1
A Busy-Tone based MAC Scheme for Wireless Ad Hoc Networks using Directional Antennas Hong-Ning Dai and Kam-Wing Ng Department of Computer Science and Engineering The Chinese University of Hong Kong, Hong Kong
[email protected],
[email protected] Min-You Wu Computer Science & Engineering Department Shanghai Jiao Tong University, China
[email protected]
Abstract Applying directional antennas in wireless ad hoc networks offers numerous benefits, such as extended communication range, increased spatial reuse, improved capacity and suppressed interference. However, directional antennas can cause new location-dependent carrier sensing problems, such as new hidden terminal and deafness problems, which can cause severe penalties to the performance. Recently, a few schemes have been proposed to tackle these problems. However, these methods can provide limited solutions on the hidden terminal and deafness problems. We propose a new MAC protocol, termed the busy-tone based directional medium access control (BT-DMAC) protocol. When the transmission is in progress, the sender and the receiver will turn on their omnidirectional busy tone to protect the transmission. By combining with directional network allocation vector (DNAV), the scheme almost mitigates the hidden and the deafness problems completely. The mechanism increases the probability of successful data transmission and consequently improves the network throughput. This paper describes the BT-DMAC scheme and analyzes its performance. The simulation results also demonstrate the effectiveness of the protocol.
Index Terms Directional Antennas, Medium Access Control, Wireless Ad Hoc Networks
2
I. I NTRODUCTION There is a strong interest in applying directional antennas or smart antennas to wireless ad hoc networks recently. Directional antennas have numerous benefits, such as longer transmission range, reduced interference and increased spatial reuse. However, utilizing directional antennas in wireless ad hoc networks is still limited due to inherent constraints, such as physical size and device complexity. Besides, new problems due to directional beamforming, such as hidden terminals and deafness, also prevent directional antennas from being widely deployed. The first kind of hidden terminal problems originates in asymmetry in gain. As depicted in Fig. 1, node B initiates a transmission to node C by sending Directional RTS (DRTS). Then, C will reply with a Directional CTS (DCTS). Assume that node A is in idle mode (listens omnidirectionally) and far enough from node C, A can not hear the DCTS from C since an omnidirectional gain Go is smaller than a directional gain Gd . Then nodes B and C begin data transmission by pointing their beams to each other with the gain Gd . However, when the transmission between B and C is in progress, node A wants to communicate with node D. Node A uses the directional beam towards node D to sense the channel and finds that the channel is idle as node C is in directional receiving mode at this moment. Subsequently, it sends a DRTS to node D. Since node C is in directional receiving mode with a gain Gd , it is very possible that the DRTS from A interferes with node C. In other words, a transmitter in directional mode (with Gd ) and a receiver in omnidirectional mode (with Go ) may be out of each other’s range, but they may
reach each other if they both transmit and receive directionally (with Gd ). This kind of hidden terminals arises due to different gains with omnidirectional and directional modes of directional antennas.
E Go A
D F
Fig. 1.
B
C Gd
Hidden terminals and deafness
In the second scenario (Fig. 1), while node D is sending a packet to node E, node B sends a RTS to
3
node C. At this time, node C’s DCTS can reach node D, however D cannot hear it since D is beamformed in the direction of E. When the communication between nodes B and C is in progress, assume that node D finishes transmitting to E and has a packet for node B. Node D’s DNAV (Directional Network Allocate Vector) indicates that the direction to node B is free. Therefore node D begins to send a DRTS to node B, and it will lead to a collision with B’s transmission. This kind of hidden terminals arises as the transmitter and the receiver have not heard the DRTS or DCTS control frames. However, this problem will not happen with omnidirectional antennas. Another drawback of directional beamforming is the deafness problem [1] [2]. Briefly, the deafness is caused when a transmitter fails to communicate to its intended receiver, as the receiver is beamformed towards a direction away from the transmitter. In Fig. 1, node F wants to transmit data to node C, using the route through node B. When B gets a packet from F, it beamforms in the direction of C and forwards the packet. At this time, F is unaware of the transmission between B and C since it does not hear the communication of B and C. If it initiates the next packet to B, F will not receive the CTS reply from B since B is beamformed to C. Node F retransmits the RTS as there is no response from B. This process will go on until the RTS retransmitting limit has been reached. The excessive retransmission of control packets will bring a severe penalty on the network performance. Since F would increase its backoff interval on each attempt, this event can result in unfairness as well. Several schemes have been proposed, attempting to tackle the directional hidden terminal and deafness problems [2]–[4]. However, to the best of our knowledge, there is no protocol that solves the hidden terminal and deafness problems completely with low cost. Our main contributions of this paper are: 1) We have identified the weakness of existing solutions to the hidden terminal and deafness problems in wireless ad hoc networks using directional antennas. 2) We present a novel MAC protocol, Busy-Tone based Directional MAC (BT-DMAC) to attack the hidden terminal and deafness problems. 3) We have analyzed the performance of BT-DMAC and the numerical results demonstrate the effectiveness of BT-DMAC. Comparisons with other existing MAC schemes are also given. 4) We have also conducted simulation experiments. Our results show that BT-DMAC can achieve higher spatial reuse as compared with the existing schemes, while keeping fairness among nodes. The rest of the paper is organized as follows. The related work is presented in Section II. Then we describe the BT-DMAC protocol in Section III. The performance evaluation of the scheme is given in Section IV. Finally, we summarize the paper in Section V.
