Minimum Energy Based Efficient Routing Protocol Over MANETs

Minimum Energy Based Efficient Routing Protocol Over MANETs

D. Venkata Siva Prasad, M. Tech Student, and Mr. D. Sharath Babu Rao, Faculty

Depart ment of Electronics & Co mmunication Engineering Jawaharla l Nehru Technological Un iversity

Anantapur, India

ABSTRACT

Minimum energy (Energy Efficient) routing protocols are very essential for wireless ad hoc networks which usually consist of mobile battery operated computing devices and many such protocols and schemes have been proposed so far. However, few efforts have been spent on issues associated with such protocols such as routing overhead and route setup time and route maintenance. The energy efficient routing protocols could fail without considering the mobility of node and routing overhead involved. In this paper, a more accurate analytical model is proposed to track the energy consumptions due to various factors and impact of packet errors. A simple energy efficient routing scheme called PEER is implemented that significantly improves the performance during path discovery phase and in mobility scenarios. The simulation results show that PEER protocol can reduce up to 2/3 routing overhead during path discovery phase and delay, and 50 percentage transmission energy consumption compared to conventional energy efficient routing protocol.

1. INTRODUCTION

   In wire less ad hoc networks, mobile devices are often battery powered. But current battery technique still could not support the devices to work long enough. In addition, changing the battery may not be feasible in some application scenarios, such as sensor networks in hostile environ ment. There fore, energy saving schemes are very impo rtant in wireless ad hoc networks. Since mobile devices are getting smaller and mo re energy efficient, co mmun ication energy cost becomes a much significant part in the total energy consumed. Therefore energy effic ient commun ication scheme is one of the most effective ways to save energy.

   In wireless network, the trans mitted signa1 is attenuated at the rate of 1/ , where d is the distance to the sender and n is the path loss exponent between 2 and

   1. Then the basic energy efficient scheme would be to adjust the transmission power according to the distance between the sender and the receiver instead of using the constant ma ximu m transmission power. This is called power control scheme. However, this is not optimal in terms of end-to-end energy consumption. To achieve the optima l solution, many energy effic ient routing protocols have been proposed [1]-[8]. These protocols can be generally c lassified into two categories: Minimum Energy routing protocols[l]-[6] and Maximizing Network Lifetime routing protocols [7][8]. Minimum Energy routing protocols try to find the most energy efficient path to transmit the data packets fro m the source to the destination, wh ile Maximizing Network Lifetime routing protocols try to balance the re main ing battery power at each node .

      Minimum Energy routing protocols can be further div ided into three c lasses based on the types of

      lin k costs: Minimum Total Transmission Power (MTTP), Minimum Total TransCeiving Power (MTTCP), and Minimum Total Reliable Transmission Power (MTRTP) protocols. MTTP protocols use the transmission power as the link metric and search for the path with minimu m total transmission power between the source and the destination. MTTCP protocols use the transmission power as well as the receiving power as the link cost. MTRTP protocol uses the total transmission power for transmitting the data packets from one node to its neighboring node reliably as the link cost.

      Most of previous work concentrated on the lin k costs. Once a new lin k cost was derived, then the traditional shortest path routing protocols, such as AODV, DSR, and Be llman-ford, can be modified with the new link cost. However, there a re some proble ms with such straightforward modification. First, the routing overhead for the route discovery is very high, which consumes a lot of energy. Second, the route setup time is very long. Third, the route ma intenance scheme is not suitable for dynamic environ ments, such as mobility scenarios.

      To address these issues, we propose a Progressive Energy Efficient Routing (PEER) protocol. Contrary to other energy effic ient routing protocols that try to find the optimal path at one shot and maintain the route reactively, PEER searches for the mo re energy efficient path progressively and maintains the route continuously. It first finds a path near the most energy efficient path between the source and the destination quickly, and then adjusts the nodes whenever necessary so that the path would be energy effic ient all the time . Our performance evaluation shows that PEER achieves less routing overhead, shorter setup time, and great energy effic iency in static scenario as well as the mobile scenario.

