An Opportunistic Progressive Routing (OPR) Protocol ... - CiteSeerX

1 downloads 0 Views 176KB Size Report
the progress and the instantaneous rate when forwarding packets in the fading ... For each unicast packet, the sender chooses the best forwarder from a few candidates. ... to guarantee the delivery of a packet even in the presence of void area.
An Opportunistic Progressive Routing (OPR) Protocol Maximizing Channel Efficiency Suhua TANG, Ryutaro SUZUKI, Sadao OBANA ATR Adaptive Communications Research Laboratories 2-2-2 Hikaridai, “Keihanna Science City”, 619-0288, Japan Email: {shtang, ryutaro.suzuki, obana}@atr.jp

Abstract— In this paper we study the channel efficiency of Mobile Ad hoc Networks (MANET) and try to trade off between the progress and the instantaneous rate when forwarding packets in the fading environment. We define a new metric—bit transfer speed (BTS)—as the ratio of the progress made towards the destination to the equivalent time taken to transfer a payload bit. This metric takes the overhead, rate and progress into account. Then we propose an Opportunistic Progressive Routing (OPR) protocol, jointly optimizing routing and MAC by the cross-layer design. In OPR, a node selects the forwarder with the biggest BTS for a packet and the forwarder changes as the packet size or channel status varies. The extensive simulation shows that OPR greatly reduces channel occupation time and packet loss compared with the normalized advance (NADV) [7] scheme and contention-based forwarding (CBF) [17] scheme.

I. I NTRODUCTION Mobile Ad hoc Networks (MANET) consist of mobile nodes cooperating to support multi-hop communications. They are easy to deploy and may be used in many fields such as Inter-Vehicle Communications. Researchers have proposed various routing protocols [1] for MANET from different aspects and conventionally focused on route adaptation to topology variations and reduction of overhead. A big branch of the MANET routing protocols is the geographical routing [2]-[4]. With the position of the destination known in advance [5]-[6], a node makes the forwarding decision for each outgoing packet based on its local topology in a greedy way except when the void area is encountered. Most of the geographic routing protocols take the unit disk graph assumption. In the real system, however, this is not true and another important factor—multipath fading—needs to be taken into account. With the fading effect, the transmission range becomes variable and the communication gray zones come into being [8], where sometimes the links are available and in other times a deep fading blocks the communication. As a result, the transmission of a packet may fail, and packets may get lost after the retransmission limit is reached. During the retransmission and backoff for a packet, packets destined to other neighbors are blocked in the queue, resulting in the Head-of-Line (HOL) problem [9]. The HOL problem can be solved by scheduling in the MAC layer [12]-[13] or the opportunistic routing [14]-[16]. In this paper we analyze the channel efficiency and study the tradeoff between the progress and the instantaneous rate in the geographical routing under the fading situations. Then we

propose an Opportunistic Progressive Routing (OPR) protocol. It is a joint optimization of routing and MAC, and uses space diversity. For each unicast packet, the sender chooses the best forwarder from a few candidates. The procedure involves a contention among the forwarders within the area specified by the sender, according to their Bit Transfer Speeds (BTS, defined as the ratio of progress to the equivalent time taken to transfer a payload bit of the packet). This heuristic scheme effectively reduces the total channel occupation time. The rest of the paper is organized as follows: section II reviews link quality aware progressive routing, the multi-user diversity and opportunistic routing. Then in section III we analyze the channel efficiency and define BTS. On this basis, we present the detailed OPR protocol in section IV, evaluate it by simulation in section V, and conclude the paper in section VI. II. BACKGROUND AND R ELATED W ORK A. Link Quality Aware Progressive Routing The greedy forwarding mode in the geographic routing always tries to select the longest link towards the destination. The chosen link may have a poor quality and cause retransmissions and packet loss. Lee et al. proposed the Normalized ADVance (NADV) metric [7], adopting a tradeoff between the link quality and the progress. The link quality is reflected in the link cost which focused on Packet Error Rate (PER), and the greedy forwarding rule is embodied in the progress. NADV is a conservative strategy based on the average PER and can not utilize the instantaneous high quality of a link. Moreover, similar to most of the local routing schemes, congestion may occur when many packets go through a node. B. Multi-User Diversity MAC IEEE 802.11 [11] adopts Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA) as the main channel access method. On receiving a packet from the upper layer, an interface senses the channel (by physical and virtual carrier sense supported by Network Allocation Vector (NAV)). It sets up a backoff counter when the channel is busy. When the channel becomes idle again, the backoff counter is decreased over each time slot, until it reaches zero when the packet is transmitted immediately either in DATA-ACK sequence if the packet size is less than the RTS threshold or otherwise in RTS-CTS-DATA-ACK sequence, spaced by Short Inter Frame

