To assure load-balance and energy efficiency, EGR enhances traditional ... Keywords: Geographic Routing, Ad-hoc Wireless Networks, Energy Proportional.
國立成功大學 資訊工程學系 碩士論文
通用於無基礎架構網路利用地理資訊 之有效能源利用路由 Generic Energy-efficient Geographic Routing for Ad-Hoc Wireless Networks
研 究 生：陳易聰 指導教授：郭耀煌 博士
中華民國九十六年八月
II
III
通用於無基礎架構網路利用地理資訊 之有效能源利用路由 陳易聰*
郭耀煌**
國立成功大學資訊工程學系碩士班 摘要 由於在無線網路的環境中，節點只有有限的能源，如何節省能源使用一直是重 要的議題。本論文中，我們提出一個可用在所有無線網路上，有效減少能源消耗的 路由方法，稱為 EGR（Energy-efficient Geographic Routing）。傳統路由演算法產生 原路徑後，EGR 在原路徑中的每一段路徑中找適合的轉傳點，以達到節省能源的目 的；另一方面，也考量了平衡能源使用以延長整個網路的可使用時間。而在每一段 路徑中，使用無線電傳輸模組可計算出傳輸與接受資料所耗的能源，從而找到能代 為轉傳資料的轉傳區域。在這個轉傳區域中的節點都能作為轉傳點，可節省整體的 能源消耗；再考慮 Energy-Proportional Principle (EPP) 後，會達到平衡能源使用的目 標。 EGR 可以輕易的和傳統路由演算法作結合，我們針對兩大類型的傳統路由演算 法作了修改，其它路由演算法亦可在每一段路徑中考量 EGR 以達到減小能源消耗。 最後模擬與分析結果，發現 EGR 的確有效的節省能源使用，也增長了網路的可使用 時間。
*作者
**指導教授
關鍵字：利用地理資訊的路由, 無基礎架構網路, 比例式能源使用原則, 平衡能源使 用
I
Generic Energy-efficient Geographic Routing for Ad-Hoc Wireless Networks Yi-Tsung Chen* Yau-Hwang Kuo** Department of Computer Science and Information Engineering, National Cheng Kung University, Tainan, Taiwan, R.O.C.
Abstract We propose an energy-efficient geographic routing (EGR) mechanism which is generally applicable to reduce energy consumption in wireless communication networks. To assure load-balance and energy efficiency, EGR enhances traditional ad-hoc routing algorithms by constructing an initial routing path considering location information. Then, to further improve energy utilization it selects relay nodes of links on the initial path. The EGR finds an optimum relay node in a relay region between any two traffic nodes to conserve energy and balance traffic load. The relay region is derived from the radio propagation model constraining energy-saving when relaying transmissions between two nodes. Any node within this region is a relaying candidate to decrease total traffic energy consumption and to balance traffic load. According to the Energy-Proportional Principle (EPP), we propose an energy-saving criterion. To balance traffic load, the EGR follows the EPP and in the relay region selects the relay node with the highest score corresponding to the criterion. Compared to the traditional routing methods, EGR effectively utilizes energy and prolongs network lifetime.
*Author
**Advisor
Keywords: Geographic Routing, Ad-hoc Wireless Networks, Energy Proportional Principle, Load Balance
II
誌謝 在成大的這兩年，在學術研究上有很大的進步，這得感謝郭耀煌老師的諄諄教 誨及以國際實驗室自許為目標，除了能深入自己所研究的領域，也由每次的實驗室 討論會中涉獵他人研究領域，了解許多解決問題的方法。在實驗室裡，除了有計劃 要完成，也接觸了理論的研究，碩士生活可謂紮實，我很幸運能接受郭耀煌老師的 指導。 也感謝陳朝烈學長和李振維學長在我的研究期間一直不斷的給予指導與修正 論文研究方向，使我最後能順利克服論文研究中的種種問題，除此之外，亦師亦友 的陳朝烈學長也教了我許多人生該持有的正確態度，使我獲益良多。 在實驗中一起打拚的同伴們讓我的研究生活不會那麼的枯燥無聊，感謝和我同 窗的同學們，帶給我歡樂的兩年碩士生活，很高興大家都有相同的興趣，能常常一 起出去唱歌同樂，每年大家的生日會也讓我們吃了不少台南的美食。 最後，我要感謝我的父母，讓我在物質生活上無後顧之憂；並時常支持我鼓勵 我，在我研究陷入低潮時為我打氣，論文一路走來並不順利，感謝父母為我的精神 支柱，讓我能完成碩士學位，謹以此篇論文獻給他們。
陳易聰 96/8
III
Content Abstract ............................................................................................................................... II Content ...............................................................................................................................IV Figure List ..........................................................................................................................VI Chapter 1.