2. OBSERVATION AND MOTIVATION

   Many routing protocols have been proposed for wireless ad hoc networks. These protocols can be generally categorized as: (a) table -driven, (b) on-de mand, and (c) hybrid. For table driven routing protocols, all nodes need to advertise the routing information periodically so that they can have the up -to-date view of the network. Destination Sequenced Distance Vector (DSDV), Wireless Routing Protocol (WRP), and Cluster Switch Gateway Routing (CSGR) belong to this category. Diffe rent fro m table-driven routing protocols, on-demand routing protocols create the route only when desired by the source node. Some on-de mand routing protocols are Ad hoc On-demand Distance Vector (AODV). Dynamic Source Routing (DSR), and Temporally Ordered Routing Algorithm (TORA). The Zone Routing Protocol (ZRP) is

   a hybrid protocol with table-driven routing scheme for the intra-zone routing and on-demand routing scheme for the inter-zone routing. Most of energy effic ient schemes modified the on-demand routing protocols such as AODV or DSR since there is a lot of routing overhead if using table-driven routing protocols . So, we will only focus on the on-demand energy effic ient routing protocols.

   Fig 1: A Linear Topology

   For on-demand routing protocols such as AODV, a node will start a route discovery process if it needs a route to a destination. It broadcasts the route request packet and waits for the reply from the destination. The neighboring nodes that receive such route request packet will rebroadcast it, and so on. To reduce the routing overhead, the nodes will only rebroadcast the first route request packet received and discard the following duplicate ones. And the destination node only replies to the first route request packet, too. For e xa mple , in Fig 1, both A and B are neighboring nodes of S and D, and S needs a route to D. So S b roadcasts the route request packet first, and both A and B receive the packet. Assume A broadcasts such packet next, then node S, B and D receive such packet, however node S and B will discard it as they have already received the same route request packet. Therefore the final route is SAD. It is apparent that the routing overhead for these protocols is O(n), where n is the nu mber o f nodes in the network. Things are quite different for energy effic ient routing protocols. The nodes could not simp ly d iscard the duplicate route request packets now as they may come fro m mo re energy effic ient paths. That is, they also need to respond to the route request packets fro m a mo re energy effic ient path.

   Therefore, the nodes may need to b roadcast the same route request packet many times. For the same e xa mple in Fig. 1, node B may need to broadcast both the packets fro m S and A if the path SAB is mo re energy efficient than SB. Based on the Be llman- Ford a lgorith m , we can obtain that routing overhead for minimu m energy efficient routing protocols is ) now. Such overhead will consume a lot of energy and network resources, especially when the number o f nodes in the network is very large.

   In addition, the route setup time is much longer than the on-demand routing protocols. There are two ma in reasons for this. One is that the energy efficient route has more intermediate nodes than the shortest path in general, so it ta kes longer time for the route request and route reply packets to go through all the intermediate nodes. The other is that the energy efficient routing protocols have much more routing overhead wh ich can cause more delay at each link. The simulat ions in Glo moSim verify our observation.

   Fro m the simu lation results in Fig. 5-7, it is clea r that the routing overhead, energy consumption for routing overhead, and route setup time for the energy efficient routing protocol increase dra matica lly with the nu mber of nodes in the network, while only linearly for the on- demand routing protocol.

3. ENERGY CONSUMPTION MODEL FOR 802.11

   Link c ost is very important in energy efficient routing protocols. Without an accurate lin k cost the minimu m energy routing protocols could not find the optima l route. In this section, we will first present some physical and MAC layer assumptions used in this paper. Then we propose an efficient way to estimate the lin k cost. PEER requires that each node can adjust the transmission power dynamically and retrieve channel informat ion such as noise and received power level. Both are also common assumptions in most energy efficient routing protocols. In addition, it also desires that the MAC protocol can p rovide re liable hop-by-hop data transmission as retransmission costs a lot of energy. Therefore we use power control 802.11 for MA C protocol, in which RTS and CTS packets are transmitted at the ma ximu m powe r while DATA and ACK packets are transmitted at the minimu m required power level for the receiver to decode correctly. To avoid some collisions, PEER also requires the nodes to set their NA Vs (Network Allocation Vector) to the EIF (Extended InterFrame Space) duration if they can sense the signal but can not decode it correctly [l0]. We derived an accurate energy consumption model for 802.11 in [6]. Denote the packet sizes of RTS, CTS, DATA, and ACK packets by Nr, Nc , Nd , and Na and packet error rates for RTS, CTS, DATA, and A CK packets between node i and j by pr,i,j, pc,j,i , pi,j and pa,i,j . In addit ion, fo r a variable x, denote 1-x by x*, and the mean value of x by . Then the average total transmission power for transmitting a packet fro m node i to one of its neighboring node, node j

