• No results found

Energy Efficient Routing in Ad Hoc Networks

N/A
N/A
Protected

Academic year: 2021

Share "Energy Efficient Routing in Ad Hoc Networks"

Copied!
26
0
0

Loading.... (view fulltext now)

Full text

(1)

Institutionen för datavetenskap

Department of Computer and Information Science

Examensarbete

Energy Efficient Routing in Ad Hoc Networks

av

Emmie Nilsek och Christoffer Olsson

LIU-IDA/LITH-EX-G--14/065--SE

(2)

Examensarbete

Energy Efficient Routing in Ad Hoc Networks

av

Emmie Nilsek och Christoffer Olsson

LIU-IDA/LITH-EX-G--14/065--SE

2014-06-17

Handledare: Niklas Carlsson Examinator: Nahid Shahmehri

(3)

Students in the 5 year Information Technology program complete a semester-long software development project during their sixth semester (third year). The project is completed in mid-sized groups, and the students im-plement a mobile application intended to be used in a multi-actor setting, currently a search and rescue scenario. In parallel they study several top-ics relevant to the technical and ethical considerations in the project. The project culminates by demonstrating a working product and a written re-port documenting the results of the practical development process including requirements elicitation. During the final stage of the semester, students cre-ate small groups and specialise in one topic, resulting in a bachelor thesis. The current report represents the results obtained during this specialization work. Hence, the thesis should be viewed as part of a larger body of work required to pass the semester, including the conditions and requirements for a bachelor thesis.

(4)

This thesis presents a comparison between a basic shortest path routing policy of the Destination-Sequence Distance Vector (DSDV) protocol and two power-aware policy variations of it. In the two modified versions, the routes are selected based on the energy available on the nodes and not only the shortest path distance to the destination. Simulations are conducted for a given situation of nodes and the energy efficiency of the three aforemen-tioned policies are evaluated for example scenarios.

First, a brief overview of the theory behind the study is presented. It con-sists of an description of ad hoc networking, DSDV, and our energy-aware modifications to DSDV.

After the fundamental theory, the method is presented. It consists of a de-scription of how the simulated scenarios relates to a real-world scenario and the simplifications made in the model. We present an overview of the model used for simulation and the operation of the program. This section ends with an explanation of the three simulated policies: shortest path, simple weighted and doubled weighted.

When the theory behind the thesis are completed, the simulations are con-ducted. The results are examined and a summary of their meaning is dis-cussed. It is explained how the assumptions effect the reliability of the study and an estimation of the accuracy of the results are presented.

We find that the power-aware policy variations (simple weighted and double weighted) both achieve better network lifetime than the basic shortest path policy, at the cost of slightly longer per-packet paths. These results are encouraging and show that very simple modifications to DSDV can achieve significant gains in the network lifetime, helping users get the most out of their networks. Future investigation could try to optimize these gains.

(5)

Contents

1 Introduction 1 1.1 Background . . . 1 1.2 Purpose . . . 1 1.3 Problem statement . . . 1 1.4 Limitations . . . 2 2 Background 3 2.1 Ad hoc networks . . . 3

2.2 Destination-Sequence Distance Vector . . . 3

2.3 Energy efficient versions of DSDV . . . 3

3 Method 5 3.1 Scenario . . . 5 3.2 Simulation model . . . 5 3.2.1 Simplifications . . . 5 3.2.2 Model . . . 6 3.2.3 Program . . . 6

3.3 General policy model . . . 7

3.3.1 Shortest path . . . 7

3.3.2 Simple and double weighted . . . 7

4 Performance evaluation 9 4.1 Scenario overview . . . 9

4.2 Results . . . 10

4.2.1 Default scenario . . . 10

4.2.2 Impact of initial energy . . . 12

4.2.3 Impact of externally supplied nodes . . . 12

5 Discussions 15 5.1 Method . . . 15

5.2 Results . . . 16

5.3 Related work . . . 16 6 Conclusions and Future work 18

(6)

Introduction

1.1

Background

In situations like natural disasters or other social crises it is important for The Armed Forces and other operators within the authority to be able to maintain their communication paths. If the infrastructure of their networks and/or power stations break down they have to set up a temporary network. Ad hoc networks is a type of network that are commonly used in this type of situations. Another crucial problem in these situations is how to maintain power in their mobile devices.