Introduction ................................................................................................. 1
Chapter 2.
Related Works.............................................................................................. 4
2.1.
Energy-Proportional Routing (EPR) ..................................................................... 4
2.2.
Ad Hoc On-Demand Distance Vector (AODV) Routing Protocol...................... 10
2.3.
Destination Sequenced Distance Vector (DSDV) Routing Protocol ................... 11
2.4.
Geographical Adaptive Fidelity (GAF) ............................................................... 12
Chapter 3.
Energy of Radio Model and Relay Constraint........................................ 16
3.1.
Energy Consumption and Propagation Model..................................................... 16
3.2.
Relay Region ....................................................................................................... 17
3.3.
Optimal number of Relay Nodes ......................................................................... 20
3.3.1.
Formulation ................................................................................................. 20
3.3.2.
Analysis ....................................................................................................... 22
Chapter 4.
Construction of energy-efficient geographic routing (EGR)................. 25
4.1.
Selection of Relay Nodes .................................................................................... 25
4.2.
Ad-hoc routing with location information........................................................... 26
4.2.1.
Modification of AODV................................................................................ 26
4.2.2.
Modification of DSDV ................................................................................ 28
IV
4.3.
Generic sleeping scheduler.................................................................................. 30
Chapter 5.
Simulation and Comparisons ................................................................... 32
5.1.
EGR energy saving.............................................................................................. 32
5.2.
Enhancing Energy Utilization and Lifetime in GAF........................................... 35
5.2.1.
Traffic and mobility models ........................................................................ 35
5.2.2.
Simulation results ........................................................................................ 35
Chapter 6.
Analysis....................................................................................................... 47
6.1.
Energy consumption model for each node over GAF ......................................... 47
6.2.
Energy consumption model for each node with using EGR ............................... 49
6.3.
Simulation and comparisons................................................................................ 50
Chapter 7.
Conclusion and Future Works ................................................................. 54
References .......................................................................................................................... 55
V
Figure List Figure 2-1. Sleeping scheduler machine state of GAF .................................................................................... 14 Figure 3-1. First Order Radio Model............................................................................................................... 16 Figure 3-2. Traffic from A to B........................................................................................................................ 17 Figure 3-3. The contour of ES .......................................................................................................................... 19 Figure 3-4. Saved energy ES on the geographical plane .................................................................................. 20 Figure 3-5. When dmax is 250 metes, the number of relay nodes causes different total energy consumption 22 Figure 3-6. NMIN(E) with different dmax.............................................................................................................. 23 Figure 3-7. Nopt with different dmax .................................................................................................................. 24 Figure 4-1. Selecting relay node according to its relay score .......................................................................... 26 Figure 4-2. Flow charts of modified AODV adopting EGR............................................................................ 27 Figure 4-3. Flow charts of modified DSDV adopting EGR ............................................................................ 30 Figure 4-4. Sleeping scheduler state machine of EGR .................................................................................... 31 Figure 5-1. The energy consumption ratio of non-EGR to EGR with various dmax ......................................... 33 Figure 5-2. Comparison of the energy consumption between DSDV, AODV and EGR. ................................ 34 Figure 5-3. Simulation results of GAF-b and EGR for network lifetime with 900 pause time ..................... 36 Figure 5-4. Simulation results of GAF-b and EGR for network lifetime with 600 pause time ..................... 37 Figure 5-5. Simulation results of GAF-b and EGR for network lifetime with 300 pause time ..................... 37 Figure 5-6. Simulation results of GAF-b and EGR for network lifetime with 120 pause time ..................... 38 Figure 5-7. Simulation results of GAF-b and EGR for network lifetime with 60 pause time ....................... 38 Figure 5-8. Simulation results of GAF-b and EGR for network lifetime with 30 pause time ....................... 39 Figure 5-9. Simulation results of GAF-b and EGR for network lifetime with 0 pause time ......................... 39 Figure 5-10. Simulation results of GAF-b and EGR for average delay........................................................... 40