   , is

   + + +

   +

   where Pm is the ma ximu m power, Pi,j and Pj,i are the transmission power for DATA and ACK packets respectively. Denote the data size, the 802.11 header size , the RTS packet size, the CTS packet size, and ACK packet size by N, Nhdr, Nrts, Ncts, and Nack, respectively. And we also define the following symbols: N8=N+ Nhdr + Nphy, Nr= Nrts+Nphy, Nc =Ncts + Nphy and Na = Nack + Nphy where Nphy is the size of physical layer overhead. In addition, denoting the receiv ing power as Pr , then the average total rece iving power for successfully receiv ing a packet fro m node i to node j as

   = Pr

   where N8=N+ Nhdr + Nphy , Nr= Nrts+Nphy , Nc = Ncts + Nphy and Na = Nack + Nphy

   Assume there are M-1intermediate nodes between a source and a destination. Let the nodes along the path fro m the source to the destination be numbered fro m 0 to M in that order. Then, the average total power for re liable transmission along the path from the source (node 0) to the destination (node M) is

   =

   Based on this formu la, it is apparent that + would be the link cost between node i and i+1.

   Most of parameters in this model can be easily obtained except the transmission power and the packet error rates. PEER adopts the transmission power estimation scheme used in [10]. If node A receives a packet transmitted at the ma ximu m power leve l fro m node B, such as RTS, CTS and broadcast packets, then node A can calculate the desired transmission power to node B, Pdesired, based on the received power, Pr, and the

   ma ximu m power level (Pm) as, Pdesired = *Prthresh*c,

   where Prthresh is the min imu m necessary received signal strength and c is an constant.

4. PEER PROTOCOL

   As a routing protocol, PEER a lso consists of route discovery process and route ma intenance scheme.

   1. Route Discovery Process

      The quickest way to find a path between two nodes would be through a shortest path routing scheme. However, there may e xist a fe w shortest (smallest number of hops) paths between the source node and destination node. For e xa mple , in Fig. 3, assuming all the intermediate nodes (A, B, E, F, G, H) a re the ne ighboring nodes of both S and D while S and D are beyond transmission range, then there are six shortest (2hops) paths (SAD, SBD, SED, SFD, SGD, SHD). A mong all the shortest paths, it is better to pick the most energy- efficient one. Denote the set of paths between the source and the destination by L, the nu mber of hops for path l by Nl, and the energy consumption for link i in path l by El,i, then the set of shortest paths Ls would be

      Ls= arg min(Nl), l L

      Fig 2: The routes between S and D

      The set of minimu m energy shortest paths Lms would be

      Lms = arg min( ), l Ls

      Even though there may be more than one minimu m energy shortest path in Lms, the routing protocol can pick

      a unique one by some criterion, such as route request packet arriving time . Based on the previous definition, the basic searching algorithm would be: 1) search for all shortest (fe west hops) paths; 2) pic k the minimu m energy path(s) among the shortest paths in (1). To imple ment this algorith m, the route request packet should carry two pieces of information: one is the hop count; the other is the energy consumption. The source node first broadcasts the route request packet with both hop count and energy consumption set to 0. Once an intermediate node rece ives a route request packet, it first updates the hop count (increased by 1) and energy consumption (increased by the energy consumption between the sender and itself) informat ion in the route request packet. Then, it will rebroadcast such packet only if one of the fo llo wing conditions holds:

      1. The node hasnt received such a packet before or the packet comes fro m a shorter (s maller nu mber of hops) path.

      2. The packet comes fro m a path with the same number of hops as the best path so far, but the energy consumption is lower.

