Wireless Ad Hoc and Sensor Networks

Prof. Sudip Misra

Department of Computer Science and Engineering
Indian Institute of Technology, Kharagpur

Lecture – 02
Introduction: Wireless Ad Hoc Network-Part-II

So, this is the second module for Wireless Ad Hoc networks. So, here in the first module
we have already seen that there are different challenges that have to be overcome in
order to implement and deploy these kind of networks, Ad Hoc networks. These indeed
are very attractive, but at the same time the challenges are also huge. So, what are the
different challenges that we will see over all the different lectures on Ad Hoc networks,
but now for the next few moments we are going to go through some of the other
challenges that are involved in deploying these networks we are going to look at them at
high level and then in the next lectures we are going to go through them in more detail.

(Refer Slide Time: 01:23)

So, the first one is medium axis. So, as you know that medium axis is a property in the
second layer, is a function in the second layer which is the link layer of the OSI stack and
the medium access is all about that you have a collection of different nodes and these
nodes this we share that the same medium and how one can you know, you know one
can offer access to these different nodes in a fair manner is what medium access control
is all about.
Now, there are different protocols that have been proposed for medium access control in
different wire and wireless networks, but these are not very useful, these are not very
useful in this kind of Ad Hoc network because what is required is to offer distributed
access it is required in these networks to offer distributed access to the common channel.
So, the channel is one and with in that particular channel there is a single channel and in
this particular channel distributed axis has to be offered. So, the traditional MAC
protocols that one is familiar of basically they do not deal with you know this kind of
problem, additionally there is problem of host mobility in minutes which are also not
dealt with in the traditional MAC protocols. And additionally what we have is an
environment where there is no centralized coordinator.

So, consequently if there is no centralized coordinator the traditional schemes such as the
TDMA time division multiple access or frequency division multiple access FDMA, these
are schemes which are not very suitable for implementing in this kind of networks
because you know if there is no centralized coordinator a designated one which can help
in polling and you know in offering different slide time slots or frequencies to the
different you know the different nodes. So, who is going to take care of this, because we
do not have such an entity there is no centralized coordinator. And additionally there are
issues called the hidden terminal problem and exposed terminal problems which
basically haunt this kind of networks. So, there are multiple problems of medium access
in these networks.

(Refer Slide Time: 03:59)

Let us now talk about energy efficiency, a very important issue when it comes to Ad Hoc
networks. As I said in module 1, that these networks the nodes in these networks the
nodes in Ad Hoc networks they are battery powered. So, there is no infinite battery that
powers these nodes and it is very limited so typically you know once the battery is fully
charged it can be from you know it can last the batteries can last for you know couple of
minutes to you know to a couple of hours right, a couple of hours at the most. So, that is
the maximum that a fully charged battery in these networks, in the use, in these nodes in
these networks they can last for.

So, there are different you know there are different ways in which energy has to be
considered you know. So, energy is consumed due to transmission, due to computation,
re transmission, by the different nodes the intermediate nodes who returns meet the
packets that that it receives re transmission and reception of nodes. So, all of these
computation, communication, reception, transmission, retransmission all of these things
you know these tasks they consume power or energy.

Now, at the same time we have seen that the batteries they have very limited life cycle
they can operate you know once fully charged you know they will not be able to last for
too long. So, additionally there are other issues with these batteries they have very you
know, they have very highs discharge rate typically know the current they every batteries
that are used in these Ad Hoc networks these have very high discharge rate, low you
know life cycle and so on and so forth.

So, what is required is for using the nodes of these networks it is required to have high
performance and smart batteries and which will support mobility of the devices also; that
means, the mobile which can be used in the mobile devices and these batteries should
have low self discharge rate, this would have long life cycle, they should be able to be
operate, they should be able to operate in wide temperature ranges and they should have
high energy density.