1.2

Purpose

Today, energy awareness is very important as energy is a significant bottleneck for mobile devices with limited battery resources. When energy is in high priority, performance is compromised. This project investigates if and how an energy level parameter can be added to the shortest path routing with the DSDV protocol, such as to allow ad hoc networks to increase thier expected lifetime.

1.3

Problem statement

In this study we investigate how the DSDV routing protocol is suited for these type of situations. By comparing the original shortest path policy for DSDV routing protocol with our modified policies we answer the following questions:

• How can we do DSDV more energy efficient?

• How energy efficient is the shortest path policy in comparison with the modified policies?

• Can a more power-aware routing protocol be a better choice in these kind of situations?

• How should we weight the energy level parameter to maximize the lifetime of the network?

(7)

1.4. LIMITATIONS CHAPTER 1.

1.4

Limitations

This study is only conducted in a simulated environment. This is due to the lack of time and resources to create a real wireless ad hoc testbed. Furthermore, this study will only examine three variations of policies for the DSDV protocol and discover patterns in how the energy parameter could impact the performance of a real scenario. Given more time, a broader range of policies could be evaluated. The simulation only uses a few parameters to make it easier to see their impact on the performance. The parameters not considered and their exclusion from this study are explained and further discussed in Section 3.2.1.

This study only focus on the performance tradeoffs between the different poli-cies that is introduced in Chapter 3 and does not consider the potential uses that longer network lifetime can have on the users of the mobile devices. This means that potential risks with ad hoc networking in general (for example eavesdropping) are considered to be the same regardless of the results in this study. The reasoning behind that is that a longer network lifetime will neither produce nor reduce the possibilities of a threat being realised.

(8)

Background

2.1

Ad hoc networks

Wireless ad hoc networks is a network represented by device-to-device commu-nication without any centralized system and infrastructure [4]. It is supported only by mobile nodes who send data to each other by a wireless method such as bluetooth. The nodes act as both clients and routers and can therefore send data between several nodes only using themselves. In order to actually send data to a mobile node out of reach, the sender node pass the data to one node in reach, which pass it forward to another node, and so forth. The routing within the ad hoc network can be handled in different ways, and in this thesis we discuss a protocol named Destination-Sequence Distance Vector (DSDV).

2.2

Destination-Sequence Distance Vector

The Destination-Sequence Distance Vector (DSDV) protocol is a table-driven ad hoc routing protocol where all nodes store routing information to all other nodes in the network. Every route stored at a node holds a sequence number that indicates the freshness of the route. All the nodes sends updates to their neighbours to keep the tables fresh [8]. However, this study is focused on a situation where the stations (nodes) are stationary once the network is deployed. Therefore, the sequence numbers is irrelevant for this study; the first updated routes will always be the best routes in terms of distance. Updates will, however, have an important role in the modified version.

2.3

Energy efficient versions of DSDV

The modification that is applied to the shortest path policy for DSDV protocol for this study is that an extra parameter is added to the routing algorithm: the energy level of the nodes. This means that the protocol now weight in both the distance and energy when deciding the cost of a route. Similarly to the original protocol where the tables is updated on new routes, these tables are updated when the nodes reach different levels of energy. An important assumption is made

(9)

2.3. ENERGY EFFICIENT VERSIONS OF DSDV CHAPTER 2.

that these updates are negligible compared to the data that is transmitted in the network. This assumption is made considering the energy cost of updating all node tables being the same for all nodes, regardless of policy.

(10)

Method

To investigate ad hoc routing, we have developed a software. A real world scenario is the basis, followed by which simplifications made in order to form a model of it. This model is basis for a java program that is created to simulate a real scenario.