Space (SIFS). Unfortunately this conventional CSMA/CA scheme is susceptible to the HOL problem [9]. Knopp and Humblet [10] showed that in times of fading, the best power control scheme in a multi-user cell is to allocate power to links with the highest quality based on the feedback of channel side information (CSI). When all links experience fading independently, the scheme means that a link should only be used when its Signal to Noise Ratio (SNR) reaches the peak. This is known as the multi-user diversity. In a CSMA/CA network, the general approach to obtain CSI is to use a multicast RTS (MRTS) frame, specifying several potential communication peers to which packets are to be sent. These candidates contend according to their link quality [12]. The frame burst scheme has also been proposed to further benefit from the high rate and reduce the contention overhead [13]. Fairness is guaranteed by careful scheduling. The benefit of multi-user diversity depends on the number of active communication pairs for which packets can be scheduled. C. Opportunistic Forwarding Jain et al. explored the path diversity in the link layer [14], which benefits from the multipath routing. Instead of sticking to the same forwarder, each node specifies a candidate forwarder list, and by the contention among the candidates selects the one with the highest instantaneous SNR. This is improved by Choudhury et al. with the anycasting scheme [15], where the candidates compete for the CTS transmission in the specified order, and only the winner actually transmits a CTS, reducing the overhead. In both protocols, the routing module must build multiple routes in advance and prepare the forwarder list. With geographic information, the opportunistic routing can be more efficient, where the candidate list in MRTS is replaced by the contention area. A beaconless contention-based forwarding (CBF) protocol was proposed in [17], where each potential forwarder within the specified area contends to reply a CTS to the sender of MRTS according to their progress made towards the destination. This scheme is extended in [18] to guarantee the delivery of a packet even in the presence of void area. In the two schemes, the unit disk graph is assumed, regardless of link quality. We also focus on the opportunistic routing with geographic information. We aim to minimize the total transfer time taken for a packet from the source to the destination to improve the channel efficiency under the fading environment. Each link can usually support different rates and the rate varies over time in the presence of fading. Though it is reported in [19] that transmission over a link at a low rate may degrade the total capacity of the network, we argue that this is not always true when multi-hop transfer is concerned. Transmission of a packet at a high rate does require a short time, though it is usually at the cost of a short progress and more transmissions are needed to transfer a packet from the source to the destination. As a tradeoff, we adopt BTS as the criteria in the forwarder selection. By preferring forwarders with larger BTS, a short packet may be forwarded over a

thop MRTS tMRTS

CTS tcontention tCTS

Fig. 1.

PLCP PLCP MAC preamble header header

toverhead = thop - tpayload

DATA FCS ACK L bits tpayload

tACK

Overhead and payload of a MAC transmission.