So, the commonly used batteries in this different nodes of these networks include the
nickel cadmium NICD, NIMH nickel metal hydride, lithium ion; lithium ion is more
common you know lithium ion batteries and, because the energy is very scarce in the
different nodes of these networks what is required is to take care of energy, manage the
energy at all levels of this networks bit at the device level the different protocols that are
designed for these networks or the different applications that are you know design for
use in these networks, at all different levels energy management is a very crucial issue.

Now, one thing you might have already observed that we are using the term energy and
power in an interchangeable manner. So, as you are probably aware that energy and
power they are you know inversely related, so sorry I am sorry energy and power, power
is basically energy consumed per unit time. So, this is how energy and power are related
and you know, so that is why we often interchangeably used these two terms.

(Refer Slide Time: 07:47)

So, next is cooperation, next issue is cooperation in wireless Ad Hoc networks, this is a
very important issue this is that is typical of Ad Hoc networks because you know what
we have is a multihop scenario source node sending to the destination node and this is
something that we have already seen source node, sending the data to the destination
node, the destination node is not within the direct transmission range of the source node.

So, there are intermediate nodes which will act as relays for the data that is sent by the
source node to be delivered to the intended destination node. So, we have these
intermediate nodes which are acting as relays. So, quite understandably quite; obviously,
these nodes they have to cooperate the intermediate relays would have to cooperate with
the source and the destination node for retransmission to be successfully completed. And
at the same time these nodes as we have already seen, all these nodes the relay nodes
indeed are very much energy constrained.
So, naturally what happens like in a social network; that means a network of human
beings for examples when there is you know scarcity of resources there is an issue of
cooperativeness you know, some nodes may or may not be willing to cooperate. So, this
sort of thing is typical of Ad Hoc networks. So, there are two extremities that is
encountered with respect to cooperation in Ad Hoc networks - one is the case a case of
total cooperation. So, if all the relay requests are accepted then we have a case of total
cooperation; that means, the nodes they cooperate fully and that is not a very good thing
because the nodes they themselves have very limited energy and if they relate all the
requests that are you know that are sent to them then they are going to be there, they are
going to exhaust of their energy quite fast and we have the case of total non cooperation
on the other extremity where none of the relay requests that are sent to a particular node
are accepted by it.

So, any packet that is sent to the node would be dropped by that particular node. So, the
network throughput in such a case is going to go down quite rapidly; that means, the
network is not going to function for too long. So, neither of these two extremities; that
means, the extremity of total cooperation or the estimate of total non cooperation is ideal
for implementing in these networks.

So, there are additional issues with respect to cooperation because of the above things
that I just mentioned issues of selfishness, selfishness then self interest symbiotic
dependence; that means, the nodes you know ideally would like to you know
symmetrically depend on each other. So, if one node basically helps another node in
relaying its packets though it would also expect that the other node is also going to help
it in relaying, so there is some kind of mutual dependence symbiotic dependence that is
basically you know typified in this between the different nodes in these networks.
(Refer Slide Time: 10:55)

So, these networks because cooperation is a very important issue these networks have
nodes which have a tendency to misbehave. So, one of the first works, one of the first on
misbehavior in minutes was done by a group of a scientists Marti and others and it was
published in a paper in a c m MobiCom in 2000. So, in their work in Marti et al’s work
they basically defined the misbehaving nodes to be the ones which agree to cooperate,
but eventually who which do not write. So, initially they would agree to cooperate with
the other nodes, but eventually when it comes to actually cooperating they do not.

So, what is important is to intelligently identify the misbehaving nodes and avoid
relaying through these nodes this is what basically is important, that is this is basically
fundamental thing that has to be implemented and this is what was proposed by Marti et
al in their particular work.

So, in their proposed solution they have you know their solution to this particular
problem has two components - one component is known as the watchdog which basically
runs on every node to keep track of the behavior of the other nodes in the network and
the other one is the path rater which basically uses the watchdog information to find out
the reliable roots. So, watchdog is keeping an eye on the behavior or the misbehavior of
the different nodes in the network, where as the path rater based on the observations of
the watchdogs the path rater is determining that what which routes are the reliable ones;
that means, which nodes in the different nodes roots would help the packets to be
delivered successfully to the intended destination node.