4
II. R ELATED
WORK
Many researchers have proposed new MAC schemes based on directional antennas [1]–[12]. There are quite a few protocols based on the distributed coordination function (DCF) of IEEE 802.11, which typically uses RTS/CTS control packets to prevent interferences. However, these mechanisms can not prevent the new location-dependent carrier sensing problems: the hidden terminals (due to asymmetry in gain and due to unheard RTS/CTS) and the deafness problems [1] [2]. These problems will have major impacts on the performance of ad hoc networks. Several protocols have been proposed, attempting to tackle the hidden terminal and deafness problems. Dual Busy Tone Multiple Access (DBTMA) [13] uses omnidirectional transmitting and receiving busy tones to avoid omnidirectional hidden terminals and exposed terminals. Huang et al. [8] have extended DBTMA to directional antennas. However, these protocol have not solved the directional hidden terminal problems. The deafness problem is also not addressed in the two schemes. Circular-DMAC [3] attempts to tackle both hidden terminal and deafness problems by sending directional RTS/CTS packets before transmitting data sequentially. However, transmitting multiple RTS/CTS packets for each data packet will severely degrade the performance. Choudhury et al. [2] propose a tone-based notification mechanism which allows neighbors of a node to classify congestion from deafness and react appropriately. However this scheme can not prevent retransmitting RTS requests from other nodes and it cannot mitigate the deafness completely. Furthermore, this tone-based protocol does nothing to the hidden terminals. We propose a Busy-Tone based Directional MAC protocol (BT-DMAC) to address these problems. While the transmitter and receiver are communicating, they will turn on their busy tones to prevent possible collisions. Combining the mechanism with DNAV scheme can mitigate the hidden terminal and deafness problems almost completely. III. P ROPOSED P ROTOCOL A. Antenna Model Each node has two interfaces: one of them is equipped with a switched beam antenna and another one is attached with an omnidirectional antenna. The switched beam antenna has two modes: omnidirectional mode and directional mode. When a node is in idle state, as it does not know the arrival direction of a signal, it will listen in all directions by switching its directional antenna to omnidirectional mode. Once a signal is sensed, the antenna begins to receive with an omnidirectional gain Go . During the signal receiving period, the antenna performs an azimuthal scan in order to select the beam that acquires the
5
maximum gain. Then the node will record the beam information for future use. The directional mode will be used to transmit or receive RTS, CTS, data and ACK frames. According to the Friss equation, the maximum distance between the transmitter and the receiver is lengthened with increased antenna gains in the transmitter and the receiver. As the directional gain Gd is greater than the omnidirectional gain Go , directional antennas offer longer transmitting and receiving ranges. When both nodes are in omnidirectional modes, the maximum communication range is O-O range (Roo ). When one node is in omnidirectional mode, and another node transmits or receives directionally, the maximum communication range is D-O range (Rdo ). It is obvious that Rdo > Roo . If both nodes transmit and receive directionally, the maximum communication range can be sufficiently extended to D-D range (Rdd ), which is greater than Rdo and Roo . However, since a receiver does not know who is the exact transmitter in advance, it can only receive the RTS frame in omnidirectional mode. Hence, the effective communication range is bounded by Rdo . The omnidirectional antenna is only used to send busy tones omnidirectionally. In order to cover the range of directional transmission, the transmitting power of the omnidirectional antenna is increased suitably. Since an omnidirectional antenna is only for sending tones, it can be easily implemented and mounted in wireless stations with low cost. B. Neighbors Discovery One of the hardest problems with directional antennas is to find the directions of neighbors of a node, or neighbor discovery. A node needs to determine where and when to point the beam to transmit or receive. In this paper, we propose a neighbor discovery scheme with low cost and without additional hardwares. Each node listens omnidirectionally when it is in the idle mode. If the node hears any frames (RTS, CTS, data and ACK), no matter whether the frames are intended for the node or not, it will recognize the direction which the frames are sent from by using selection diversity and determining which beam its neighbor is located in. Then it will record the number of the beam and the identifier of its neighbor into a table, called Neighbor Location Table (NLT). Directional Network Allocation Vector (DNAV) is a directional version of NAV of IEEE 802.11, proposed by [7] and [1]. DNAV excludes the directions and sets the corresponding durations, towards which the node is not allowed to initiate a transmission to avoid collisions with data or control frames. We integrate the DNAV mechanism with the NLT. When a node receives a RTS frame and the receipt address matches its address, it beamforms towards the transmitter (switch to directional mode) and replies the RTS with a CTS frame. If the control frames are not for itself, it will update the sender’s information
6
in the NLT and set the corresponding DNAVs. Fig. 2 shows that a node A has a four-beam antenna and its neighbors, C, B and E, D located in beam 0, 1 and 2 respectively. Node A stores its neighbors’ location information into a NLT. When node B communicates with node E, node A will modify the corresponding entries in its NLT by reading DNAVs from the RTS/CTS frames.