VI
Figure 5-11. Simulation results of GAF-b and EGR for average data delivery ratio....................................... 40 Figure 5-12. Simulation results of GAF-b and EGR for average data received .............................................. 41 Figure 5-13. Simulation results of GAF-ma and EGR for network lifetime with 900 pause time ................ 42 Figure 5-14. Simulation results of GAF-ma and EGR for network lifetime with 600 pause time ................ 42 Figure 5-15. Simulation results of GAF-ma and EGR for network lifetime with 300 pause time ................ 43 Figure 5-16. Simulation results of GAF-ma and EGR for network lifetime with 120 pause time ................ 43 Figure 5-17. Simulation results of GAF-ma and EGR for network lifetime with 60 pause time .................. 44 Figure 5-18. Simulation results of GAF-ma and EGR for network lifetime with 30 pause time .................. 44 Figure 5-19. Simulation results of GAF-ma and EGR for network lifetime with 0 pause time .................... 45 Figure 5-20. Simulation results of GAF-ma and EGR for average delay........................................................ 45 Figure 5-21. Simulation results of GAF-ma and EGR for average data delivery ratio ................................... 46 Figure 5-22. Simulation results of GAF-ma and EGR for average data received ........................................... 46 Figure 6-1. The transition of each node’s state................................................................................................ 47 Figure 6-2. The transition of each node’s state................................................................................................ 51 Figure 6-3. The transition of each node’s state................................................................................................ 52 Figure 6-5. The transition of each node’s state................................................................................................ 53
VII
Chapter 1.
Introduction
The classification of ad-hoc routing algorithms includes on-demand and table-lookup driven classes. The most representative routing protocols of these two classes are the Ad hoc On-demand Distance Vector (AODV) routing [1] and Destination-Sequenced Distance Vector (DSDV) routing [2] respectively. In AODV, nodes use the maximum transmission range to communicate with each other, but nodes can adaptively adjust their power level for conserving transmitting energy by the Transmit Power Control (TPC) [3]. Lower power level means reduced interference problems and increased energy utilization. Namely, an energy-efficient routing protocol should be able to control transmission power dynamically. For load sharing and energy-saving, two communication nodes should dynamically tune down transmitting power when there are suitable relay nodes for the communication. The same situation happens in table-lookup driven routing algorithms such as DSDV. The routing table construction should consider the load sharing and energy saving especially when the location information is available. In this thesis we propose the Energy-efficient Geographic Routing (EGR) algorithm which is generally applicable to reduce energy consumption in wireless communication
1
networks. The proposed EGR improves energy efficiency and load balance for both classes of ad-hoc routing algorithms. While a traditional routing protocol finds the routes in the path, EGR uses location information for initial path construction and then finds more route nodes in the initial routing path to minimize the total energy consumption. Since many later algorithms are developed from the two classes of ad-hoc routing, the EGR is a generic energy-efficient mechanism that is applicable to wireless networks. We illustrate the geographical routing as examples. To extend the lifetime of the ad-hoc wireless networks, many articles proposed geographic routing as LAR [4] and GEAR [5] use location information to find a better routing path for saving transmission energy. Further, using energy-aware routing protocols like GAF [6] and SPAN [7] can save more energy consumption. The main idea is that they choose a node in a region as a coordinator to forward data, and other nodes go to sleep. Among them, Geographical Adaptive Fidelity (GAF) is one of the most representative routing algorithms that effectively use geographic information for coordinator selections and sleep-time scheduling for energy conservation. GAF divides the communication area into geographic grids. Though the grids ensure that all the nodes in a grid square are able to connect other nodes in any adjacent grid square, energy constraint in radio propagation is not considered. Therefore, any communication between two grids could violate the constraint and degrade performance in energy conservation. As an application, the
2
proposed EGR improves energy conservation in wireless ad-hoc networks by constraining relay nodes selection while at the same time balances traffic load by applying the energy-proportional principle (EPP) in the relaying. The EPP originates from the energy-proportional routing [8][9], which effectively balance intra- and inter-cluster energy utilization and thus extend lifetime of clustered sensor networks. The EPP considers the total energy-proportional balance among nodes, rather than either merely simple balance of energy consumption or communication distance. In this way, the EGR effectively utilizes energy, balances the load, and thus prolongs network lifetime. The thesis is organized as follows. Chapter 2 includes related works for EGR. In Chapter 3, we reduce the relay inequality to a circular relay region and analyze the optimum number of the relay node. Chapter 4 depicts the construction process of EGR algorithm. Chapter 5 shows the experimental results of energy utilization. Finally, we give conclusions in Chapter 7.