3.1

Scenario

Consider a group of soldiers located in a forest, each carrying a mobile device. There is no infrastructure of any kind, but they still need to communicate with each other. The mobile devices has a limited signal reach and the area they are located in is way too wide for the signal to transmit over. This indicates that they have to use ad hoc network to be able to get the message from A to B. Every mobile device has different energy levels, and power is a hard currency in these kind of situations. They are sending data through the network with the help of their peers’ devices. For each time a device send and/or receive data it loses energy. The group want to maintain the network alive, which fails when one of the mobile devices is out of power. A limited few of the soldiers’ devices are considered having unlimited power. That is, if they have access to extra batteries, a solar panel or a charger connected to a vehicle, e.g a tank. The challenge of this is to make the network survive as many transactions as possible and we will determine this by using different routing policies.

3.2

Simulation model

3.2.1

Simplifications

Since reality is very complex, it is necessary to form it into a simple and under-standable model in order to examine the scenarios. In order to do that, a couple of simplifications are done. To begin with, the nodes will have fixed positions. Having the nodes in motion would be a lot more complicated to implement and only affect how the path was calculated, not how energy decreases during trans-actions. Without having a significant impact on the end results, it is not included as a factor during the study.

(11)

3.2. SIMULATION MODEL CHAPTER 3.

The second simplification is how the reduction of energy level of the nodes are calculated. Updating tables at the nodes requires energy, as well as energy con-sumed when the device is actively managed by the user. In fact, the device loses energy as long it is turned on. The adoption done of this it that these types of factors has an equal impact on each node and network. For that reason, a node in this study is only considered to lose energy when sending or receiving data since that is the only factor that matters.

Another simplification is that only one transaction is made at the time. This is not happening in reality, but saves a lot of implementation issues and is not considered to have any significant impact of the end results. More simultaneous transaction would cause the network to die faster but not affect the relation be-tween the policies.

In the simulation we assume that the network already is deployed, that is the nodes are aware of the topology of the network from start, so there is not consider any network set up. It is reasonable to disregard this, because all the nodes in the network are involved equally and the energy level at the nodes are therefore considered to be affected equally.

3.2.2

Model

The model contains a network of nodes with fixed positions in a given area. Ev-ery node has a given signal reach, and are connected with the nodes which are located within it. Each node have an initial energy level, and some of them are considered to be externally supplied with energy (they have a fixed energy level at 100%). A transaction has a start node and an end node, and has to travel across other nodes if the start and end nodes are not directly connected. The routing policies determine the path. Later on, we discuss which polices we use. Each node along the path that act as a sender and/or receiver loses energy. Transactions are simulated until the network dies. In our simulation we let this point be the time instance when the first node runs out of energy. The lifetime of the network is measured in transactions made before it dies. In this study, three scenarios are executed. These scenarios are described in Chapter 4.

3.2.3

Program

The simulator is created as a java program of the model described in the previous section. The program include two random generators that creates two files: one for the topology of the network and one for the transactions that will be simulated. The topology file describes the placement of the nodes within a given area and in-formation about their initial energy level. The second file, the trace file, describes a large amount of transactions in form of source node and destination node. The program calculates the edges (connections between nodes) based on their signal reach. When the network is set up, the transactions can begin. Dijkstras algo-rithm is used to determine the shortest path for the transactions based on the weight/costs. Each node involved in a transaction get its energy level reduced and updated, as well as the cost depending on which policy that is simulated. The

(12)

made from the trace file. The program run one transactions at the time until the energy level of one of the nodes reach zero. The network is called dead, and the program output the number of transaction and hops that was made. The input parameters for the program is described together with the scenarios in Chapter 4.

3.3

General policy model

The simulation consists of three different policies that will be tested with a default scenario as base. The three policies tested are the original DSDV (shortest path) policy and two different modifications of it that weigh in the energy of the nodes when deciding which path to take. These policies is called simple weighted and double weighted. The two policies differ by how much the energy level affect the cost of transmitting to neighbouring nodes when making route decisions. The two weighted policies are used to evaluate the input of the energy awareness in the performance and network lifetime of the overall system. The interesting param-eters to investigate is the network life length and the number of hops needed to complete the transactions when this factor increases.

3.3.1

Shortest path

This policy chooses the shortest path between the two nodes that wants to com-municate with each other. It pays no attention to the energy levels at all since the only mission is to send the data from source to destination based on the distance. Thus the cost of a connection/edge between two nodes is the same as the distance between them.