         However, the destination node D has no such informat ion so that it could not pic k the minimu m energy shortest path even if it a lready receives all route request packets from all shortest paths. There are several ways to deal with this issue at the destination node. One option is that the destination sends a route reply packet for each route request packet it receives. This method will waste some energy as the destination will send out many route reply messages and the source node might transmit some data packets on less energy effic ient path. The other one is that the destination sets up a timer a fter receiving first route request packet. If it rece ives another route request packet before timeout, it will reset the timer. Otherwise, it will select the best path so far and reply with a route reply packet when the timer goes off. This method help reduce the energy consumption, but it may increase the route setup time. In this paper, we use the second one. The minimu m energy shortest path may still not be energy efficient enough since it tends to use the long -distance lin k. A llo wing a route to pass through some intermediate nodes may help to sav energy. To speed up the route optimization process, this can be done in parallel as the route reply message travels from the destination to the source. When the nodes that are not on the minimu m energy shortest path overhear such route reply message, they will check whether they are on a lower energy path between the sender and the receiver.

   2. Route Maintenance

      As described in section III, each node can estimate the necessary transmission power and the link cost to one of its neighboring node once it receives RTS, CTS or broadcast packet from such node. PEER requires that each node adds the link cost to the receiver in the IP header as an IP option for each data packet it transmits, and monitors the data pac kets transmitted in its neighborhood. For each data packet trans mitted, received, or overheard by the node, it will record the fo llo wing informat ion into a lin k cost table: (a ) sender; (b) rece iver;

      (c) link cost between the sender and the receiver; (d) source; (e) destination; (f) IP header ID; (g) the current time. A mong these parameters, (a) and (b) can be obtained from the MAC header, while (c) to (f) can be obtained from the IP header. The informat ion for a lin k will be kept only for a short time for accurate in formation and reducing storage overhead.

      Fro m the link cost table, a node can know how a packet passes through its neighborhood and the total lin k cost for that. For e xa mp le, node Ds lin k energy table is in Table I. As the parameters (source, destination, and IP header ID can identify a packet, we can see in the table that node D records the path info for three pac kets: Pl(S1, D1,1), P2(S2, D2,3) and P3(S3, D3,5). The first packet (Pl) uses two-hop path (A B C) in Ds neighborhood

      TABLE 1

      A LINK ENERGY TABLE

      (a)	(b)	(c)	(d)	(e)	(f)	(g)
      A	B	5	S1	D1	1	0
      B	C	4	S1	D1	1	1
      D	B	3	S2	D2	3	3
      F	G	7	S3	D3	5	4
      B	E	2	S2	D2	3	5

      and the total lin k cost is 9(5+4). The second packet (P2) uses another two-hop path (D B E) and the total link cost is 5(3+2). The third packet (P3) uses one-hop path (F G) and the lin k cost is 7. Based on the information in the link cost table, each node can help improve the loca l path as well as its corresponding end-to-end path with the three operations (Re move, Replace, and Insert) illustrated for node D in Fig.3

      Fig 3: Remove, replace and insert

      1. Re move

         The rule for Re move operation is as follows:

         Assume there is a two-hop path X A B wit h destination D and total link cost T in Xs lin k cost table. If X finds the link cost between X and B is smaller than that of the two-hop path, it will update its routing table by setting the next hop for destination D to B.

         In Fig.3(a ), node D has the two-hop path info (D B E) fro m its lin k energy table with destination D2 and the total link cost (5) for such path. If node E is one of Ds neighboring nodes, D can estimate the link cost to E fro m the RTS or CTS packets transmitted by node E. If <5, then D will update its routing table by setting the next hop for destination D2 to E. The following packet for destination D2 will go through E directly.