3
Chapter 2.
Related Works
2.1. Energy-Proportional Routing (EPR) In order to understand the relation between the residual energy of a sensor node and the cost (the dissipated energy from the sensor node to the other sensor nodes), EPR defines data capacity to show this relation in Equation (2-1) It actually shows the maximal amount of data that can be forwarded. Di j ,total =
Ei Ei = ETx (1, dij ) α t + α amp d ij2
(2-1)
where Dij,total is the maximum amount of data to be transmitted from node i to node j, Ei is the remaining energy in Joules of node i, and ETx(1,dij) is the additional estimated energy dissipation per bit when the bit is transmitted from node i to node j. According to the radio
frequency
transmitting
model
for
energy
widely
accepted
in
[10][11][12][13][14][15] and [16] transmitting k bytes across distance d will consume ETx(k,d)=kαt+kαampd2. To predict sensing data amount rather than to predict energy consumption such as in [17] of a node in each round, EPR applies the Markov model to save computation overhead since it is much simpler. The computation of prediction itself dissipates energy. Therefore, low complexity is an important requirement for the prediction. The proposed
4
prediction steps are as follows: First, EPR divides the period of a round into T phases each of which is identified with a time stamp. Fro one time-stamp to the next one, the state of a sensor node will change (state i changes into state s). During a round, a node’s state transits at each time stamp t. By using the Chapman Kolmogorov equations [18], the t-step transition probabilities are represented as M
Pis(t ) = ∑ Pij( r ) Pjs(t − r ) , 0 < t < T
(2-2)
j =1
Therefore, EPR calculates the summation of all probability over T phases to obtain the number of time steps that a node stays in state s in a round. Suppose a node at state s each time transmits collected ks-byte of data to its cluster head. EPR obtains the total predicted data amount in a round that a node n at state s transmits as: T
Dn( s ) = k s ∑ Pis( t )
(2-3)
t =1
Next, to obtain the predicted data amount of a node, EPR summates the values Dn( s ) for all state s in 1, 2, K, M. Suppose a cluster is composed of N sensor nodes, the predicted data amount of a cluster is as (2-4). N
M
Di , predict = ∑∑ Dn( s )
(2-4)
n =1 s =1
Here EPR assumes that there are four states (M = 4) in the proposed algorithm. The four states and their respective ks’s are defined as follows: State 1: sensing off, radio off, ks = 0;
5
State 2: sensing on, radio off, ks = 0; State 3: sensing on, radio receiving, ks = 0; State 4: sensing on, radio transmitting, ks = k. These states are sufficient and necessary for predicting the amount of data to be transmitted. In inter-cluster communication, only in State 4 can a cluster transmit data to an intermediate cluster head or base station while in the other states it keeps sensing or receiving. When entering State 4, the data packet of length k bytes is sent. Note that the probabilities of transitions among states depend on the past history of sensor nodes. Consequently, the probabilities are updated per round in order to be adaptive to the historical statistics. EPR updates the probability, used by (2-4), in the beginning phase of each round. EPR finds that the prediction method is very important. The performance of the proposed algorithm depends on the method of prediction. If the method of prediction does not precisely predict the future amount of data for transmission in each node, energy saving is limited. However a trade-off arises in that the time complexity of the prediction algorithm also affects the lifetime of sensor network. While calculating a new routing path, all nodes must be awake and they must receive topology information from the base station within each period. If EPR executes complex computations in order to acquire precise predictions, it consumes more energy. Therefore, EPR adopts a simple and widely
6