3.3.2

Simple and double weighted

The simple weighted policy is used to make the energy of the nodes in the net-work a factor when deciding new routes. In this policy the weight is based of both distance and the current energy level at the nodes. The cost to travel through a node is the distance and an extra cost for the current energy level. The weight-formula used to weigh the cost of transmitting to neighbour nodes in this policy use the distance d between the nodes, as in the default policy, plus the energy E that has been drained from the next node, as measured in percentage of the total battery capacity. The weight d + E evenly balance the advantage of selecting short hops against the advantage of selecting to route over nodes with much remaining energy. The proportion between the distance and energy drainage for this policy should be normalized so the energy drainage maximum impact will be half of the maximum distance. The maximum distance is determined by the signal reach and the energy drainage depends of the measurement of energy level. Which numbers used in this study is explained in Chapter 4.

The double weighted policy is used as a complement to the simple weighted policy. In this policy the energy of the nodes have a greater impact on the route of choice. This means that it will rather take a longer route than both the shortest path and the simple weighted if it is more energy efficient. The weight of the edges are based on both distance and the current energy level as the simple weighted

(13)

3.3. GENERAL POLICY MODEL CHAPTER 3.

formula, but it has greater impact. The weight formula used to weigh the cost of transmitting to neighbour nodes in this policy is d + 2E, giving twice the weight to the energy drainage.

(14)

Performance evaluation

4.1

Scenario overview

The scenarios use a number of different input and output parameters that forms the base of the tests that is conducted. The input parameters of the scenarios is: • Number of nodes: The number of nodes placed randomly on the 1000x1000

map.

• Externally supplied nodes: The number of nodes with external energy supply (node with constant energy level of 100%).

• Signal reach: The distance that every node is capable of transmitting. • Energy loss: The energy loss for a node when sending and/or receiving

measured in percentage points.

• Energy variance: The interval of energy that the nodes is given at the start of the test. Every node starts with a random level of energy within the interval.

The different scenario we are going the simulate is as follows:

• Default scenario: Tests with 50-500 nodes including 10 externally supplied nodes, high energy case (initial 50-100%), 200 meters signal reach and 1 percentage point energy loss.

• Impact of initial energy: Like default scenario but with an low energy case (initial 20-50%).

• Impact of externally supplied nodes: Test with 0-90 externally supplied nodes out of 100 nodes and same values as default scenario for the other parameters.

To allow fair head-to-head comparison, for each scenario, each policy was evalu-ated under multiple random topologies and transaction traces. To measure the performance we measured the number of transactions that could be performed before the network dies and the total number of hops needed to do those trans-actions. The network lifetime is used to see how the different policies perform in terms of network lifetime and the number of hops is used to evaluate the route

(15)

4.2. RESULTS CHAPTER 4.

distances used in every test. This is to find a relation between energy saving and the total distance needed for the transactions since one might compromise the other.

4.2

Results

4.2.1

Default scenario

We wanted to investigate how the lifetime of the network differs with the shortest path, simple weighted, and double weighted policy. We also wanted to investigate how the number of hops differs between the policies. Furthermore, we want to see how the lifetime and hops relates to each other and compare the different policies. For each policy, the network was simulated with 50 to 500 nodes, changing the number of nodes by intervals of 50 (50, 100, ..., 500). To give a fair evaluation, every test was made ten times and data used in the graphs is an average of those. Figures 4.1 and 4.2 show the number of transactions and hops made, before the network dies, respectively, as a function of the number of nodes in the network.

Figure 4.1: Transactions for the default scenario

Figure 4.2: Hops for the default scenario

(16)

life duration for a policy who use energy level as a parameter to the weight for-mula. Another graph to look at the number of hops in Figure 4.2. The number of hops increases here also, which is relevant because more transactions causes more hops. An important point here is that the weighted policies actually visits a lower amount of nodes per transactions. The shortest path policy with 500 nodes makes an average of 4,3 hops per transaction, while simple weighted and double weighted have an average of 3,4 and 3,5 hops per transaction, respectively. This means that even if the routes are more energy aware they are not necessarily much longer. To explain what happens with the energy level at each node, we took a closer look at an example simulation. The simulation is made with 50 nodes, high energy case, and without any external supplied node. Figures 4.3 and 4.4 show the current energy level of the nodes when the network dies and how it differs from the initial state for the shortest path and the double weighted policies, respectively. We see that the double weighted policy have a more balanced energy level. The shortest path policy takes no account of the energy level at all, which is shown is this example.