longer link with less overhead. As a result the total transfer time is reduced. III. S YSTEM M ODEL AND A NALYSIS We take the following assumptions: All nodes are equipped with a single radio and share the same channel. The radio runs IEEE802.11b, where each rate has a different transmission range at the same SNR. Transmission of each unicast packet always involves the four frames RTS/CTS/DATA/ACK. The node density is high enough so that a sender seldom fails to find a forwarding node. In times of fading, the block fading is assumed and SNR keeps almost unchanged during the period of a single packet transmission. A. Forwarder Selection Metric under a Gaussian Channel In a MANET communication usually takes place between pairs of nodes multiple hops away, and requires multiple transmissions over consecutive links. The rate of a link varies with the distance between its two end nodes. During the transmission of a packet over a link, both the nodes around the sender and those around the receiver should yield their access to the channel. To improve the system capacity, the basic requirement is to minimize the total channel occupation time taken for a packet during its transfer from the source to the destination. We use the following symbols in the analysis. d progress made towards the destination at a node. L number of bits in the DATA payload. H hop count between the source and destination. R MAC layer communication rate for the DATA part. bit error rate at the rate of R. pR b PLR delivery ratio of a packet (payload size L) at rate R. Vbit bit transfer speed (BTS). Figure 1 shows the procedure for the transmission of a single packet over a link. It is obvious that some overhead is necessary to transfer the payload DATA. The single hop transmission time (thop ) consists of the time taken for the overhead (toverhead ) and the payload DATA (tpayload ). The overhead (except the MAC header) is usually transmitted at a basic rate, while the MAC header (28 bytes) and the payload DATA are transmitted at a flexible rate based on the instantaneous link quality such as SNR. As a result, the payload is transmitted at the cost of overhead, and the equivalent transfer time for a payload bit is tbit−hop = thop /L. It is a function of the payload length L and the transmission rate R, as shown L in Eq. 1. Packet delivery ratio (PLR ≈ (1 − pR b ) , [20]) is also included in Eq. 1 to reflect the potential retransmissions.

䎥䏌䏗䎃䏗䏕䏄䏑䏖䏉䏈䏕䎃䏖䏓䏈䏈䏇䎃䎋䏐䎒䏑䏖䎌



high instantaneous rate will have a large BTS and win the contention. The selection diversity effectively increases the average SNR, however, in a non-linear way. The biggest gain is obtained by going from no diversity to two-branch diversity and in general increasing M yields diminishing returns in terms of SNR gain. This suggests that in the opportunistic routing protocols a moderate node density is enough to achieve most of the diversity.

䎔䎰 䎕䎰 䎘䎑䎘䎰 䎔䎔䎰 䎰䎤䎻

䎓䎑䎛 䎓䎑䎙 䎓䎑䎗 䎓䎑䎕 䎓

䎔䎓䎓

䎕䎓䎓 䎖䎓䎓 䎳䏕䏒䏊䏕䏈䏖䏖䎃䎋䏐䎌

䎗䎓䎓

䎘䎓䎓

IV. D ESCRIPTION OF THE OPR P ROTOCOL

Fig. 2. Bit transfer speed under different distances (Gaussian channel, L=1000bytes).

To maximize the channel efficiency, we try to minimize tbit−total , the sum of tbit−hop over all links between the source and destination. When the nodes are densely distributed, tbit−total equals the product of tbit−hop and the hop count H, as shown in Eq. 2. H roughly equals the ratio of Dsrc−dst (the distance between the source and the destination) to the progress d. For a specific pair of nodes, Dsrc−dst is a constant, and minimizing tbit−total is equivalent to maximizing Vbit defined in Eq. 3. tbit−hop = [toverload + L/R]/L /PLR

(1)

tbit−total = tbit−hop · H ≈ tbit−hop · Dsrc−dst /d

(2)

Vbit =

d tbit−hop

=

d · PR toverhead /L + 1/R L

(3)

Figure 2 shows Vbit at different rates, where SNR at a distance is the same for all rates (this is true in 802.11b where the transmission power of all rates is the same). In such cases, when SNR is big enough (the distance is short enough), the maximum rate should always be used. However, as the distance increases, Vbit of a specific rate reaches its peek and then decreases quickly. Heuristically, the routing policy is to choose the forwarder with the largest Vbit , similar to the real transportation system where the speed is of a great concern. B. Under Multipath Fading Under the multipath fading environment, the capacity of a fading channel is less than that of a non-fading channel with the same average SNR. A node with M antennas separated far enough can receive M independent copies of the signal with the same average SNR (γ). Then the selection diversity can be used to improve the channel capacity. In times of Rayleigh fading, by always selecting the branch the highest SNR, with M the output SNR increases to γ Σ = γ i=1 1/i. Though the links from a sender to its neighbors usually have different average SNR, the selection diversity can be applied in a distributed way in a MANET as follows: before transmitting a packet, the sender transmits a MRTS, trying to recruit a forwarder from the specified candidates. A candidate detects the SNR on receiving the MRTS and chooses its best rate. Then each candidate contends according to its BTS. It is expected that the candidate with a long progress and a

