• Share
  • Email
  • Embed
  • Like
  • Save
  • Private Content
Overlapped carrier sense multiple access

Overlapped carrier sense multiple access






Total Views
Views on SlideShare
Embed Views



0 Embeds 0

No embeds


Upload Details

Uploaded via as Adobe PDF

Usage Rights

© All Rights Reserved

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
Post Comment
Edit your comment

    Overlapped carrier sense multiple access Overlapped carrier sense multiple access Document Transcript

    • IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. 8, NO. 3, MARCH 2009 369 Overlapped Carrier-Sense Multiple Access (OCSMA) in Wireless Ad Hoc Networks Surendra Boppana, Student Member, IEEE, and John M. Shea, Member, IEEE Abstract—In wireless ad hoc networks (WANets), multihop routing may result in a radio knowing the content of transmissions of nearby radios. This knowledge can be used to improve spatial reuse in the network, thereby enhancing network throughput. Consider two radios, Alice and Bob, that are neighbors in a WANet that does not employ spread-spectrum multiple access. Suppose that Alice transmits a packet to Bob for which Bob is not the final destination. Later, Bob forwards that packet on to the destination. Any transmission by Bob not intended for Alice usually causes interference that prevents Alice from receiving a packet from any of her neighbors. However, if Bob is transmitting a packet that he previously received from Alice, then Alice knows the content of the interfering packet, which can allow her to receive a packet from one of her neighbors during Bob’s transmission. In this paper, we develop overlapped transmission techniques based on this idea and analyze several factors affecting their performance. We then develop a MAC protocol based on the IEEE 802.11 standard to support overlapped transmission in a WANet. The resulting overlapped carrier-sense multiple access (OCSMA) protocol improves spatial reuse and end-to-end throughput in several scenarios. Index Terms—Wireless ad hoc networks, MAC protocols, multipacket reception, interference cancellation. Ç1 INTRODUCTIONW IRELESS networks present several challenging issues for the network designer that are quite different fromtheir wired counterparts. An impairment that is due to the the PHY leads to greater spatial reuse in the network. MAC protocols were proposed in [5] and [8] that take advantage of the MPR capabilities of the PHY to increase the spatialbroadcast nature of the wireless network is interference. reuse in networks to provide high throughput in heavySince all the nodes share the same physical medium, traffic and low delay in light traffic.simultaneous transmissions may result in interference at the In most cases, mobile radios do not have sufficientreceiving nodes. In networks that do not employ code- processing power to perform complex MUD schemes.division multiple access, medium-access control (MAC) Recent work on the transport capacity of wireless networksprotocols such as IEEE 802.11 [1] are used to allocate the [9] indicates that in the low-attenuation regime, multistagechannel resources to specific transmitters and receivers so relaying using cancellation of known interference is orderas to minimize the interference in the network. Tradition- optimal. Here, the interference is known from the use ofally, the design of the MAC protocol is carried out multihop routing. Using interference cancellation (IC) forindependently of the physical-layer (PHY) design, assum- only known interference may significantly improve net-ing a simplistic collision channel model. In these models, a work performance at a reasonable complexity.packet is successfully received by a node if there are no To explain how an interfering signal may be known inother transmissions in its interference range. These MAC multihop routing in a wireless ad hoc networks (WANet),protocols schedule transmissions such that the collisions in consider a four-node linear network consisting of nodes A,the network are minimized. B, C, and D, in which A transmits a packet to D using Multiuser detection (MUD) [2], [3], [4], [5], [6], [7] in multihop routing. In a slotted communication systemwireless networks has been proposed as a means to increase employing a conventional MAC protocol, a typical se-spatial re-use by increasing the number of simultaneous quence of transmissions for a packet would betransmissions in the network. MUD techniques are em-ployed at the PHY to recover information from colliding 1 : A ! B; 2 : B ! C; 3 : C ! D;packets at the receiver. These signal processing techniquesused at the PHY enable a node to receive packets in the where the notation 1 : A ! B indicates that node Apresence of other transmissions in its communication range. transmits a packet to node B in time slot 1, etc. UnderThis multipacket reception (MPR) capability of the nodes at conventional MAC protocols, in the time slot when C forwards a packet to D, A is not allowed to transmit to B since C’s transmission will cause interference at B. How-. S. Boppana is with Qualcomm Inc., 