used method - the Markov model to predict the data transmissions. There are, of course, other precise prediction algorithms. For example, fuzzy algorithms, generic algorithms, and neural network algorithms are all good candidates for predictions. However, they are not suitable for sensor networks because they waste too much time in learning and obtaining rules. Before introducing the energy-proportion threshold, EPR introduces how to define the utilization ratio of sensor nodes in order to observe the conditions of all sensor nodes including cluster heads while working. The equation for the utilization ratio is as follows: U ij =
Di , predict Di j ,total
=
Di , predict (α t + α amp dij2 )
(2-5)
Ei
where Di,predict is the predicted amount of data of the node i obtained in (2-4), and Di,totalt obtained in (2-1) is the capacity from node i to node j. EPR defines a threshold value to determine whether a node is energy proportionally balanced or not. Equation (2-6) is for calculating the threshold. N
ωth =
∑U i =1 N
Ei
i0
(2-6)
∑E i =1
i
where Ei is the residual energy of the i-th node, and Ui0 is the utilization ratio of transmission from the i-th node to its destination such as the base station. By (2-1)-(2-5), the threshold value of a round is the proportion of total predicted energy consumption in
7
the total remained energy. That is, N
ωth =
∑E i =1
i , predicted
(2-7)
N
∑E i =1
i
where summation of Ei,predicted is the total predicted energy consumption for transmission and receiving (if with relaying). In fact, the threshold is the average energy utilization proportion among sensor nodes when accounting for intra-cluster communication. When accounting for inter-cluster communication, it is the average energy utilization proportion among cluster heads. Therefore, EPR arrives at the following trivial proposition to explain how the predicted data amount and energy consumption relates to distances.
Proposition 1: The total predicted energy of N-cluster sensor network in a round is the product of the energy proportional threshold and the total energy remaining. That is, N
∑ D (α i =1
i
N
t
+ α amp di20 ) = ωth ∑ Ei
(2-8)
i =1
where Di is the predicted data amount. Note that this proposition also holds when EPR considers the intra-cluster communication energy consumption in an N-sensor cluster. An example for determining and applying the threshold in inter-cluster communication is as follows. Suppose there is cluster heads A, B, C, D and E with a common destination node O such as the base station. Suppose applying clustering and routing that the utilization ratios of cluster heads A, B, C, D, E are 22.22%, 62.5%,
8
13.95%, 28.31%, and 25% respectively, while the residual energy of cluster heads A, B, C, D, E are 2 joules, 1.5 joules, 2 joules, 0.5 joules, and 1 joule respectively. We then have 29.3% as the mean utilization of energy by using (2-6). Therefore, for load balancing, cluster head B should neither directly send its aggregation to the base station nor serves a confluent node for forwarding since its utilization ratio is much higher than the threshold. The cluster heads A, C, D, and E who have lower utilization ratios than the threshold 29.3% are candidates for sharing load of cluster head B. The proposed EPR algorithm then route transmission paths of the network by assigning A, C, D, and E as intermediate nodes that forward data from node B. Hence, the utilization ratios of A, C, D, and E will increase while B’s will decrease until they are very close to ωth 29.3%. Since the EPR considers the total energy-proportional balance among clusters, rather than either merely simple balance of energy consumption or communication distance, the network lifetime is extended even in scenarios with quite different parameters. For this example, cluster B will not quickly die out of energy and the lifetime of the network is extended no matter how data amounts and residual energy are distributed. For intra cluster communication, if we replace the role of cluster heads A, B, C, D, and E with common sensor nodes in a cluster and to replace the base station as the cluster head, this example also show the application of the EPR in intra-cluster communication. EPR has two ideas we depend on it when we develop EGR. One is using relay to
9
reduce total energy consumption and the other is the energy proportional concept is used for balance the residual energy and extend the network lifetime further.