3
D
Fig. 2.
C A
2
0
E B
1
The NLT of Node A beam No.
0 1 2 3
neighbors C B, E D -
available duration
Yes No 10(ms) Yes Yes -
An example of the Neighbor Location Table
C. The BT-DMAC Protocol In the BT-DMAC protocol, two busy tones are implemented with enough spectral separation on the single shared channel. When the transmission is in progress, the transmitter and the receiver turn on the transmitting busy tone BTt and the receiving busy tone BTr , respectively. Each BTt comprises two subtones, an ID tone and a beam number tone for the transmitting node. Each BTr comprises two sub-tones, an ID tone and a beam number tone for the receiving node. To encode several-bit information into the sine wave, there is an easy way to achieve this: Pulse Modulation, which sends signals by turning the sine wave on and off (in Fig. 3). Any node hearing the busy tones learns node identifiers and beam numbers from the tones and deduces whether the potential sending will interfere with the current transmission. Any attempts that may cause potential collisions are prevented. The busy tones occupy only a small portion of the whole frequency. If a node has data to send, it will search the NLT to find the beam for the destination and check the availability of the beam in the DNAVs. If the beam is available, the node will listen directionally by using that beam. If no busy tone is detected, the node will send a RTS immediately. If a busy tone is sensed, the node will identify the corresponding sender ID and the beam number from the tones. If the ID matches the destination’s, it is obvious that the destination is busy now and the attempt will be deferred. If the ID does not match the destination’s, then the sender will compare the beam number with
7
Pulse modulation
Bits to be transmitted
1 Fig. 3.
0
1
1
0
1
0
1
The Pulse Modulation
Available Spectrum beam No. sub-tone ID sub-tone data channel
Frequency (Hz) Fig. 4.
The Frequency chart of the BT-DMAC protocol
that one used to communicate with the destinations in its NLT. If the beam number is identical, the node will defer its transmission to avoid collision. The receiver and the transmitter will turn on the busy tones until the ACK frame is received. Fig. 5 depicts the finite state machine (FSM) of the BT-DMAC scheme. In BT-DMAC, a node is in one of the following states: IDLE, WF CTS, S DATA and WF DATA. When a node has no data frames to send and has not received any requests, it will stay in the IDLE state. If it has data frames to send, it will sense the medium first. If the channel for the destination direction is free, it will send a RTS to the destination and enter the WF CTS state. Otherwise, it will go back to the IDLE state. If the sender in the WF CTS state receives a CTS, it will turn on its BTt and begin to transmit. If there is no CTS within the retry timer, it will go back to the IDLE state. If the sender gets the ACK reply correctly, it will turn off its BTt and go back to the IDLE state. However, if the ACK cannot reach the sender within the retry timers, the sender will retransmit the data and increase the retry counter until it reaches the maximum value. On the other hand, when a node in IDLE state hears a RTS, it will point its beam towards the sender direction and send a CTS, then turn on its BTr . If the data frame is correctly received, the receiver will reply the sender with an ACK and turn off its BTr . The BT-DMAC protocol can be illustrated by an example, shown in Fig.6. In this figure, there are
8
T3 WF_CTS
T1 WF_DATA
T7 T9 T8
T4
T2 IDLE
T5
T6 S_DATA
T1 : data ready AND the beam for the destination available/send RTS and set timer T2 : time out AND no CTS / Ti (E / A): the transition
T3 : no CTS AND retry < RETRY LIMIT / retransmit RTS
E: the trigger event
T4 : CTS received / BT on, set timer
A: the action of the transition
T5 : ACK received OR retry >= MAX RETRY / BT off T6 : no ACK AND retry < MAX RETRY/ retransmission T7 : DATA received / send ACK, BT off T8 : time out / BT off T9 : RTS received / send CTS, BT on, set timer
Fig. 5.