      2. Replace

         The rule for Replace operation is as follo ws:

         Assume that there is a two-hop path A B C with destination D and total link cost T in Xs link cost table. If X finds the total cost for the path A X is smaller than that of the two-hop path A B C, X will update its routing table by setting the next hop to destination D to

   3. In addition, it will request A to update As routing table by setting the next hop to the destination D to itself (X).

      In Fig. 3(b), Node D has the two-hop path info (A B C) in its link cost table with the destination D1 and the total link cost (9). If both A and C are Ds neighboring nodes, D can estimate the link costs to them ). If <9, then the path A D C is mo re energy e ffic ient than A B C. So node D will update its routing table by setting the next hop to destination D1 to C and request A to update As routing table by setting the next hop to destination D1 to

   4. If A accepts the request fro m D, then the follo wing packets for D1 at node A will be transmitted to node D and D will forwa rd them to C. If A does not accept the request from D, the routing info for destination Dl at node D will be purged after some time .

   1. Insert

   The rule for Insert operation is as follows:

   Assume that there is a one-hop path A B with destination D and total link cost T in Xs lin k cost table. If X finds the total cost for the path A X B is smaller than that of one-hop path, it will update its routing table by setting the next hop to destination D to B. In addition, X will request A to update As routing table by setting the next hop to the destination D to itself (X).

   In Fig. 3(c ), Node D has the one-hop path info (F G) in its link cost table with the destination D3 and the total lin k cost (7). If both F and G are Ds neighboring nodes, D can estimate the link costs to them ). If <7, then the path F D G is more energy efficient than F G. So node D will update its routing table by setting the next hop to destination D3 to G and request F to update Fs

   routing table by setting the next hop to destination D3 to

   D.

   Fig 4: An undesired improve ment

   Only Rep lace and Insert operations need the control message. The control messages are only sent out when a better path is noticed so that the maintenance overhead is very lo w. The control message includes: operation ID, requester ID, destination, ne xt hop, the total lin k cost for new path.

   The control message that D sends to A for Replace operation is [Replace, D, D1, B, the total lin k cost for ADC]. While the control message that D sends to F fo r Insert operation is [Insert, D, D3, G, the total lin k cost for FDG]. Once a node rece ives a control message, it will first check the routing info for the destination in its routing table. If the ne xt hop for such destination is diffe rent fro m that in the control message, it will d iscard such control message since the route has been changed. Within these three operations, Insert may have higher priority than the other two since it only needs to check one-hop transmission. This may not be desirable. For e xa mple , in Fig 4, node A transmits the data packet to node B. D overhears such data packet so that it sends a packet to A indicat ing that it can save energy between the lin k A B. Simila rly, node E may be inserted between

   nodes B and C. Therefore, the fina l path will be ADBEC.

   800

   TABLE-II

   Para meter	Value	Para meter	Value
   Nu mber Of Nodes	60	Packet Size(Byte)	512
   Connection Arriva l Rate	30	Connection Duration(min)	6
   Max. Speed(m/s)	10	Min. Speed(m/s)	0.5

   DEFAULT SETUP PARAM ETERS

   Normal

   However, there are t wo more options, AC and AFC, and AFC is the best path. So it would be better to let Re move and Replace have hiher priority than Insert. In PEER each node receiving Re move or Insert requests will wa it for some time before ma king the decision. If it has Insert and any other operation request, it will take the other operation. If it has both Remove and Replace operation requests, it will select one by the energy saving percentage. For the same e xa mp le, node A has the Insert (by node D), Re move, and Replace (by node F) requests, then it will on ly process Re move and Replace operations. And as AFC is better than AC <. so it takes the Replace operation.

5. PERFORMANCE EVALUATION

   We have simulated PEER, MTRTP, as well as norma l AODV protocols in Glo mosim. We modified AODV with the new link cost derived in [4] for MTRTP protocol. And the power control scheme is also applied to the norma l A ODV p rotocol. The network area is 1200(m)× 1200(m.) and the nodes are randomly distributed over the network. The available trans mission

   700 PEER

   Num of Routing Packets per Request

   MTRTP

   600

   500

   400

   300

   200

   100

   0

   40 50 60 70 80 90 100

   Num of Nodes

   Fig 5: Routing overhead

   4.5

   Normal

   power levels are 1, 5, 10, 15, 20, 25, 30, 35 mW. The

   4.0

   PEER

   Energy Consumption per Request (mJ)

   connection arriva l rate follows Po is on distribution and the MTRTP

   connection duration follows Exponential distribution. The application protocol is CBR (Constant Bit Rate) and the source and destination pairs are randomly selected. The mobility model is random waypoint with 30-second pause time. So me other default setup para meters are in Table II.