In our OPR protocol, the routing module at a node knows its own position (by the GPS module). It also learns the position of the destination [5]. Instead of predetermining a forwarder, the routing module sends down each outgoing unicast packet to the MAC module, specifying a contention area within which any node with a big BTS can opportunistically forward the packet. The MAC module of the sender broadcasts an MRTS that carries the data packet size L, its own position and the destination’s position. This MRTS is to recruit a forwarder, which continues the packet transmission. On receiving a MRTS, a node detects the instantaneous SNR of the link. With its own position, the node determines whether it lies in the specified area. If confirmative, it acts as a competitive candidate and performs the following operations. A. Calculation of Bit Transfer Speed First of all, each candidate calculates the progress d made by itself towards the destination according to the positions of the sender and the destination carried in the MRTS and its own position, as shown in Fig. 3. Then each candidate checks all the available rates and calculates the packet delivery ratio PLR under the measured SNR. Next the candidate calculates the overhead time toverhead by Eq. 4. toverhead consists of three parts: (1) the time taken for the transmission of MRTS (tM RT S ), CTS (tCT S ), ACK (tACK ) at the basic rate, (2) the time thead taken for the transmission of the PLCP preamble, PLCP header, MAC header and FCS of the DATA frame, and (3) the contention time (tcontention ) associated with Vbit in Eq. 5, where x is the integer part of x, and Nslots is the maximal contention slots. tM RT S , tCT S and tACK include the SIFS time, as shown in Fig. 1. The solution of Eq. 3-5 gives the accurate Vbit . When tcontention is small compared with other terms in Eq. 4, its expected value, half of the contention window, can be used instead. Then Vbit is calculated by Eq. 3. toverhead = tM RT S + tcontention + tCT S + thead + tACK (4) tcontention = {(1 − Vbit /Vmax ) · Nslots +Nrand }·tslot (5) B. Contention Policy Similar to [17], the contention area is specified by the position of the sender and the destination, and an angle parameter θ, as shown in Fig. 3. Each node within the area

G

E A

S

TABLE I MAC PARAMETERS .

contention area

progressSE

D

T T

B C

progressSC

Fig. 3.

F

Contention area.

acts as a candidate on hearing the MRTS and contends to reply a CTS to the sender. When the sender of MRTS can learn the positions of its neighbors, the angle θ in Fig. 3 can be made adaptive so that there are always enough candidates. To prefer nodes with a big Vbit , each candidate waits for a period, tcontention , that is inversely related with its BTS, as shown in Eq. 5. The bigger Vbit is, the smaller tcontention is. A small random integer, Nrand , is added in Eq. 5 to avoid the collision when two candidates happen to have the same BTS. The contention is on the unit of MAC time slot tslot . At each slot, a candidate decreases its contention counter if the channel is sensed idle. When the contention counter reaches 0, a CTS frame, carrying the rate R for the next DATA frame, is transmitted immediately. If a candidate node senses the channel to get busy during its contention waiting, CTS transmission by another contender is inferred and it cancels its own contention timer. C. MAC Sequence Though MAC in OPR uses similar frames as IEEE 802.11, two distinct differences exist. One is the post-determination of the receiver and the other is the post-determination of the DATA rate. Hence, the NAV mechanism is modified as follows: A MRTS carries a tentative duration, setting up a NAV at all the nodes around the sender. This NAV is the sum of the CTS transmission time and the longest contention time tslot · Nslots so that the sender can receives the CTS from the node with the least BTS. The duration field of a CTS is the same as the normal CSMA/CA scheme. However it is calculated by the receiver according to the rate R and the data length L. DATA/ACK are exchanged the same as before. The sender of a MRTS may fail to receive a correct CTS or ACK due to collisions. Then a backoff and retransmission is necessary. In such cases the backoff scheme is the same as IEEE802.11, after which the retransmission takes place, where another contention recruits a new forwarder that may be different from before. V. S IMULATION E VALUATION The simulation is done with network simulator QualNet 3.8 [23]. OPR is implemented by the modification of IEEE 802.11 distributed coordination function (DCF) [11]. The Rician model is used to emulate the channel fading. The ratio of the power of Light-of-Sight (LOS) signal to that of non-LOS signals, K, is the Rician fading parameter. Rician fading degenerates to Rayleigh fading when K decreases to