Figure 4.3: Example of initial and final energy level for 50 nodes with short-est path policy

Figure 4.4: Example of initial and final energy level for 50 nodes with double weighted policy

Another way to display it is through the Cumulative Distribution Function (CDF) chart in Figure 4.5. This shows the distribution of the energy level at the nodes. The percent of participating nodes is a function of current energy level at the nodes when the network dies. When reading from the chart, the curves indicate how many of the nodes (y-axis) has an energy level less or equal to the value on the x-axis. For example; 70% of the nodes has an energy level of 70% or less with the shortest path policy while the nodes have an energy level of 40% or less with the double weighted policy. This show how the network make use of the energy level

(17)

4.2. CHAPTER 4.

of the nodes. Again, the double weighted policy brings a more smooth distribution than the shortest path policy.

Figure 4.5: Cumulative Distribution Function (CDF) over energy level of 500 nodes at the end of a simulation

4.2.2

Impact of initial energy

The previous results where computed with a high energy case. To investigate the impact of initial energy we repeated the experiment from Section 4.2, but with a lower initial energy at the nodes. Figure 4.6 and 4.7 show the number of transactions and hops as a function of number of nodes in the network for each policy.

Figure 4.6: Transactions for scenario with less initial energy

Comparing the results with those in Figure 4.3 and 4.4 we see that the results are consistent. Notice that the scale is different, but the curve remains alike. Whether the lifetime nominally reduces due to the lower energy level, the performance of the policies is not affected. This is intuitively since the relation between the energy level at the nodes and the energy level at their peers remains the same as in Section 4.2. This scenario also give the same hops per transaction as the default scenario.

4.2.3

Impact of externally supplied nodes

The results from the default scenario (Section 4.2) were computed with a low and fixed amount of externally supplied nodes. To investigate the impact of externally

(18)

Figure 4.7: Hops for scenario with less initial energy

supplied nodes we did a experiment similar to the one in Section 4.2, but instead of vary the number of nodes we varied the number of externally supplied nodes from zero to ninety with interval by 10 (0, 10, ..., 90) out of hundred nodes. Figures 4.8 and 4.9 show the results for each policy described by number of transactions and hops as a functions of number of externally supplied nodes.

Figure 4.8: Transactions with different number of externally supplied nodes

Figure 4.9: Hops with different number of externally supplied nodes

In the result from Section 4.2, we can see that the performance of the energy efficient polices was higher than the performance of the shortest path policy, but

(19)

4.2. CHAPTER 4.

there was no significant difference in performance between the weighted polices. In contrast to this similarity, we can see in Figures 4.8 and 4.9 that the performance of the double weighted policy is higher then simple weighted. The impact of externally supplied nodes for the shortest path policy is barely noticeable compared to the weighted polices. The shortest path policy with 90 externally supplied nodes makes an average of 4,0 hops per transaction, while simple weighted and double weighted have an average of 3,9 and 3,7 hops per transaction, respectively.

(20)

Discussions

5.1

Method

The method used contains a few simplifications that might impact the result. To begin with, the nodes lose the same amount of energy whether they send, re-ceive or do both. This is not in line with reality since doing both the sending and receiving should in reality demand more energy since they do two things. It is difficult though to determine how much this simplification actually affect the end result. We think this simplification do not affect the end result in any major way. This assumption is based on the fact that there is only two nodes in a specific chain that is not doing both receiving and sending; that is, the sender and receiver. In our model the nodes have fixed positions. By having the nodes in motion, the paths would be calculated differently, as mentioned in Section 3.2.1. This will not impact how the energy level decreases at each individual node. However, a changing topology may impact the lifetime of the network due to more possible paths, but this will apply for all policies. Our model may not beneficial to use on a more dynamic network since it is needed to update the network both when the topology changes (nodes move) and when the energy level decreases. Since this affects all nodes equally, it affects lifetime of the network. This means, even if we could get better energy savings when sending information, it could be at the cost of energy losses by constantly updating the costs of sending information.