   We first studied the route discovery performance for each protocol, and then the energy consumption as well as the retransmission rate in static as well as the mobile scenarios.

   Routing Overhead and Setup Time

   In this study, we simulated 10,000 connection requests for each protocol and collected the total number of routing packets, total energy consumption, and total

   3.5

   3.0

   2.5

   2.0

   1.5

   1.0

   0.5

   0.0

   40 50 60 70 80 90 100

   Num of Nodes

   setup time on each simu lation. The simu lation results are in Fig.5-7.

   Fig 6: Energy Consumption for Routing overhead

   110

   Normal

   Average RTS Retransmission per Data Pack et: It is defined by the total number of RTS retrans mission divided by the total number of packets received. As the

   RTS pac ket is transmitted at the ma ximu m powe r leve l

   100 PEER

   MTRTP

   Average Route Setup Time (ms)

   90

   80

   70

   60

   and the packet size is very small, most of RTS retransmission is because of collision. There fore, this metric can reflect the collision rate for each protocol. Higher collision rate will cause more energy consumption, higher end-to-end delay, and lowe r throughput.

   50

   65

   40 PEER

   MTRTP

   Energy Consumption per Packet (nJ)

   30 60 Normal

   20

   10 55

   40 50 60 70 80 90 100

   Num of Nodes

   50

   Fig 7: Route setup time

   It is clear fro m the results that the normal on- demand routing protocol performs the best in terms of routing overhead, energy consumption for routing overhead, and setup time, followed by PEER and minimu m energy routing protocol. Both the routing overhead and setup time for the minimu m energy routing protocol are much more than the on -demand routing protocol, and increase dramatica lly with the number of nodes, That is because the routing overhead for minimu m energy routing protocol is O( ) (n is the number of

   nodes) as discussed in Section II. Therefore the minimu m energy routing protocol could not scale well with the

   45

   40

   40 50 60 70 80

   Num of Nodes

   Fig 8: Different density (static)

   number of nodes. While for PEER protocol, the performance is quite we ll. Even though both the routing overhead and route setup time a re still higher than the on- demand routing protocol, they are much less than the minimu m energy routing protocol. Most importantly, both routing overhead and route setup time increase very close to linearly with the number of nodes in

   the network. So PEER has high scalability with the number of nodes.

   Static Scenario

   In the static scenario, we studied the energy consumption and RTS retransmission rate performance for each protocol in three diffe rent groups: different density, different packet size, and different connection arrival rate. The simu lation t ime for each protocol is 5 hours. We monitored the total energy consumption, the total number of packets received at all destination nodes, and the total number of RTS retransmission for each

   1.50

   Average RTS Retransmission per Data Packet

   1.45

   1.40

   1.35

   1.30

   1.25

   1.20

   1.15

   1.10

   PEER MTRTP

   Normal

   40 45 50 55 60 65 70 75 80 85

   Num of Nodes

   Fig 9: Different density

   simu lation. The two metrics we used to evaluate the protocols are:

   Energy Consumption per Pick et: It is defined by the total energy consumption divided by the total number of packets rece ived. This metric reflects the energy efficiency for each protocol.

   The simulat ion results are in Fig. 8-13. For all three diffe rent groups of studies, PEER protocol performs the best in terms of Energy Consumption per Pack et as well as Average RTS Retransmission per Data Pack et, followed by MTRTP protocol and norma l p rotocol.