Parameter PLCP Preamble and header Slot Time SIFS time Contention slot time (OPR only) Maximum contention slots Nslots (OPR only) Minimum contention window size Maximum contention window size

7

10

8

1 SRC

3

2

Value 192 us 20 us 10 us 20 us 20 31 1023

4 CBR traffic

5

6 DST

11

9

Fig. 4.

A simple topology.

0. The channel experiences no fading when K approaches ∞. Qualnet emulates the bit error of 802.11b caused by the wireless channel. We modified it according to [20] to make the BER-SNR relation more realistic. In the simulation, we compare three protocols, NADV [7], CBF [17] and the proposed OPR protocol. In NADV, the rate is adjusted by the Auto Rate Fallback (ARF) scheme [21]. In CBF, the progress determines the contention priority. Therefore, the fixed rate (2Mbps) is used. OPR uses RBAR [22] to calculate the instantaneous rate. The three protocols adopt similar position distribution scheme like DREAM [5]. Each node builds and relies on its own position table to find the position of the destination. The main parameters of the MAC module are listed in Table I. We mainly use two metrics to evaluate the routing performance: Packet Error Ratio (PER), and Packet Transfer Time (PTT) which is defined as the sum of channel occupation time taken for a packet over all links during its transfer from the source to the destination. The less PTT is, the more efficient the channel becomes. Simulation results are averaged over 50 runs with different random seeds. Figure 4 shows a simple topology where a Constant Bit Rate (CBR) flow is established from node 1 to 6. In the communication, nodes 2 and 4 move up and down while the destination node 6 moves towards the source node 1 and back to its original place. The speed is about 5m/s. Other nodes are kept static. A. Effect of Fading Parameter Figure 5-6 show PER and PTT respectively under different Rician fading parameters. In Figure 5, when there is no fading (K = ∞), NADV and OPR have a PER approaching zero, while CBF has a relatively high PER because it always select the same longest link with a relatively high BER. As the fading

1.6

2 NADV

NADV

1.5

CBF

PER (%)

PER (%)

1.2

OPR

0.8

0.4

CBF OPR

1

0.5

0

0

K=0.0

K=1.0

K=10.

K=100

K=㺙

32

64

Rician fading parameter

Fig. 5.

128

256

512

1024

Packet size (byte)

PER under Rician fading (payload size=512 bytes).

Fig. 7.

PER under different packet size (Rayleigh fading).

20 30 Packet transfer time (ms)

Packet transfer time (ms)

NADV CBF

15

OPR 10

5

0

NADV CBF

20

OPR

10

0

K=0.0

K=1.0

K=10.

K=100

K=㺙

32

Rician fading parameter

Fig. 6.

Packet transfer time under Rician fading (payload size=512 bytes).

parameter decreases, in CBF, the long link may happen to have high quality and the packet can be received correctly; or its quality is so low that MRTS over it is not heard and a shorter link with higher quality is used instead. As a result the PER is decreased. In NADV, fading increases the link quality variation and packets transmitted at high rates tend to have a high PER. In contrast, the rate in OPR is chosen according to the instantaneous SNR. As a result, OPR still has a low PER even at Rayleigh fading. Figure 6 is the PTT. It depends on the transmission rate and hop count. With large fading parameters, the high rate in NADV and OPR greatly reduces the packet transfer time compared with CBF. As fading gets serious, retransmissions in NADV increase the PTT. Under all cases OPR has the lowest PTT since it can benefit from the instantaneous high rate of the long links, reducing both transmission count and the per-transmission time. This results in the highest channel efficiency, as is desired. B. Effect of Packet Size Figure 7-8 show PER and PTT under different packet size at Rayleigh fading. CBF sticks to a low rate and has a low PER. In NADV, the high rate may not match the link quality and cause an obvious PER, which increases as the packet size does. OPR always has the lowest PER. For a short packet the long progress of a link is preferred. However a long link is more susceptible to packet errors and results in a little increase of PER. When the packet size is small, in OPR, a long progress is preferred to a high rate in the forwarder selection, and