2.2. Ad Hoc On-Demand Distance Vector (AODV) Routing Protocol AODV is a routing protocol designed for ad hoc mobile networks. It is an on demand algorithm, meaning that it builds routes between nodes only as desired by source nodes. It maintains these routes as long as they are needed by the sources. AODV uses sequence numbers to ensure the freshness of routes. It is loop-free, self-starting, and scales to large numbers of mobile nodes. AODV builds routes using a route request / route reply query cycle. When a source node desires a route to a destination for which it does not already have a route, it broadcasts a route request (RREQ) packet across the network. Nodes receiving this packet update their information for the source node and set up backwards pointers to the source node in the route tables. In addition to the source node's IP address, current sequence number, and broadcast ID, the RREQ also contains the most recent sequence number for the destination of which the source node is aware. A node receiving the RREQ may send a route reply (RREP) if it is either the destination or if it has a route to the destination with corresponding sequence number greater than or equal to that contained in the RREQ. If this is the case, it unicasts a RREP back to the source.
10
Otherwise, it rebroadcasts the RREQ. Nodes keep track of the RREQ's source IP address and broadcast ID. If they receive a RREQ which they have already processed, they discard the RREQ and do not forward it. As the RREP propagates back to the source, nodes set up forward pointers to the destination. Once the source node receives the RREP, it may begin to forward data packets to the destination. If the source later receives a RREP containing a greater sequence number or contains the same sequence number with a smaller hopcount, it may update its routing information for that destination and begin using the better route. If a link break occurs while the route is active, the node upstream of the break propagates a route error (RERR) message to the source node to inform it of the now unreachable destination(s). After receiving the RERR, if the source node still desires the route, it can reinitiate route discovery.
2.3. Destination Sequenced Distance Vector (DSDV) Routing Protocol DSDV is a table-driven algorithm based on modification made to the Bellman-Ford routing mechanism. Each node in the network maintains a routing table that has entries for each of the destination in the network and the number of hops required to reach each of them. Each of the entry has a sequence number that helps in identifying stale entries, thus avoiding formatting of routing loops. Each node periodically sends updates tagged
11
through out the network with a monotonically increasing even sequence number to advertise its location. New route broadcasts contain the address of the destination, the number of hops to reach the destination, the sequence number of the information received regarding the destination, as well as a new sequence number unique to the broadcast. The route labeled with the most recent sequence number is always used. Once its neighbors receive this update, they recognize that they are one hop away from the transmitting node and include this information in their distance vectors. Every node stores the "next routing hop" for every reachable destination in their routing table. The route used is the one with the highest sequence number, i.e., the most recent one. When a node A detects a broken link to another node B, it will advertise a sequence number one greater than what it had advertised for that route before with a hop count metric set to infinity. When other nodes hear this message they will update their routing tables and stop using this broken link until they hear an advertisement from B with a higher sequence number.
2.4. Geographical Adaptive Fidelity (GAF) GAF is an energy efficient protocol in mobile ad hoc networks because it elects the proper coordinator between mobile nodes based on position information. Nodes which are data sources or sinks do not participate in the GAF procedure. Sleeping nodes also do not buffer pending traffic.
12
GAF divides the simulation area in a lot of geographic grid. If the transmission range is R and the edge length of grid is r, then r must less or equal than R / 5 . It is because GAF has to ensure that all the nodes in a grid square are equivalent with respect to providing connectivity to any adjacent grid square. One waking node (coordinator) in each grid is supposed to maintain the connectivity of the original network. Because GAF elects a coordinator for each grid square and requires no explicit exchange of connectivity information. Each node in the same grid transits independently among three states: sleep, discovery, and active such as Figure. When nodes wake up from sleep state, they transit themselves to discovery state. Nodes in the discovery state listen to other nodes' announcements and announce its own grid id and expected node active time. If the node hears no “higher ranking” announcement during a period of time, it transits to the active state. Or it come back to sleep. A coordinator in the active state wants to maintain network connectivity between grids must periodically announce its state so that it can tell the other nodes to sleep in the same grid. When a node has been a coordinator for a while, it transits back to the discovery state, letting the other nodes be a coordinator in the same grid square.