Another simplification is that only one chain of transaction is made at a time. When a node is sending information via a chain of other nodes to a receiver, no other transactions are made in the network. It is something that we do not think affect the end result at all since the simulation would work the same even for more transactions. The simulation starts with a sender node A, a receiver node B and some data to be sent. When node A is done sending data to node B another random transaction is starting between two other nodes. This is done in a chain and should not do any difference to the end evaluation of the protocols compared to a simulation where the transactions is done simultaneously. This is difficult to prove to be true so it would be an improvement of the study to conduct more scenarios where more simultaneous transactions is made.

(21)

5.2. RESULTS CHAPTER 5.

Another simplification is that we assume that the route does not change during a transaction. The study would be more realistic if the end-to-end transactions were broken into data packages and the route was able to change during the trans-action, when the costs are updated, for example. With our current assumptions each transaction should probably best be interpreted as sufficiently small to avoid route changes during a transaction.

The last and most obvious limitation of the study is that this is only a simu-lation. A study of a real scenario with real nodes and information would allow evaluations under more realistic conditions. If this study was conducted in a real scenario, some of the simplifying assumptions would not be necessary, as param-eters like signal reach would be determined by the physical limitations of the physical devices themselves. However, our approach have certain advantages over a real scenario. It is easy implemented and can generate several test results much cheaper than a real scenario. It may not give the exact results of a real scenario but it is probably good enough to prove that the benefits and drawbacks of the policies presented. It can be used as a base for further development. What if we used a perfect real scenario, use a lot of money and resources and get all the ”perfect” (perfect in the sense of 100% real) results that eventually show us that we were wrong? It would be better to have a simpler and cheaper simulation with a few simplifications to see if what we want to test give us the result that we want (or something close to it) before we invest money and resources.

5.2

Results

As shown, the modified policies make the network live longer. In order to keep the network alive, it requires different paths for the transactions. Because these paths considered energy level as parameter rather then just the shortest path measured in distance, the actual distance of the paths becomes longer. Despite to the longer path, the number hops do not necessarily increases (which mentioned in Section 4.2). An assumptions made in the beginning of this study was to disregard the loss of energy over time. If the network manage to survive more transactions with the modified protocol, it is also a reasonable conclusion that the nodes of that network loses more energy caused by the actual time taken. Another disadvantage is that protocol affect the network throughput negatively if it takes longer time to send data.

An important note here is that these parameters is based on the weight formulas in order to get a normalized result. If someone else were to try this study in a smaller network (for example, smaller distance between nodes), they have to nor-malize the values for the weight formulas so they give similar costs for the same kind of transactions as in this study when energy levels are considered.

5.3

Related work

Kawadia and Kumart [6] discuss how to use long or short hops to transmit packets. Longer hops requires more transmission power and have a bigger risk of

(22)

interfer-more hops. They also present three different protocols that is used to control power and joint clustering that can be used by any other routing protocol. Li, Aslam and Rus [7] discuss and simulate similar model as the aforementioned. They create an own online algorithm used to route messages in a large network. This algorithm use zone-based power-aware routing to partition the network into smaller zones where each zone can evaluate their power quickly. They use these evaluations as a weight-parameter for the zones and this method achieved good results in terms of energy efficiency.