64

128

256

512

1024

Packet size (byte)

Fig. 8.

Packet transfer time under different packet size (Rayleigh fading).

the per-bit overhead in Eq. 3 is effectively reduced. As a result, OPR has similar PTT as CBF, which is much better than NADV. As the packet size increases, PTT increases in all protocols, however with different trends. In OPR, higher rates are gradually preferred in order to increase BTS in Eq. 3. The superiority of OPR is very obvious when the packet size increases to 1024 bytes, where OPR has much less PTT than both NADV and CBF. Meanwhile PTT of CBF exceeds that of NADV due to the low transmission rate. C. Effect of Mobility In the following we show the effect of mobility with the general topology. We generate topologies, each having 50 nodes randomly distributed in the square with the size 4000m x 300m. All nodes moves according to the random way point model. There is no pause and the maximum speed is changed in different simulations. At each speed, the results obtained by 50 random topologies are averaged. In this simulation, each packet has a payload size of 512 bytes. When a node receives an MRTS, it measures SNR, calculates BTS and determines the best rate, which is used for the transmission of the payload of the DATA frame. It takes 0.87ms to transmit the CTS frame at the basic rate (2Mbps) and the DATA frame at 11Mbps. The block fading assumption is accurate when the variation of distance is less than 1/4 wavelength (3.125cm at 2.4GHz), or the channel coherence time must be no less than the time taken for the transmission of CTS and DATA. Then the relative speed between two nodes must be less than 35.8m/s, or the maximal speed is less than 17.9m/s.

12

R EFERENCES

8 PER (%)

NADV CBF OPR

4

0 0

5

10

15

20

25

Maximal Speed (m/s)

Fig. 9.

PER under different speed (Rayleigh fading)

60 Packet transfer time (ms)

NADV CBF OPR

40

20

0 0

5

10

15

20

25

Maximal Speed (m/s)

Fig. 10.

Packet transfer time under different speed (Rayleigh fading)

Figure 9-10 show PER and PTT. OPR works well and has a lower PER and PTT compared with both CBF and OPR. As the speed increases, the probability that the actual SNR of the data frames differs from the SNR measured on receiving MRTS gets large. As a result, PER of OPR also gradually increases. As the speed gets larger than 17.9m/s, the block fading model becomes inaccurate. Therefore OPR is suitable for the low and moderate mobile environment. VI. C ONCLUSION We have analyzed the channel efficiency of progressive routing and proposed a new metric for the opportunistic forwarder selection, which takes into account the tradeoff between the progress and the instantaneous high rate in the fading environment. Calculation of the metric involves the instantaneous SNR, the packet size, the propagation distance, and the potential transmission failure. And the best forwarder is the one with the biggest instantaneous bit transfer speed towards the destination. As a result a packet is opportunistically forwarded over a link with both a long progress and a high rate. The total result is the improved channel efficiency with a reduced end-to-end packet transfer time and less delay. The simulation results confirmed that OPR has a much better performance compared with NADV and CBF in a moderate mobile environment with multipath fading. ACKNOWLEDGMENT This work was supported by the National Institute of Information and Communications Technology (NICT), Japan.