   PEER

   MTRTP

   Normal

   norma l p rotocol than the energy efficient routing protocols. As the link cost for MTRTP underestimates the real energy consumption, it tends to use larger number of

   75 hops. This will also increase the chance of RTS packets

   Energy Consumption per Packet (nJ)

   being lost and hence the retransmissions. So PEER

   70 protocol performs the best in terms of RTS retrans mission rate. It is interesting to observe that the RTS

   65 retransmission rate increases with the density in Fig. 9 for

   60 all protocols, while the energy consumption per packet in Fig. 8 has no such trend. This is because even though

   55 higher retransmission rate can cause more energy consumption, it can be co mpensated by the more energy

   50 efficient paths found by the routing protocols with higher number of nodes.

   45

   40 58

   35 56

   Energy Consumption per Packet (nJ)

   400 500 600 700 800

   Packet Size 54

   52

   Fig 10: Different packet size (static)

   50

   PEER MTRTP

   Normal

   PEER

   MTRTP

   Normal

   Average RTS Retransmission per Data Packet

   1.50

   1.45

   1.40

   1.35

   1.30

   48

   46

   44

   42

   40

   20 25 30 35 40

   Average Connection Arrival Rate per Hour

   Fig 12: Different connection arrival rate (static)

   1.25

   1.20

   400 500 600 700 800

   Packet Size (byte)

   Fig 11: Packet size (static)

   Both PEER and MTRTP protocol search for energy effic ient path instead of shortest path in norma l protocol so that they can perform better in terms of energy consumption. PEER perfo rms better than MTRTP in terms of energy consumption. There a re several reasons for that. First, PEER p rotocol uses a more accurate link cost. Second, there is a lot of routing overhead in MTRTP that the route request packet fro m the most energy effic ient path has higher probability of being lost in some intermed iate node, Th ird, PEER protocol can adapt the path with the environment change quickly. With power control scheme in a ll three

   1.05

   Average RTS Retransmission per Data Packet

   1.00

   0.95

   0.90

   0.85

   0.80

   0.75

   0.70

   0.65

   PEER MTRTP

   Normal

   20 25 30 35 40

   Average Connection Arrival Rate

   protocols, RTS retransmission is mainly caused by asymmetric power. For norma l protocol, the distance on each lin k can be quite different, ranging fro m very sma ll up to the transmission range. While the two energy efficient routing protocols try to use some short distance lin ks. Therefore , the retransmission rate is higher for

   Fig 13: Different connection arrival rate (static)

   Mobile Scenario

   For mob ile scenario, we also studied the same metrics as in static scenarios for each protocol. And the

   three groups of simulations are different speed, different packet size, and diffe rent connection rate. The simulation results are in Fig. 14- 19. For a ll three d ifferent groups of studies, PEER protocol performs the best in terms of Energy Consumption per Pack et as well as Average RTS Retransmission per Data Pack et.

   PEER

   Normal MTRTP

   Average RTS Retransmission per Data Packet

   1.5

   1.4

   As mentioned in s tatic scenarios, the RTS retransmission is ma inly caused by asymmetric power. Because of node mobility, MTRTP will have simila r asymmetric powe r issue as normal protocol now. In addition, due to larger number of hops, the RTS retransmission rate is larger for MTRTP than norma l protocol. Again, because PEER protocol could adapt the path with the mobility, it still t ries to use some short distance link in spite of node mobility. So it performs better than norma l protocol.

   1.3

   80

   PEER

   Energy Consumption per Packet (nJ)

   75 Normal

   1.2

   1.1

   1.0

   0 5 10 15 20

   Maximum Speed (m/s)

   Fig 14: Different speed (mobile)

   MTRTP

   70

   65

   60

   55

   50

   45

   40

   35

   PEER

   Energy Consumption per Packet (nJ)

   80 Normal MTRTP

   30

   400 500 600 700 800

   Packet Size (byte)

   Fig 16: Different packet size (mobile)

   70

   1.55

   Average RTS Retransmission per Data Packet

   60

   1.50

   1.45

   50

   1.40

   PEER

   Normal MTRTP

   40

   0 5 10 15 20

   Maxium Speed (m/s)

   Fig 15: Different speed (mobile)

   MTRTP performs the worst in terms of energy consumption, as its route maintenance scheme could not adapt with the mob ility we ll. So the o rig inal minimu m energy path would not be energy efficient any mo re because of node mobility. MTRTP even consumes much more energy than normal protocol as its path normally has more hops. As PEER adapts the path with the mobility, it could get an energy e fficient path all the time . Therefore, it performs much better than normal protocol and consumes several times lowe r energy as compared to MTRTP.