The Finite State Machine of BT-DMAC
several nodes, A, B, C, D, E and F, which are equipped with four-beam antennas. When node A has data frames to send to node B, it will sense the channel towards node B by using Beam 0. If Beam 0 is free, node A points the beam towards node B, sends a RTS frame to node B, and then goes into the WF CTS state. After node B receives the RTS, it switches to directional mode (beamforms towards the direction of A using Beam 2) and replies with a CTS, turning on its BTr (B, 2). Then it sets up a timer and enters the WF DATA state. After receiving the CTS from node B, node A turns on its BTt (A, 0) and goes into the S DATA state and sends the data frames. Upon successful reception of the data frame, node B replies to node A with an ACK and turns off the BTr , entering the IDLE state. If, for any reason, node B does not receive the data packet before the timer expires, it turns off the BTr and enters the IDLE state. If node A receives the ACK frame successfully, it turns off its BTt and goes into the IDLE state. Otherwise, it will retransmit the data frames until the timer expires. In this scenario, node C within the busy tone range of node B has data to send to B (using Beam 2). Since it senses the BTr (B, 2) from node B, it will defer its transmission to avoid colliding. If node D within the BT ranges of node A and B wants to communicate with node E (D and E are close enough), it detects the channel first. When it senses the BTt (A, 0) from node A, it decodes the beam number (0)
9
from the tone and deduces that its transmission (using Beam 0 also) will cause interferences with nodes A and B. Therefore node D will defer its transmission. The time diagram with the operations of node A and node B is shown in Fig. 8.
BTt(A,0) 3
0
2
F1
Fig. 6.
3
0
3
0
2
D1
2
B1
3
0
2
A1
3
0
2
E1
3
0
2
C1 E
BTr(B,2)
AG
o
D
Rdd B
C G
d
F
a scenario illustrates the operation of BT-DMAC
Fig. 7.
BT-DMAC solves the hidden terminals and deafness
problems
RTS
A: Data transmission A: Busy tone B: Data transmission B: Busy tone Fig. 8.
DATA BT
t
CTS
ACK BT
r
The Time Diagram of BT-DMAC
D. The Hidden Terminal and Deafness Problems with BT-DMAC As we have mentioned before, since a node receives the RTS frame only in omnidirectional mode, the effective communication range is bounded by Rdo . However, the busy tone can be sensed in D-D range Rdd since nodes listen to busy tones with directional mode and busy tones are sent to reach the range of
directional transmission. The extended busy tones make it possible to reduce potential collisions further.
10
The BT-DMAC can effectively solve two kinds of hidden terminal problems (due to asymmetry in gain and unheard RTS/CTS) and mitigate the deafness problem. As the example shown in Fig. 1, if BT-DMAC is implemented in this scenario, the nodes B and C will turn on the busy tones during their transmission period. Since the busy tones are sent to reach the range of directional transmission and node A will listen directionally towards node D, it can sense the busy tone sent by node B in D-D range. Therefore, node A will deduce that the direction to node D is busy and defer its transmission to node D (in Fig.7). BT-DMAC offers a complete solution to the second kind of hidden terminal problems as well. In Fig.7, although node D can not hear the DRTS/DCTS due to directional beamforming towards node E, D can diagnose that the direction towards B and C is unavailable as it can hear the busy tones before it begins to send to node B. Hence, the mechanism can effectively solve the two kinds of new hidden terminal problems. Before node F begins to send, it will sense the channel first. It will detect the busy tone of node B and deduces that B is busy. Then it will defer the transmission to node B. Therefore, the deafness has been settled by using BT-DMAC. Furthermore, BT-DMAC does not prohibit other normal transmissions that will not collide with the current communication. Take the example in Fig. 6 as well, the node F wants to transmit to the node E. It also senses the BTt from node A, however it realizes that it will not interrupt the transmission between A and B when it uses the beam 1. Hence, the transmission between the nodes F and E can be carried on in parallel with A and B’s. Therefore, the mechanism can improve the spatial reuse. IV. P ERFORMANCE A NALYSIS A. Assumptions The discrete Markov chain model, used in [14] [15] to study CSMA and BTMA is adopted here to evaluate the saturation throughput of BT-DMAC. We have extended the throughput model to support directional antennas and range extension is also considered in our model. For easy discussion, we assume that all nodes have a uniform setting (the identical beamwidth θ and antenna gains) and the nodes are deployed in two-dimensional Poisson distribution with density λ. According to the Friss equation, the transmission range is proportional to the product of the gains of the transmitter and receiver antennas, we have γ =
Gd Go ,
where Gd and Go are the directional gain and the
omnidirectional gain respectively. Suppose N is the average number of nodes within a circular region of 2 , λπR2 = γN and λπR2 = γ 2 N . an O-O radius. Hence, we get: N = λπRoo do dd
We assume that each node operate in time-slotted mode, with a time slot τ . As mentioned in [15], when the time slot τ is very small, the performance of the time-slotted protocol is almost the same as the
11
PII
PSI=1
SUCCEED
PIS
IDLE
PIC PCI=1
Fig. 9.
COLLISION
The Markov chain model for a node
performance of the asynchronous version of the protocol. The transmission time of RTS, CTS, data and ACK frames are depicted as the multiple of τ , i.e. trts , tcts , tdata and tack . The throughput of BT-DMAC is based on a heavy-traffic assumption that a node always has a packet to send. The probability that a node transmits in a slot is denoted as p, which is smaller than the ready probability pr due to potential interferences within the channel.