13
lower rank nodes
sleeping
lower rank nodes
active
after TS after Td after Ta
discovery
Figure 2-1. Sleeping scheduler machine state of GAF
The goal of the ranking function is used to balance the residual energy among all nodes in the same grid square. The ranking function is based on the residual energy of nodes and length of time it is projected to remain in its current grid square. Therefore, a node has more residual energy, it has higher ranking. The sleep intervals are set according to the random number between fourth to half of expected node active time in order to transit from the sleep state to the discovery state in time to replace an active node, if needed. Now GAF runs independently with mobile ad hoc routing protocol. It makes that the mobile ad hoc networks still have connectivity when the active node in grid square changes state on GAF. Because routing protocol interprets this situation as a route failure from which it must recover. The effect of GAF on the energy consumption and performance of AODV routing
14
protocol has been studied in ns2 [19] packet level simulation. In [6], we have seen that AODV and AODV/GAF show the similar performance about data delivery ratios and transfer delays. But the mean energy consumption per node in AODV/GAF has reduced about 40-50% than AODV. GAF also extends proportional network lifetime in increasing node density (defined as 80% data delivery ratio). Similar results were obtained using DSR [20] as the routing protocol. GAF needs to select an appropriate value of grid size in order to keep the connectivity. The announcement mechanism in GAF can adjust the connectivity in a grid square. If the grid size is not selected appropriately, it may cause that some nodes in the grid square may not receive an announcement and the connectivity is poor. Therefore, energy consumption is increased. We use GAF as a sleeping scheduler implemented in networks for energy saving. Over GAF, traditional on-demand routing protocols can operate well. EGR enhances energy saving over these routing protocols. We simulate GAF in NS2 and show the comparison to EGR.
15
Chapter 3.
Energy of Radio Model and Relay Constraint
3.1. Energy Consumption and Propagation Model In this thesis, we apply the first order radio model as Figure 3-1 commonly used in low-energy radios.
Figure 3-1. First Order Radio Model
When two nodes are d meters apart and sender transmits k bits data to receiver, the energy consumption can be calculated as follows [21]： ETx ( k , d ) = ETx ,elec × k + ETx , amp × k
(3-1)
ERx ( k ) = ERx ,elec × k
The radio dissipates ETx,elec or ERx,elec in transmitting or receiving one bit data. ETx,amp is depleted in amplifier for transmitting data, and determined by crossover distance (d0). At the crossover distance, the power for receiving predicted by the two-ray ground (TR) reflection model equals to that predicted by the free-space (FS) propagation model. If the
16
transmitter is within the crossover range, using the FS model is appropriate. Otherwise, use the TR model. That is, ⎧⎪ε FS × d 2 , when d ≤ d 0 ETx ,amp ( d ) = ⎨ 4 ⎪⎩ε TR × d , when d > d 0
(3-2)
where εFS and εTR are the respective amplifier parameters in FS and TR models. In our simulation, we set the communication energy parameters as: ETx,elec = ERx,elec = 50 nJ/bit,
εFS =100 pJ/bit/m2, εTR = 0.013 pJ/bit/m4 and d0= ε FS / ε TR . Moreover, we let αt = ETx,elec = ERx,elec and αamp = εFS.
3.2. Relay Region
Figure 3-2. Traffic from A to B
Considering a simple illustration shown in Figure 3-2, suppose that there are three nodes A, B and C. Assume all nodes use the same circuitry for transmission and receiving. The source node A sends data to the destination node B. For simplification, node A is located at the origin and B with coordinate (d, 0) is d meters apart from node A. The
17
coordinate of relay node C is set to (x, y). Assume C is p meters apart from A and q meters apart from B. For power saving, node C must be in a region to satisfy the following inequality (we call the region relay region):
(α
t
+ α amp d 2 ) + α t > (α t + α amp p 2 ) + (α t + α t + α amp q 2 ) + α t
(3-3)
where the right hand side is the energy dissipation spent by the three nodes when using C for relay and the left hand side is the energy dissipation spent by the node A and node B when A send data directly to B. Substituting the coordinates into distances d, p, and q, we have p2 + q2 < d 2 −
2α t
α amp
2α t 2 ⇒ ( x 2 + y 2 ) + ⎡( d − x ) + y 2 ⎤ < d 2 − ⎣ ⎦ α
amp
Thus, we have the relay inequality (3-4) be reduced to a circular relay region. x2 − d × x + y 2 +
αt