   1.35

   1.30

   1.25

   1.20

   1.15

   1.10

   400 500 600 700 800

   Packet Size (byte)

   Fig 17: Different packet size (mob ile)

   ACKNOWLEDGMENT

   Energy Consumption per Packet (nJ)

   55

   PEER

   Normal

   50 MTRTP

   45

   Siva Prasad would like to thank Mr. D. Sharath Babu Rao, who had been guiding through out to complete the work successfully, and would also like to thank the HOD, ECE Depart ment and other Professors for extending their help & support in g iving technical ideas about the paper and motivating to complete the work e ffective ly & successfully.

   40

   35

   Average RTS Retransmission per Data Packet

   1.0

   0.9

   0.8

   0.7

   0.6

   20 25 30 35 40

   Average Connection Arrival Rate per Hour

   Fig 18: Different connection arriva l rate (mob ile )

   PEER

   Normal MTRTP

   20 25 30 35 40

   Average Connection Arrival Rate per Hour

   Fig 19: Different connection arriva l rate (mob ile )

6. CONCLUSION

REFERENCES

1. K. Scott and N. Bambos, Routing and Channel Assignment for Low Power Transmission in PCS, ICUPC 96, Oct. 1996.

2. S. Doshi, S. Bhandare, and T . X Brown, An Ondemand Minimum Energy Routing Protocol for a Wireless Ad Hoc Network, ACM Mobile Computing and Communications Review, vol. 6, no. 3 , July 2002.

3. V. Rodoplu and T. Meng, Minimum Energy Mobile Wireless Networks, IEEE Journal on Selected Areas on Communications, vol. 17, Aug. 1999.

4. S. Banerjee and A. Misra, Minimum Energy Paths for Reliable Communication in Multi-hop Wireless Networks, MOBIHOC02, June. 2002.

5. J. Gomez, A. T. Campbell, M. Naghshineh, and C. Bisdikian, Conserving Transmission P ower in Wireless Ad Hoc Networks, IEEE Conference on Network Protocols, Nov. 2001.

6. J. Zhu, C. Qiao and X. Wang, A Comprehensive Minimum Energy Routing Protocol for Wireless Ad Hoc Networks, INFOCOM04 , Mar. 2004.

7. C. K. Toh, H. Cobb and D. Scott, Performance Evaluation of Battery-Life-Aware Routing Schemes for Wireless Ad Hoc Networks, ICC01, June 2001.

8. A. Misra and S. Banerjee, MRPC: Maximizing Network Lifetimefor Reliable Routing in Wireless Environments, WCNC02, Mar. 2002.

9. ANSI/IEEE Std 802.11, 1999 Edition.

10. E. Jung and N. H. Vaidya, A Power Control MAC Protocol for Ad Hoc Networks, MOBICOM02, Sept. 2002.

11. G. Bianchi and I. Tinnirello, Kalman Filter Estimation of

    the Number of Competing T erminals in an IEEE 802.11 network, INFOCOM03, 2003.

12. C-K Toh, Ad Hoc Mobile Wireless Networks Protocols and Systems, Prentice Hall, 2002.

13. T. H. Cormen, C. E. Leiserson and R. L. Rivest, Introduction to Algorithms, MIT Press, 1998.

.

It is important to design energy efficient routing protocols for mobile ad hoc networks. Spec ially, an energy e fficient routing protocol could incur much h igher control overhead and path setup delay as demonstrated by our simu lations, and consume even more energy than a norma l routing protocol in mob ile environ ment. PEER performs much better than norma l energy efficient protocol in both static scenario and mobile scenario, and under all circu mstances in terms of node mobility, network density and load. In mobile scenarios, PEER can reduces about 2/3 routing overhead and path setup delay and reduce transmission energy consumption up to 50% in a ll simu lation cases compared to the conventional energy effic ient routing protocol MTRTP.