B. Throughput In this paper, we adopt a simple model to derive the saturation throughput of the BT-DMAC scheme. The throughput is calculated by the proportion of time that a node spends transmitting data packets successfully in the average. We use a Markov chain model, as shown in Fig. 9. Let P (S), P (I) and P (C) denote the steady-state probability of state SUCCEED, IDLE and COLLISION respectively. From
the Markov chain model of Fig. 9, we have: T hroughput =
P (S) · tdata P (C)TC + P (S)TS + P (I)TI
(1)
where TC , TS and TI are the duration of states COLLISION, SUCCEED and IDLE respectively. Then we derive the steady-state probabilities, transition probabilities and times spent at different states of BT-DMAC. Firstly, the duration in time slots of a node in the SUCCEED state is: TS = (trts + 1) + (tcts + 1) + (tdata + 1) + (tack + 1) = trts + tcts + tdata + tack + 4
(2)
12
where trts , tcts , tdata and tack are the duration times of transmitting RTS, CTS, data and ACK frames respectively. As the handshake between the sender and the receiver may be interrupted during the collision period that varies from T1 = trts + 1 to T2 = trts + tcts + 2. The length of collision period follows a truncated geometric distribution with parameter p, low bound T1 and upper bound T2 . The mean value of the truncated geometric distribution as TC , which is given by: TC =
T 2 −T1 1−p pi (T1 + i) 1 − pT2 −T1 +1 i=0
(3)
The duration of a node in IDLE state TI is 1 (τ ). Since a node in the IDLE state listens omnidirectionally, any node which can reach it by directional beamforming is a Directional-Omnidirectional neighbor [1]. Thus the potential interference range is D-O range instead of Omnidirectional-Omnidirectional (O-O) range. Hence, the transition probability PII that the node continues to stay in IDLE state in a slot is equal to the probability that it does not initiate any transmission and there is no node around it initiating a transmission towards it. As the two events are independent, we have: PII
= (1 − p) ·
∞
(1 − p)i ·
i=0 2 −pλπRdo
= (1 − p)e
2 )i (λπRdo 2 e−λπRdo i!
= (1 − p)e−pγN
(4)
From Fig. 9, the steady-state probability of the IDLE state equals: P (I) = P (I) · PII + P (S) + P (C)
(5)
Noting that P (S) + P (C) = 1 − P (I), hence we have: P (I) =
1 1 = 2 − PII 2 − (1 − p)e−pγN
(6)
From Fig. 9, the steady-state probability of SUCCEED state P (S) can be calculated by P (S) = P (I) · PIS . Substituting for PIS , P (S) can be expressed as follows: P (S) = P (I) · PIS =
PIS 2 − (1 − p)e−pγN
(7)
To derive the transition probability PIS from IDLE to SUCCEED, we need to calculate the probability Pis (r) that node A successfully initiates a four-way handshake with node B which is r distance away
from A. Pis (r) equals the probability that node A transmits in a given time slot, node B does not transmit
13
1 Rdo
RTS
A
r
2 Rdd
CTS
B
3
4
Fig. 10.
The interference region for BT-DMAC
in the same time slot, and none of the nodes around them interfere with the handshakes. Then Pis (r) can be calculated as: Pis (r) = p · (1 − p) · P1 · P2 · P3 · P4
(8)
where P1 is the probability that no node interferes with the RTS reception, P2 is the probability that no node interferes with the CTS reception, P3 is the probability that no node interferes with the data reception and P4 is the probability that no node interferes with the ACK reception. As depicted in Fig. 10, the sender A transmits a RTS to the receiver B. However, when the node B is in idle mode, it listens omnidirectionally. When it hears the signal from A, it will point its beam towards A and listens directionally. During the reception time of RTS, the node B may be interfered by the nodes within the sector region with the radius Rdd (D-D range) and the angle θ. Since the RTS frame sent by node A can prevent the nodes within region 1 from interfering with node B (by setting DNAVs), node B may only be interfered by the nodes within region 2 and 4. Therefore, the probability that no node interferes with the reception of RTS is equal to the product of the probability that no node in the region with the radius Rdo (D-O range) transmits in the same time slot as node A does, and the probability that no node in the region 2 transmits towards B during the (RTS+1) period: r2
P1 = e−p λπRdo · e−p (trts +1)λ( 2 Rdd − 2
2
θ
2
tan( θ2 ))
(9)
where p = pθ/(2π). After the receiver B hears the RTS correctly, it will beamform towards node A and response with a CTS and node A will listen directionally. As illustrated in Fig.10, as the RTS sent by A can block the nodes within the region 1 (setting DNAVs), the node A may only be interfered by the nodes within the
14
region 3(the size of region 3 is S3 =
θ 2
2 − · Rdd
θ 2
2 ). Hence, we get the probability that no node · Rdo
interferes with the reception of CTS to node A: ∞
P2 = (
i=0
(1 − p )i ·
(λS3 )i −λS3 (tcts +1) ) = e−p λS3 (tcts +1) e i!