Networks can also reach a great level of energy savings by switching off the wire-less communication for nodes when they are not used. Doing it this way mean that a number of nodes in the network that must be active to maintain connectiv-ity while other nodes sleep. Cerpa and Estrin [1] uses a self-made algorithm for this where, for example, nodes only join the network when they are necessary to transmit data. This algorithm called Adaptive Self-Configuring Sensor Network Topologies (ASCENT) is shown to provide significant amount of energy savings. Others have presented how Connected to Dominant Set (CDS) approaches can be used for sleep management [10]. Here, the CDS acts as a topology backbone where all nodes stay active and nodes not in the CDS can sleep when necessary. Similar energy effectiveness studies as in this paper has been performed based on the on-demand routing protocol Dynamic Source Routing (DSR) [2]. Also they use a formula for energy-based routing computation but they add transmis-sion power control. They found that their modification was three times better than the original DSR. Also, geographic routing protocols such as GPSR and the recovery mechanisms thereof have been considered [3]. Based on power-aware ran-dom walks and power aware greedy routing, the authors form a new power aware geographic protocol. Their results show that their recovery mechanism gave an increase in delivery rates compared to power-aware greedy protocols without it. This improvement was at the cost of a small impact on energy efficiency. A Minimum Power Configuration (MPC) approach for wireless sensor networks are analyzed and discussed by Xing, Lu, Zhang, Huang and Pless [9]. In the paper they propose a protocol based on MPC, and use simulations to show that their modifications were significantly better than similar topology control and energy conserving protocols. Other general design challenges when creating an energy efficient ad hoc protocols, are presented by Goldsmith and Wicker [5]. They found that the biggest challenges when designing a protocol based on energy efficiency is how to handle limitations of node capabilities, topology variations and the lack of centralized control. Some solutions like node control, node redundancy, and adaptation at each protocol layer are also presented.

(23)

Chapter 6

Conclusions and Future

work

To relate our results with what we want to investigate we should answer the ques-tions from Section 1.3.

First, our results show that DSDV can be made more energy-efficient with the help of simple modifications to the protocol. The simulations of our modified policies showed that the simple weighted policy increased the lifetime by 104% measured in transactions compared to the shortest path policy, and the double weighted policy managed to increase the lifetime by 132%. With that as basis, it appears that a power-aware routing policy would be a much better choice. However, there are still many aspect we have not considered, causing us to raise new questions and therefore unable to make a definitive answer.

This study is in many ways different from a study with real scenarios. This is something that has both advantages and disadvantages that is presented in Sec-tion 5.1. Regardless of those differences, we can make the conclusion that our weighted policies are likely to perform better even in a real scenario. This is be-cause we have an awareness of why we can do the simplifications that we do and explain how they might impact end results. Even if the simulator itself does not work as it probably would in reality, it does not make that much difference. A real scenario could work just like this, it is just unlikely. The model would work similar even if we tried to make the simulation program more like reality and would give a similar end result. Therefore, these results are considered accurate enough to prove that the weighted policies presented in this thesis works better for energy efficiency.

The two created energy awareness polices uses two different weight formula, and that is the only weight formulas tested. With that said, we are not ready to answer how to the optimal weight formula to maximize the network lifetime would look like. A conclusion is that this subject is worth further investigation.

(24)

become more general. This means, it is highly possible to apply our energy effi-cient polices to other routing protocol. This is something we have not considered, but would be interesting to investigate further.

In the future, if we would have more time and resources on our hand, we could certainly investigate more aspects and reach a more concrete result. An interesting aspect is how the energy saving routing affects delivery time, throughput and if the network could suffer from congestion when there is more pressure on the static nodes. The model and simulation program could be developed further to reflect reality more. Work with the policies also remains. Testing several other weight formula was next to do if we had more time. Other aspects that related works have been studying that are interesting is how sleep management can be used and how longer or shorter hops affect end-to-end delay.

(25)

Bibliography

[1] A. Cerpa and D. Estrin. ASCENT: Adaptive Self-Configuring Sensor Net-works Topologies. IEEE INFOCOM, Volume 3:1278–1287, 2002.

[2] S. Doshi, S. Bhandare, and T. X. Brown. An On-demand Minimum Energy Routing Protocol for a Wireless Ad Hoc Network. Mobile Computing and Communications Review - (MCCR), Volume 6(No. 3):50–66, 2002.

[3] A. Dvir and N. Carlsson. Power-aware Recovery for Geographic Routing. IEEE Wireless Communications and Networking Conference (WCNC), pages 1–6, 2009.

[4] M. Frodigh, P. Johansson, and P. Larsson. Wireless ad hoc networking - The art of networking without a network. Ericsson Review, Volume 4:248–263, 2000.