(Refer Slide Time: 12:48)

So, there are you know there were different other works with respect to cooperation
works on incentive offering incentives for cooperation a bunch of European scientists
Buttyan and Hubaux in the year 2000-2001. They published couple of works in
MobiHoc and IEEE Communications Magazine where did suggest providing incentives
for cooperation without overloading the network. So, what they proposed is some kind of
a reward penalty model. In this particular model the nodes that are receiving a service are
charged and the nodes which are providing the service they are immunerated and based
on this particular model the different relay requests at the nodes they are either accepted
or rejected.

So, the drawback of such a model is that there is large overhead that is involved in
implementing such a model or such a solution and that is not very good in resource
constrained networks that of networks and sensor networks alive.
(Refer Slide Time: 13:59)

One of the ways of handling the issue of cooperation is to promote reputation promote
you know keeping track of reputation of the behavior of the different nodes in the
network. So, reputation means that how the different nodes are behaving so; that means,
that if a particular node is sent a packet, keeping track that whether it is forwarding it or
it is dropping it in between.

So, keeping track of this thing is a very important thing and how one keeps track of the
reputation of the difference. So, if a node basically drops the packet; that means, it is not
a node which is reliable. So, its reputation level is going to go down and if a node
forwards then its reputation level is going to go up. And at the same time it is also
important to track, to identify and track the liars in these networks you know which
nodes are liars; that means, that the nodes which are believed to be able to cooperate, but
which basically do not you know cooperate much they are not much cooperative. So,
these are termed as the liars.
(Refer Slide Time: 15:11)

There are security challenges in cooperation as well because you know ideally
cooperation is very good you know it is a open medium you know which is shared by
different nodes, same medium is shared by different nodes and these nodes they
dynamically change positions. At the same time there is no centralized network
management or certification authority and at the same time because it is open medium
any node can get into the network and can be malicious and can hurt the functioning of
any of the existing nodes in the network. So, these nodes are prone to attacks infiltration
eavesdropping interference so on and so forth, all these different types of common
attacks.

The main problem is that with the existing solutions one could have handled these
problems because there are well known solutions how to handle attacks infiltration
eavesdropping it etcetera etcetera, but here we cannot have we cannot assume that we
will have you know a server which basically a server or a you know centralized authority
which can you know keep track of the behavior and which can help in preventing these
attacks. So, such a thing is not possible in these networks the Ad Hoc networks ok.

And on top of that if a node is cooperating with the other nodes it might so happen that it
can fall victim from a malicious node because you know a malicious node can take
advantage of the cooperating node it can send number of you know it continuously it can
stream different data and this good node the co operative node it is going to soon run out
of its existing limited resources, energy etcetera etcetera. So, it is very imp very easy to
launch a denial of service kind of attack in these you know cooperative network
environments.

(Refer Slide Time: 17:11)

There are some routing challenges as well because you know, so not only that we have a
multihop kind of scenario, but at the same time we also have an environment which has
high rate of consumption of power at the different nodes which has low bandwidth high
error rates because the environment is such that there is high error rate and you know,
what we have are the nodes that randomly they move from one point to another and so
on. So, routing becomes a very important challenge and secured routing privacy aware
routing are other different routing challenges that are you know very important
considerations in these networks.
(Refer Slide Time: 17:53)

So, when we talk about routing in Ad Hoc networks let me just give you a very high
level overview of what are the different routing solutions that are proposed for Ad Hoc
networks. For using Ad Hoc networks the there are two classes of two main classes there
are few others also, we will talk about now two main classes of routing protocols that are
proposed for Ad Hoc networks. So, one is called the proactive routing protocols or
periodic routing protocols sometimes these are also known as table based routing
protocols.