(10)
where p = pθ/(2π). Since nodes A and B turn on their busy tones after starting transmission, and the nodes within the D-D range are suppressed by the busy tones. Thus, the probability that no node interferes with the reception of data frames to node B is considered to be one, i.e. P3 = 1. Similarly, the probability that no node interferes with the reception of ACK to node A is one, i.e. P4 = 1. We also consider the spatial reuse of directional antennas [15]. The spatial reuse factor σ(r) is defined to be the number of possible concurrent transmissions in the combined region covered by nodes A and B. The σ(r) is the radio between the total region covered by nodes A and B and the actual area that excludes the region covered by the handshake between nodes A and B. If there is one handshake in areas 1 and 4, then in theory there may be Stotal /(S1 + S4 ) concurrent handshakes in the region excluding 1 and 4. Therefore, σ(r) = Stotal /(S1 + S4 )
(11)
Here, we just give a very rough estimation of the spatial reuse factor for BT-DMAC, and the practical parameter is lower than the theoretical value. Then we have: PIS =
Rdo 0
2rσ(r)Pis (r)dr
(12)
C. Numerical Results We have compared the throughput of our proposed BT-DMAC with Basic DMAC [5], and IEEE 802.11 MAC (omnidirectional antennas) in Fig. 11 and 12. Fig. 11 shows the throughputs of the three protocols when the directional gain is regarded to be equal to the omnidirectional one (γ = 1). The results show that BT-DMAC has performed much better than Basic DMAC and IEEE 802.11 MAC at different values of the beamwidth θ (π/12, π/6, π/3 and π/2). Directional antenna gain is considered in Fig. 12 (γ = 2). Basic DMAC works well when the beamwidth is narrow. When the antenna has a wider beam, Basic DMAC is more vulnerable to interferences as the number of neighbor nodes is increased. As a result, the throughput degrades conspicuously. Basic DMAC performs even worse than IEEE 802.11 when the average number of nodes is increased further. However, BT-DMAC has still outperformed Basic DMAC and IEEE 802.11 MAC when directional antenna gain is
15
BT-DMAC Basic DMAC IEEE 802.11
1.0
0.9
0.8
0.8
0.7
0.7
Throughput
Throughput
0.9
0.6 0.5
0.6 0.5
0.4
0.4
0.3
0.3
0.2
0.2
0.1
0.1
0 1
2
3
4
5
6
7
8
9
BT-DMAC Basic DMAC IEEE 802.11
1.0
0 1
10
2
3
4
5
N
(a) θ = π/12 BT-DMAC Basic DMAC IEEE 802.11
0.9
0.8
0.8
0.7
0.7
0.6 0.5
0.4 0.3
0.2
0.2
0.1
0.1 5
6
7
8
9
10
0 1
2
N
3
4
5
6
7
8
9
10
N
(c) θ = π/3 Fig. 11.
10
0.5
0.3
4
9
0.6
0.4
3
8
BT-DMAC Basic DMAC IEEE 802.11
1.0
Throughput
Throughput
0.9
2
7
(b) θ = π/6
1.0
0 1
6 N
(d) θ = π/2
Throughput comparison when γ = 1 and p = 0.008 (trts = tcts = tack = 5τ , tdata = 100τ )
considered. For example, when θ = π/3 and N = 10, the throughput of BT-DMAC is almost 3.5 times of that of IEEE 802.11. One possible reason is that spatial reuse brought by directional antennas can counteract the bad effect of increasing interferences (Basic DMAC can work well when the beamwidth is narrow and the number of nodes is small). When the antenna has a wider beamwidth, a transmitting nodes is more vulnerable to more interferences (Basic DMAC performs worse). As BT-DMAC deploys busy tones BTt /BTr to protect the ongoing transmission of data and ACK packets, it has gained a better performance. Furthermore, the hidden terminal and the deafness problems, which cannot be completely mitigated by other exiting MAC schemes, have been alleviated by BT-DMAC. V. S IMULATION R ESULTS We have extended GloMoSim 2.03 [16] with the support of directional antennas. We try to compare our proposed protocol with other existing MAC schemes (Basic-DMAC and IEEE 802.11 MAC). To
16
BT-DMAC Basic DMAC IEEE 802.11
1.0
0.9
0.8
0.8
0.7
0.7
Throughput
Throughput
0.9
0.6 0.5
0.6 0.5
0.4
0.4
0.3
0.3
0.2
0.2
0.1
0.1
0 1
2
3
4
5
6
7
8
9
BT-DMAC Basic DMAC IEEE 802.11
1.0
0 1
10
2
3
4
5
N
(a) θ = π/12 BT-DMAC Basic DMAC IEEE 802.11
0.9
0.8
0.8
0.7
0.7
0.6 0.5
0.4 0.3
0.2
0.2
0.1
0.1 5
6
7
8
9
0 1
10
2
3
4
N
5
6
7
8
9
10
N
(c) θ = π/3 Fig. 12.