[5] A. J. Goldsmith and S. B. Wicker. Design challenges for energy-constrained ad hoc wireless networks. IEEE Wireless Communications, Volume 8(No. 4):8–27, 2002.

[6] V. Kawadia and P. Kumart. Power Control and Clustering in Ad Hoc Net-works. IEEE INFOCOM, Volume 1:459–469, 2003.

[7] Q. Li, J. Aslam, and D. Rus. Online Power-aware Routing in Wireless Ad-hoc Networks . MobiCom Proceedings of the 7th annual international conference on Mobile computing and networking, pages 97–107, 2001.

[8] C. E. Perkins and P. Bhagwat. Highly Dynamic Destination-Sequenced Distance-Vector Routing (DSDV) for Mobile Computers. SIGCOMM Pro-ceedings of the conference on Communications architectures, protocols and applications, Volume 24(No. 4):234–244, 1994.

[9] G. Xing, C. Lu, Y. Zhang, Q. Huang, and R. Pless. Minimum Power Config-uration in Wireless Sensor Networks. Mobile Ad Hoc Networking and Com-puting - MobiHoc, pages 390–401, 2005.

[10] R. Zheng and R. Kravets. On-demand power management for ad hoc net-works. Ad Hoc Networks, Volume 3(No. 1):51–68, 2005.

(26)

Detta dokument hålls tillgängligt på Internet – eller dess framtida ersättare –

under en längre tid från publiceringsdatum under förutsättning att inga

extra-ordinära omständigheter uppstår.

Tillgång till dokumentet innebär tillstånd för var och en att läsa, ladda ner,

skriva ut enstaka kopior för enskilt bruk och att använda det oförändrat för

ickekommersiell forskning och för undervisning. Överföring av upphovsrätten

vid en senare tidpunkt kan inte upphäva detta tillstånd. All annan användning av

dokumentet kräver upphovsmannens medgivande. För att garantera äktheten,

säkerheten och tillgängligheten finns det lösningar av teknisk och administrativ

art.

Upphovsmannens ideella rätt innefattar rätt att bli nämnd som upphovsman i

den omfattning som god sed kräver vid användning av dokumentet på ovan

beskrivna sätt samt skydd mot att dokumentet ändras eller presenteras i sådan

form eller i sådant sammanhang som är kränkande för upphovsmannens litterära

eller konstnärliga anseende eller egenart.

För ytterligare information om Linköping University Electronic Press se

förlagets hemsida

http://www.ep.liu.se/

In English

The publishers will keep this document online on the Internet - or its possible

replacement - for a considerable time from the date of publication barring

exceptional circumstances.

The online availability of the document implies a permanent permission for

anyone to read, to download, to print out single copies for your own use and to

use it unchanged for any non-commercial research and educational purpose.

Subsequent transfers of copyright cannot revoke this permission. All other uses

of the document are conditional on the consent of the copyright owner. The

publisher has taken technical and administrative measures to assure authenticity,

security and accessibility.

According to intellectual property law the author has the right to be

mentioned when his/her work is accessed as described above and to be protected

against infringement.

For additional information about the Linköping University Electronic Press

and its procedures for publication and for assurance of document integrity,

please refer to its WWW home page:

http://www.ep.liu.se/

References

Related documents

To visualize the energy and cost saving capacity of 5G wireless networks in the rural area by comparison, we designed Case 2, in which we investigate the possibility of LTE

When it comes to security, there is an obvious trade-off. We get more protection if we have endless energy. In this paper we show that we can adapt the sensitivity of

The metaphor of network has been used as a perspective to capture the links between the local learning centres organisation and the actors in the context of

Maximized power transfer from the source output to the load input, enables the load to save excess energy in a battery or similar, while high efficiency from the load input to

4 Power management methods in computer graphics software 25 4.1 Energy efficient graphics

Med egen lagring skulle Väla Gård endast kunna köpa in ett 1 kWh batteri för att uppnå en viss lönsamhet med dagens batteripriser.. Solproduktionens

By working to facilitate the use of small scale power systems using renewable energy sources in general, and PV-diesel hybrid micro-grids in particular, the project

The most significant difference is that Circuitus has better heat exchanger and building envelope; lower U-value and better airtightness which results to better energy performance