So, the traditional weared routing protocols for example, OSPF which are table based
belong to this particular category. So, these protocols they maintain consistent up to date
routing information of the whole network and examples. So, that means, that at every
node in the network the routing information is stored and these tables are updated
periodically. So, examples of these networks are the DSDV protocol, the WRP protocol,
GSR, FSR and so on and so on, there are many many more you know Ad Hoc network
routing protocols belonging to this particular category that have been that has been
proposed in the literature.
(Refer Slide Time: 19:10)

The other category of Ad Hoc network routing protocols is called the on demand or
dative routing protocols. So, as this name suggests that whenever there is a demand the
source node basically initiates route exploration discovery of route you know disk; that
means, the you know the source node whenever it has to send some data and that you
know it is not able to find the route to the intended destination node it is going to start the
route discovery process. So, these solutions basically unlike the previous category which
where table based here basically there are that you know these different nodes in these
networks they do not maintain routing information like in the previous category.

So, the routes are created only on demand by the source node and, so typically there is a
three phase kind of approach that is used in any kind of our most of the kinds of on
demand reactive routing protocols. It first starts with the route discovery process phase
route discovery phase then comes the route maintenance phase; that means, the roots are
once discovered they are maintained until the root is no longer required and finally, when
the root is no longer required there is the teardown phase; that means, the root is gone
down.

So, these are better in Ad Hoc networks you know these reactive routing protocols they
are better in Ad Hoc networks as they are you know resource limited and the nodes are
mobile, so because the nodes are mobiled, because the resource is very limited, so it is
very important to you know to track the mobility of the nodes in the network and
discover the roots whenever it is require. Because it is quite unlikely that in a very
mobile environment you know the existing topology information is going to stay interact
over time. So, the topology information gets changed over time and consequently it is
better to you know discover the roots in these networks.

But keeping in mind this particular advantage we also should understand that when we
talk about Ad Hoc networks it is not that all the time the topology is going to be you
know made and broken quite fast in these networks. There are some networks where the
topology may not be changing quite fast in such a case every time discovering the route
maintaining etcetera etcetera as in a reactive routing protocol may not be a very good
approach. So, in such cases maybe that using the proactive approach; that means, the
table based approach would be a better solution. So, probably that you know in such a
situation it might be better, it might be efficient to use the proactive routing protocols.

Deactivate routing are on the other hand as I just mentioned would be more efficient
when we are using them in networks where the nodes are changing their positions quite
fast; that means, the nodes are quite mobile and it is important to keep track of the
mobility of the different nodes and based on that not keep track of the mobility of the
nodes it is difficult to keep track of the mobility of the nodes and, so accordingly find the
different routes in the network. So, some of the you know examples of Ad Hoc routing
protocols belonging to the reactive category are the DSR dynamic source routing,
associatively based routing ABR, UDV is a very popular protocol Ad Hoc on demand
distance vector routing protocol and power aware routing protocol PAR.

So, like in the proactive category there are many more types of routing protocols
belonging to this category that has been proposed in the literature.
(Refer Slide Time: 23:03)

Another very important you know issue is fault tolerant routing which is something that I
have worked on in my research before. So, this is this has been one of my research paper
this led to one of my research papers that was published in the telecommunication
systems journal.

So, here we talk about that you know false are quite common in minutes, so that is
something that we have seen and packet drops result due to false, false could arise
because maybe two nodes which were in the proximity of each other they are no longer
in the proximity, maybe when one node moves and consequently packet is dropped in
between. What there could be different other reasons why you know false could develop
between different nodes in the network. So, there could be problems in the link; that
means, failure in the link there could be faults in the node and so on. So, we basically
proposed a fault tolerant solution to this particular problem.
(Refer Slide Time: 24:09)

So, the fault tolerant routing problem it basically says that in the presence of fault how to
efficiently route packets. So, faults are inevitable in Ad Hoc network. So, in the presence
of faults how to route packets. So, one of the you know before our work, one of the
previous works was the E to E F T algorithm, end to end fault tolerant routing algorithm
this algorithm basically involves two major phases - one is called the route estimation
phase the other one is the route selection phase. In the route estimation phase as the name
suggests what is required is to estimate the packet delivery probability of all the routes at
the disposal at any time instant.