10
0.5
0.3
4
9
0.6
0.4
3
8
BT-DMAC Basic DMAC IEEE 802.11
1.0
Throughput
Throughput
0.9
2
7
(b) θ = π/6
1.0
0 1
6 N
(d) θ = π/2
Throughput comparison when γ = 2 and p = 0.008 (trts = tcts = tack = 5τ , tdata = 100τ )
ensure equal conditions between Basic-DMAC, IEEE 802.11 and our proposed BT-DMAC, we consider an identical scenario to Basic-DMAC in Fig. 13. The five nodes are linearly arranged. The distance between each two nodes is 360 meters. The transmission range of each node is 376.78 meters. The TCP packet size is 1460 bytes and the bandwidth is set to 2Mbps.
0 Fig. 13.
1
2
3
4
The linear topology
In the first scenario, we simulate two single-hop TCP connections, in terms of TCP(1) (from node 1 to node 0), TCP(2) (from node 2 to node 3). The results of Table I show that both Basic DMAC and BT-
17
TABLE I S IMULATION R ESULT (I)
Connections
IEEE 802.11
Basic DMAC
BT-DMAC
TCP(1) (node 1 to node 0)
427.77
838.60
880.39
TCP(2) (node 2 to node 3)
436.37
839.61
828.51
Overall throughput (Kbps)
864.14
1678.21
1708.90
DMAC have outperformed IEEE 802.11 MAC. Due to the benefits of spatial reuse of directional antennas, multiple simultaneous transmissions are allowed, hence the higher throughputs have been gained. Scenario 1 shows the best case for using Basic DMAC, which performs almost the same as BT-DMAC. The reason is that the collision probability of control packets of Basic DMAC is quite small in this scenario (DRTS packets are sent to two opposite directions). TABLE II S IMULATION R ESULT (II)
Connections
IEEE 802.11
Basic DMAC
BT-DMAC
TCP(3) (node 1 to node 2)
10.72
475.90
802.10
TCP(4) (node 3 to node 4)
821.77
633.21
819.95
Overall throughput (Kbps)
823.49
1109.11
1622.05
The second scenario also consists of two TCP connections: TCP(3) from node 1 to node 2 and TCP(4) from node 3 to node 4. The simulation results are given by Table II. The result of the IEEE 802.11 deployed network shows that TCP(3) is jammed by TCP(4) due to the omnidirectional transmission. Basic DMAC has gained much better performance than IEEE 802.11 scheme, as DRTS and OCTS are used to reduce the interferences. However, OCTS packets sent by node 2 can still interfere with the reception of ACK from node 4 to node 3. Therefore, the aggregated throughput has been degraded by the collisions of OCTS and ACK packets. Our proposed BT-DMAC has acquired much higher throughput than Basic DMAC and IEEE 802.11. Furthermore, the fairness using BT-DMAC is also much better than Basic DMAC and IEEE 802.11, for instance, TCP(3) and TCP(4) almost keep in the same throughput. One of possible reasons is that DCTS can further improve the spatial reuse (DCTS of node 2 won’t interfere with the transmission of node 3 and node 4 again). Besides, busy tones can provide better protection of on-going transmissions.
18
VI. D ISCUSSION A. Mobility Our proposed BT-DMAC can be adapted to mobile environments. Here, we adopt a similar method which is used in [7]. Each node updates the NLT every time it receives a newer packet from the same neighbor and invalidates the item in the NLT if it fails to get the CTS reply from its neighbor within a certain retry limit (e.g. 3). Let us take the scenario shown in Fig. 2, when node D is moving out the coverage range of node A, node A may not know this. If node A has a sending request, first, it will send a DRTS to node D by using the original location info of node D. But it fails to receive the CTS reply within 3 retries. Therefore, it will invalidate the corresponding item which contains node D (e.g. beam no.2). The mechanism can also be applied to the dead nodes.
3
D
Fig. 14.
C A
2
0
E B
1
The NLT of Node A beam No.
0 1 2 3
neighbors C B, E -
available duration
Yes No 10(ms) Yes Yes -
the Mobility of BT-DMAC
B. The Possible Collision of Busy Tones Busy tones are often implemented by sine waves which have no correction mechanism to ensure accurate reception of busy tones. Therefore, we propose a method to solve it. When busy tones are sent, a special duration ti is added before each tone sequence, as shown in Fig. 15. The duration ti taken from an interval [tmin , T can uniquely identify the node i in local. To ensure to decode the busy tones with some degree of accuracy, the detecting program of busy tones will continually take several samples for a period (e.g. 4 times for a sequences). If one of the sequences is collided by the tones sent by other nodes, the program can still keep a high probability that the busy tones are correctly received. Since the starting duration ti is identical in local for each node, the
19
Pulse modulation TA
TA
1 0 1 1 0 1 0 1 Fig. 15.