[1] E.M.Royer and C.-K.Toh, A review of current routing protocols for ad hoc mobile wireless networks, IEEE Personal Communications, Vol.6, No.2, pp.46–55, 1999 [2] B.Karp and H.T.Kung, GPSR: Greedy Perimeter Stateless Routing for Wireless Networks, MobiCom’00, pp.243–254, 2000. [3] F.Kuhn, R.Wattenhofer, Y.Zhang, and A.Zollinger, Geometric Ad-Hoc Routing: Of Theory and Practice, ACM symposium on Principles of distributed computing, pp.63–72, 2003 [4] L.Blazevic, J.-Y.Le Boudec, and S.Giordano, A location-based routing method for mobile ad hoc networks, IEEE Transactions on Mobile Computing, Vol.4, No.2, pp.97–110, 2005 [5] S.Basagni, I.Chlamtac, V.R.Syrotiuk, and B.A.Woodward, A distance routing effect algorithm for mobility (DREAM), MobiCom’98, pp.76– 84, 1998. [6] R.Morris, A Scalable Location Service for Geographic Ad Hoc Routing, MobiCom’00, pp.120–130, 2000. [7] S.Lee, B.Bhattacharjee, and S.Banerjee, Efficient geographic routing in multihop wireless networks, MobiHoc’05, pp.230–241, 2005. [8] H.Lundgren, E.Nordstrom, and C.Tschudin, Coping with communication gray zones in IEEE 802.11b based ad hoc networks, Proc. 5th WoWMoM, pp.49–55, 2002 [9] P.Bhagwat, P.Bhattacharya, A.Krishna, and S.K.Tripathi, Enhancing throughput over wireless LANs using channel state dependent packet scheduling, INFOCOM’96, Vol.3, pp.1133–1140, 1996. [10] R.Knopp and P.A.Humblet, Information capacity and power control in single-cell multiuser communications, ICC’95, Vol.1, pp.331–335, 1995. [11] IEEE Computer society LAN MAN Standards Committee, Wireless LAN Medium Access Protocol (MAC) and Physical Layer (PHY) Specification, IEEE Std 802.11-1997, IEEE, 1997. [12] J.Wang, H.Zhai, and Y.Fang, Opportunistic packet Scheduling and Media Access control for wireless LANs and multi-hop ad hoc networks, WCNC, Vol.2 pp.1234–1239, 2004. [13] Z.Ji, Y.Yang, J.Zhou, M.Takai, and R.Bagrodia, Exploiting medium access diversity in rate adaptive wireless LANs, Mobicom, pp.345–359, 2004. [14] S.Jain and S.Das, Exploiting path diversity in link layer in wireless ad hoc networks, CEWIT 2003. [15] R.R.Choudhury and N.H.Vaidya, Mac-layer anycasting in ad hoc networks, ACM SIGCOMM Computer Communication Review, Vol.34, No.1, pp.75–80, 2004. [16] S.Biswas and R.Morris, ExOR: Opportunistic routing in multi-hop wireless networks, SIGCOMM’05, 2005. [17] H.Fuessler, J.Widmer, M.Kasemann, and M. Mauve, Beaconless position-based routing for mobile ad-hoc networks, Ad Hoc Networks, Vol. No.1, pp.351–369, 2003. [18] M.Chawla, N.Goel, K.Kalaichelvan, A.Nayak, and I.Stojmenovic, Beaconless position-based routing with guaranteed delivery for wireless ad hoc and sensor networks, IFIP 1st Conference on Ad-Hoc Networking, 2006. [19] M.Heusse, F.Rousseau, G.Berger-Sabbatel, and A.Duda, Performance anomaly of 802.11b, INFOCOM’03, Vol.2, pp.836–843, 2003 [20] Jd.P.Pavon and S.Choi, Link adaptation strategy for IEEE 802.11 WLAN via received signal strength measurement, ICC’03, Vol.2, pp.1108–1113, 2003 [21] A. Kamerman and L. Monteban, WaveLAN-II: A high-performance wireless LAN for the unlicensed band, Bell Labs Technical Journal, Vol.2, No.3, pp.118–133, 1997. [22] G.Holland, N.H.Vaidya, and P.Bahl, A rate-adaptive MAC protocol for multi-Hop wireless networks, MobiCom’01, pp.236–251, 2001. [23] QualNet 3.8, http://www.scalable-networks.com/