So, you have number of possible routes. So, what is required between a source and
destination pair. So, what is required is to estimate probabilistically estimate the packet
delivery probability; that means, you know the probability of packet delivery through
each of these different you know options which of these different routes. After that
estimate is obtained then what the E to E F T algorithm does is it you know selects one
of these routes that have confirmed to have satisfied a certain optimization constraint a
particular threshold let us say a particular, particular threshold when it crosses that
threshold. So, that particular route is selected and the other routes other possible routes
are basically dropped and are not made available for routing.
(Refer Slide Time: 25:37)

So, our algorithm that we proposed was as I told you it was published in the
telecommunication systems journal by Springer and it is known as the WEFTR week
estimation based FTR fault tolerant routing algorithm. So, here also we have the route
estimation on the route selection phase, but the way we estimate the packet delivery
probability through each of these different alternatives that we have seen in the you
know E to E F T algorithm. So, that basically does not exist you know, so we have
improved upon that we have proposed a week estimation based learning mechanism for
estimating the probability of delivering packets through the different routes. So, this is
the improvement that we have done over the E to E F T algorithm.
(Refer Slide Time: 26:31)

And here is a comparison of performance with respect to criteria such as the percentage
of delivered packets the overhead etcetera and we have seen that you know the proposed
algorithm performs you know better with respect to you know; that means, that it has a
lower overhead and hire packet delivery probability; that means, that there is better
probability of delivering packets to the intended destination by using the WEFTR
algorithm. So, what we talked about is unicast routing.

(Refer Slide Time: 27:03)

Now, what about multicast? So, we have as we have seen in minutes we have an
environment where the topology changes quite first we have frequent changes in the
network topology and the nodes they move on the fly connections are made and broken
on the fly and so on and on at the same time we do not have we cannot assume the
existence of a centralized coordinator like a base station or a mobile repeater.

So, the traditional approaches to multi casting which are typically based on forming and
maintaining trees known as multicast trees are not quite useful in these environments. So,
protocols have to be proposed, new protocols have to be proposed which will be highly
adaptive to be able to cope with the high degrees of dynamism that is typical of these
environments.

(Refer Slide Time: 28:00)

So, there are different classes of multicast routing algorithms that have been proposed we
will talk about these in detail later on. So, broadly they can be classified as flooding
algorithms or source based tree algorithms, core based multicast routing and group based
multicast routing. So, these are some of the different categories of multicast routing
algorithms that have been proposed purely for use in Ad Hoc networks. Flooding
algorithms are like other flooding algorithms that you might be familiar with, if you have
taken a networks course before.

So, where global, where basically you know the packets the multicast packets are
basically flooded globally in the network. So, you know, so that this sort of approach is
good if the environment itself is highly mobile; that means, the nodes are moving quite
fast; however, it leads to wasted bandwidth due to unnecessary forwarding of duplicate
data and that is also not a quite good thing for use in Ad Hoc networks because we have
a very limited you know resource constrained environment and that is not very good. But
at the same time if the nodes are moving quite fast you do not have any other option was
do, but to go for a knife flooding based approach.

Source based multicast tree, so here multicast trees are established and maintained for
each multicast source node in each multicast group. So, if we have G number of
multicast groups and S number of source nodes in each group. So, what is required is to
have maintained, it is required to maintain g times s number of multicast trees and that is
not a very good thing because that will lead to you know a solution which is not very
scalable. That means, if the number of nodes increase if the number of you know
multicast groups increase or the number of source nodes per group increases. So, it will
lead to soon it will lead to a very you know not a very elegant solution in these networks.

Examples of this include you know different multicast source based multicast tree
approach protocols include the DVMRP protocol, MOSPF protocol and so on.

(Refer Slide Time: 30:14)

Core base tree basically you know what is done is it is, what it does is it creates and
maintains a single shear tree connecting different multicast group members, so that is the
approach that is adopted in the core based multicast trees. And this is a very you know
this is a better solution than the SBT is; that means, the source based trees, but at the
same time this also had this disadvantages because traffic is concentrated in the solutions
on the share links only and that can often lead to condition in these share links.