1 0 1 1
the possible collision of busy tones
probability that a collision happens again after the first collision is quite low. Meanwhile, as there are only several bits of info encodes in each busy tones, the detecting duration only occupies a short period. VII. C ONCLUSION Directional antennas offer numerous benefits, but they also cause new collisions, such as new hidden terminal and deafness problems. Although a few schemes have been proposed to address these problems, the solutions are not satisfactory. In this paper, we propose a new MAC protocol (BT-DMAC), which combines busy tones with the DNAV mechanism and can solve the hidden and the deafness problems completely. The mechanism increases the probability of successful data transmission and consequently improves the network throughput. This paper describes the BT-DMAC scheme and analyzes its performance. We have also presented the analytical model of BT-DMAC and calculated its performance. The numerical and simulation results show that BT-DMAC can achieve much higher throughput than other schemes. The BT-DMAC also maintains a high spatial reuse and alleviates the interferences from hidden terminals and deafness. Our future work is to simulate BT-DMAC in large-scale networks and implement it in realistic networks to determine how well it performs. R EFERENCES [1] R. R. Choudhury, X. Yang, N. H. Vaidya, and R. Ramanathan, “Using directional antennas for medium access control in ad hoc networks,” in Proc. MobiCOM’2002, Atlanta, Georgia, USA, 2002, pp. 59–70. [2] R. R. Choudhury and N. H. Vaidya, “Deafness: a MAC problem in ad hoc networks when using directional antennas,” in Proc. ICNP’2004, 2004, pp. 283 – 292. [3] T. Korakis, G. Jakllari, and L. Tassiulas, “A MAC protocol for full exploitation of directional antennas in ad-hoc wireless networks,” in Proc. MobiHOC’2003, Annapolis, Maryland, USA, 2003, pp. 98 – 107. [4] H. Gossain, C. Cordeiro, D. Cavalcanti, and D. P. Agrawal, “The deafness problems and solutions in wireless ad hoc networks using directional antennas,” in IEEE Global Telecommunications Conference Workshops, 2004.
20
[5] Y. B. Ko, V. Shankarkumar, and N. H. Vaidya, “Medium access control protocols using directional antennas in ad hoc networks,” in Proc. INFOCOM’2000, Tel Aviv, Israel, 2000, pp. 13 – 21. [6] A. Nasipuri, S. Ye, and R. E. Hiromoto, “A MAC protocol for mobile ad hoc networks using directional antennas,” in Proc. WCNC’2000, vol. 3, Chicago, IL, USA, 2000, pp. 1214 – 1219. [7] M. Takai, J. Martin, R. Bagrodia, and A. Ren, “Directional virtual carrier sensing for directional antennas in mobile ad hoc networks,” in Proc. MobiHOC’2002, Lausanne, Switzerland, 2002, pp. 183 – 193. [8] C. S. Z. Huang, C.-C. Shen and C. Jaikaeo, “A busy-tone based directional MAC protocol for ad hoc networks,” in Proc. MILCOM’2002, 2002. [9] H. Singh and S. Singh, “Smart-802.11b MAC protocol for use with smart antennas,” in Proc. ICC’2004, vol. 6, Paris, France, 2004, pp. 3684 – 3688. [10] L. Bao and J. Garcia-Luna-Aceves, “Transmission scheduling in ad hoc networks with directional antennas,” in Proc. MOBICOM’2002, Atlanta, Georgia, USA, 2002, pp. 48 – 58. [11] Z. Zhang, “Pure directional transmission and reception algorithms in wireless ad hoc networks with directional antennas,” in Proc. ICC’2005, vol. 5, Seoul, Korea, 2005, pp. 3386 – 3390. [12] R. Ramanathan, “On the performance of ad hoc networks with beamforming antennas,” in Proc. MobiHOC’2001, Long Beach, CA, USA, 2001, pp. 95 – 105. [13] J. Deng and Z. Haas, “Dual busy tone multiple access (DBTMA): A new medium access control for packet radio networks,” in Proc. IEEE ICUPC, 1998. [14] L. Wu and P. K. Varshney, “Performance analysis of CSMA and BTMA protocols in multihop networks (i). single channel case,” Information Sciences, vol. 120, pp. 159–77, 1999. [15] Y. Wang and J. J. Garcia-Luna-Aceves, “Directional collision avoidance in ad hoc networks,” Performance Evaluation Journal, vol. 58, pp. 215–241, 2004. [16] The
global
mobile
information
http://pcl.cs.ucla.edu/projects/glomosim/
systems
simulation
library
(GloMoSim).
[Online].
Available:
