these Reverse-path entries from the routers
that do not have a route to the destination after a certain timeout. It makes sure
only those nodes are active in the protocol that are in the path to destination.

• Route Caching Timer : There are two circumstances underwhich a route will be
removed from the routing cache maintained by the protocol. If a RERR packet
announcing a link failure is received by a node, it immediately removes the route
from its cache. A route caching timer is also used by the node to remove routes
after the timeout has ocurred if the route has not been confirmed valid (or re-
freshed) by a successfully routed packet. Due to the on-demand nature of the
protocol, that route is not calculated again until a packet is to be routed to the
destination.

Overall, AODV proves to be a useful and efficiency protocol for applications that do not
constantly require new route discovery and where bandwidth conservation is given more
importance than data latency. This protocol reduces the control overhead at the cost
of delay caused by route discovery process. These efficiencies have led to this protocol
being adopted as the standard Zigbee Distance vector routing protocol.

12
3.3 Collection Tree Protocol (CTP)
Collection Tree Protocol, detailed in [3] was developed by the Network Protocol Working
Group at UC Berkeley and is used as the standard collection routing protocol in TinyOS.
This protocol has been tailored to the requirements of WSNs. It provides best-effort
anycast datagram communication to one of the collection roots (the gateway node) in
a network.

3.3.1 The Concept of Collection Protocols


Collecting data at the base station (BS) is one of the most common requirements of
WSNs. The approach used is to build one or more trees, rooted at the base station.
There can be multiple tree roots. Hence, using anycast delivery (where data received
by any one of the base stations) is acceptable.
When a node has data that needs to be collected, it is sent up the tree. The other
duty of a node is to forward data that is being sent upstream to the node by others.
If there are multiple tree roots (data collection base stations) then multiple trees are
maintained other nodes join these trees. The route is decided by simply chosing the
correct tree (with the destination node at the root) and sending the data upstream.
These protocols make a good effort to deliver the data to atleast one of the many roots
in the network and do not offer any delivery or ordering guarantees for the network.There
are several challenges faced by the protocols in this family.

1. Loop detection: Due to simplicity of the tree-based routing scheme used by these
protocols, the networks are highly susceptible to routing loops in case a parent
node choses a descendant node as its next hop. This has been explained in more
detail under CTP features later.

2. Duplicate Supression: The protocols need to save network bandwidth by supress-


ing duplicate packets that are sent due to a lost acknowledgement or congestion
during transmission of first packet.

3. Link Estimation: Link quality is subject to effects like fading in wireless networks.
Due to these inherent properties of WSNs, the protocol needs to use a clever
link estimation scheme. This aims to shield the protocols functioning and routing
decisions from small variation in link state.

4. Self-interference: The protocol needs to prevent the packets it forwards from in-
terfering with packets that follow.

13
3.3.2 Key Features of CTP
CTP is a tree based collection protocol for WSNs where a number of nodes broadcast
themselves as root nodes and others join them leading to a tree structure. CTP is
address-free in nature, where a node does not send a packet to a particular destination
node, but simply moves towards any root by choosing a next hop. Some features and
problems faced by CTP have been listed below.

• Routing Gradient Metric: A routing gradient metric is used to generate the


routes to the base stations. CTP suggests Expected Transmissions (ETX) as its
routing gradient. In this scheme, ETX of Root is set to 0 and ETX of any other
node is calculated using:

ET X(N ode) = ET X(P arent) + ET X(Link) (1)

A lower ETX value is always preferred for a transmission, leading to a diffusion


from the root.

• Loop Detection and Prevention: Loops are a result of a parent node chosing
its descendant as the next hop. This is shown in Fig. 4

Figure 4: Figure explaining the formation of loops in WSNs using CTP. A being the parent
of B is the usual next hop. If the A-B link goes down, its next hop would be changed to D,
thereby leading to a loop.

CTP uses beaconing as a partial fix to this problem. At the time of next beacon
transmission, the faulty links are removed and the tree structure is cleaned up.

• Packet Duplication: Packet duplication takes place at a node when the first
packet is received and an ACK is sent, however the ACK is not received at the
sending node. Hence, it retransmits the packet leading to duplicates at the receiver.
This is also seen frequently in cases of looping, where a packet with the same

14
sequence number is seen multiple times. To deal with this situation, a Time has
lived (THL) field has been added to the CTP packets, which is incremented by
routing layer each time the packet is forwarded. This helps in differentiating looped
packets from link-layer retransmitted packets (as they have the same THL value).

3.3.3 CTP Noe : Improvements over Basic CTP


CTP noe [2] offers improvements over the basic CTP by introducing the following key
features:

• Datapath Validation suggests the use of data packets along with control packets
for updating routing information and in detecting loops.

• Adaptive Beaconing: Beacons are useful in collection based protocols but are
expensive in terms of overhead. Hence this scheme suggests to start with a small
beacon delay and keep doubling it upto a certain threshold. In case something
changes (loop detected or link state changed), switch it back to the low delay
(more beacons in a certain time frame). This ensures there is fast beaconing at
changes but slow at steady state.

• Transmit Timer is used to avoid self-interference. This ensures that the trans-
mitter waits for two packet intervals before sending the second packet.

• Transmit Cache suggests that the transmitted packet should be stored tem-
porarily to detect duplicates arriving at a node.

The addition of these features helped in improving the performance of CTP. Exten-
sive testing was performed by authors in [4]. Tests were performed using 12 testbeds,
7 different platforms and 6 link layer protocols. The results showed that it was able to
achieve 90% packet delivery (on average). The use of beacons was also reduced by 73%.

15
4 Discussion and Conclusion
The report discussed how the different requirements of WSNs have influenced the vari-
ous Routing protocol design ideas that have originated in this field. There are a variety
of protocols that have been suggested and choice of a protocol for a system depends on
the requirements of the application. Pro-active protocols with readily available routes
would be preferred for a network with low latency needs but sufficient bandwidth avail-
ability. Reactive protocols, on the other hand would be more suited for dense networks
with only a few nodes, acting as destinations (end-points), where data needs to be sent.

Each of the listed classes were discussed and selected state of the art protocols from
a few of these were explained in depth. AODV and OLSR protocols were studied, with
emphasis on their key features, functioning and benefits. These two protocols served
as excellent examples to compare and contrast the properties of reactive and proactive
protocols. They build on commonly used Distance Vector and Link State algorithms
that were optimised further to aid in functioning of ad-hoc networks in general. The
design of the Collection Tree Protocol was also described. This protocol, having being
tailored to WSNs, gives a clear understanding of the requirements and acceptable trade-
offs in WSNs. Improvements to CTP leading to the development of CTP Noe were also
discussed. Routing, like MAC is also a vast area, where different designs are aiming
to address different challenges in a variety of ways. This variety makes these protocols
quite difficult to compare on a single platform. Due to this, there does not exist a single
source that compares the performance of all these protocols. Hence, this report mainly
focussed on the design aspects of these protocols and the ideas that originated from them.

For each of the protocols chosen, the report highlighted the problems addressed,
basic functionality and key features. It was seen how different types of protocols trade-
off different properties in order to achieve system-wide efficiency gains.

16
References
[1] Jamal N. Al-karaki and Ahmed E. Kamal. Routing techniques in wireless sensor
networks: A survey. IEEE Wireless Communications, 11:6–28, 2004.

[2] Rodrigo Fonseca, Omprakash Gnawali, Kyle Jamieson, Sukun Kim, Philip Levis,
and Alec Woo. Collection. Technical report, UC Berkeley Network Protocol Work-
ing Group, February 2007.

[3] Rodrigo Fonseca, Omprakash Gnawali, Kyle Jamieson, Sukun Kim, Philip Levis,
and Alec Woo. Tinyos tap 114: The collection tree protocol (ctp). Technical report,
UC Berkeley Network Protocol Working Group, February 2007.

[4] Omprakash Gnawali, Rodrigo Fonseca, Kyle Jamieson, David Moss, and Philip
Levis. Collection tree protocol, 2008.

[5] Wendi Rabiner Heinzelman, Anantha Ch, and Hari Balakrishnan. Energy-efficient
communication protocol for wireless microsensor networks. In 33rd Hawaii Inter-
national Conference on System Sciences, pages 3005–3014, jan 2000.

[6] Wendi Rabiner Heinzelman, Joanna Kulik, and Hari Balakrishnan. Adaptive pro-
tocols for information dissemination in wireless sensor networks. In ACM/IEEE
5th Mobicom, pages 174–185, Seattle, WA, aug 1999.

[7] Philippe Jacquet, Paul Mhlethaler, and Amir Qayyum. Optimized link state routing
protocol. Internet-draft, IETF MANET Working Group, November 1998. Expira-
tion: May 1999.

[8] Brad Karp and H. T. Kung. GPSR: Greedy Perimeter Stateless Routing for Wireless
Networks. In Proc. ACM Mobicom, pages 243–254, Boston, MA, August 2000.

[9] J. Moy. The OSPF Specification. Technical report, RFC, 1989.

[10] University of Luxemberg SECAN lab. Ad-hoc Routing protocols.

[11] Charles E. Perkins and Elizabeth M. Royer. Ad-hoc on-demand distance vector
routing. In 2nd IEEE Workshop on Mobile Computing Systems and Applications,
WMCSA ’99, February 25-26, 1999, New Orleans, Lousiana, USA, pages 90–100.
IEEE, IEEE, February 1999.

[12] Cisco Systems. Routing Basics — CISCO Technology Handbook, 2010. [Online;
accessed 6-June-2010].

[13] Laurent Viennot. Complexity results on election of multipoint relays in wireless


networks. Technical Report RR-3584, INRIA, 1998.

17
[14] Wikipedia. Distance Vector Routing Protocol — Wikipedia, The Free Encyclope-
dia, 2010. [Online; accessed 22-April-2010].

[15] Wikipedia. Routing — Wikipedia, The Free Encyclopedia, 2010. [Online; accessed
4-June-2010].

18

Вам также может понравиться