Multicast mesh is another one where a node can have multiple parents and mesh is
formed spanning all multicast group members and it is good in terms of ensuring
reliability, but it is bad, it is advantageous because there is unnecessary forwarding of
multicast packets along all the redundant paths. Example of multicast mesh routing
protocol for Ad Hoc networks is the camp protocol which is the full form of which is the
core assisted mesh network protocol.

(Refer Slide Time: 31:22)

And finally, the group based you know a multicast forwarding where a group of
multicast forwarding nodes is maintained for each multicast group only the forwarding
nodes are responsible for forwarding the packets processing at other nodes is simplified.
All received multicast packets that are not duplicated and rebroadcasted by the
forwarding nodes to the all received pack or multicast packets are not duplicated and are
rebroadcast by the forwarding nodes to their neighbors. And, example of group based
multicast forwarding include on demand multicast routing protocol ODMRP and the
location based multicast routing protocol.
(Refer Slide Time: 32:03)

And few you know a few other important issues in Ad Hoc networks power is an
important issue that we have already seen. So, what is required is not only at the protocol
level not only at the algorithmic level it is also required to have devices manage the
power right. So, there are different manager device manager, device power management
techniques that have been proposed APM is one, ACPI is another. APM stands for
advanced power management which basically comes with a comprehensive power
management scheme at individual device level of the mobile computing devices and
which divine defines a hardware independent software solution. ACPI is also quite
similar, we will talk about this later, but just as a cursory level this is what device power
management does.
(Refer Slide Time: 32:54)

Protocol power management is required, protocol we know at all different layers the
different protocols that are proposed in the protocol architecture, they you know they all
have to ensure that you know they consume very less power, reduction there should be
reduction in the data link layer, reduction in power consumption in data link layer. There
should be deduction in power consumption at the network layer at the transport layer and
so on. So, these are the different things that have to be taken care of at the different
layers.

(Refer Slide Time: 33:24)

So, one other important thing is that we have talked about routing on top of routing is the
transport layer protocol common transport layer protocol that is used in the internet is the
transmission control protocol TCP, and as we know that in TCP there is some kind of a
handshaking door that goes on between the sender and the receive receiver right. So,
basically you know segments are sent from the sender and in a sequence and these
segments in the sequence they are supposed to be received at the receiver intended
receiver.

But it might so happen that some of these segments can be lost in between and these lost
segments have to be retransmitted. So, there are different other issues that are associated
which are like you know control of flow of these segments between the intended transmit
transmitter between the transmitter and the receiver. So, flow control, congestion control;
that means, that you know if the receiver is receiving at a rate which is more than how it
can handle or the rate at which it can handle then what is going to happen is that at the
receiver end the you know there will be congestion and consequently there will be packet
drop, so that is undesirable ok.

So, this problem has to be you know handled, but the main problem is that if you want to
use the TCP in Ad Hoc networks. So, typically you know the network layer first what it
does is once a link is broken one of the existing routes is broken, due to maybe no
mobility or something like that the you know it basically you know initiates, the route
the network layer would initiate the route repair and this route repair basically creates
delay of acknowledgement arrival and this delay is mistaken by the TCP sender; that
means, the transmitter as a sign of network congestion and consequently the congestion
control mechanisms are invoked by the transport layer and because of this basically the
communication throughput who is going to go down drastically.

So, this is the problem with using the traditional TCP. The TCP that was proposed for
you know internet for use in the internet using it for Ad Hoc networks is a big challenge.
(Refer Slide Time: 35:57)

So, there are different protocols we will talk about these protocols that perform better
than the traditional TCP for use in Ad Hoc networks. TCP feedback is one TCP bus is
another one there are many other different types of mechanisms that have been proposed
for use transport, for taking care of transport layer issues in Ad Hoc networks.

(Refer Slide Time: 36:21)

So, here are some of the references for the first two modules you know, so you know one
can consult any of these different books some of these materials and the different
concepts are borrowed from these books. So, the first two modules you know the
references are given over here.

Thank you.