6455 Lusk Blvd., San Diego, CA ever, when an MPR-based MAC protocol is employed, 92121. E-mail: sboppana@qualcomm.com.. J.M. Shea is with the Department of Electrical and Computer Engineering, simultaneous transmissions of A ! B and C ! D are University of Florida, 439 ENG Bldg. #33, PO Box 116130, Gainesville, possible, since MUD techniques can be employed at B to FL 32611-6130. E-mail: jshea@ece.ufl.edu. recover the packet transmitted by A. Note that the packetManuscript received 21 Dec. 2007; revised 25 May 2008; accepted 7 July transmitted by C to D is the same packet that B forwarded2008; published online 30 July 2008. to C in an earlier time slot (ignoring the differences in theFor information on obtaining reprints of this article, please send e-mail to:tmc@computer.org, and reference IEEECS Log Number TMC-2007-12-0392. headers). If B were to retain a copy of the packet that itDigital Object Identifier no. 10.1109/TMC.2008.114. forwarded to C, B would have information regarding the 1536-1233/09/$25.00 ß 2009 IEEE Published by the IEEE CS, CASS, ComSoc, IES, & SPS
    • 370 IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. 8, NO. 3, MARCH 2009interfering transmission. This greatly reduces the complex-ity of the MUD algorithms employed at the PHY to recoverthe packet transmitted by A. This example is revisited inSection 2. The idea of employing this type of known-IC techniqueto increase simultaneous transmissions in WANets was firstanalyzed in [10]. In our previous work, knowledge of theinterfering signal is assumed at both the transmitter and thereceiver, and the receiver performs MUD/IC to recoveradditional messages. Limitations on scheduling suchsimultaneous transmissions were analyzed, and a MACprotocol that supports such simultaneous transmissionswas proposed. Fig. 1. Four-node linear network with conventional scheduling. The idea of employing network coding to increase spatialreuse and throughput in WANets has recently received The rest of the paper is organized as follows: Section 2considerable attention from the research community [11], introduces the idea of employing overlapped transmission[12], [13], [14], [15], [16], [17], [18], [19], [20], [21]. A in a linear network. In Section 3, some limits on performingtransmitting node exploits the broadcast nature of the overlapped transmissions in wireless networks are evalu-physical medium along with the knowledge of the inter- ated. Section 4 describes the overlapped carrier-sensefering messages at the receiving nodes to combine/encode multiple access (OCSMA) MAC protocol. The design issuesmultiple independent messages at the network layer and of the protocol are considered in Section 5, and Section 6transmit to several nodes. A node receiving the encoded provides performance evaluation of the protocol. The papermessage uses the knowledge of the other interfering is concluded in Section 7.messages available at the network layer to recover themessage intended for it. Practical channel sharing schemes 2 MOTIVATIONthat support network coding in WANets were proposed in[12], [19], [21], and [15]. In this section, we illustrate the idea of overlapped Our approach is similar to some network coding transmissions in a four-node linear network, which isapproaches to increase simultaneous transmissions in illustrated in Fig. 1. We assume that the nodes canWANets [22], [23], [24], [25]. In PHY network coding [22], communicate only with the adjacent nodes and operate inrelay nodes may receive signals consisting of several the half-duplex mode. Node A transmits packets to node Dsimultaneous transmissions. These signals are decoded, through multihop routing. A typical transmission sequencereencoded, and relayed on to their final destinations. The under a conventional scheduling scheme is depicted indestination, which has the knowledge of the other interfer- Fig. 1, in which it takes three time slots for a packet from A to reach D. The scheduled transmissions in a given time sloting signals, mitigates the interference and recovers the are marked by solid directed arrows along with the packetintended transmission. However, this approach requires identifiers, and the interference caused by these transmis-perfect synchronization among those transmissions that sions are marked by dashed arrows. Under typical carrier-interfere at a relay node. An alternative strategy called sense multiple access protocols with collision avoidanceanalog network coding [24] does not require the inter- (CSMA/CA), when packet m1 is being forwarded by C inmediate relay nodes to decode the signal. When a relay time slot t3 , A cannot transmit the message m2 sincenode receives a signal consisting of interfering transmis- C’s transmission will cause interference at B.sions, the node amplifies the signal and broadcasts it. Only The throughput of this network can be improved bypacket-level synchronization is necessary between the employing simultaneous transmissions as described below.interfering transmissions. The intended destinations use We observe that in the time slot t3 , C forwards the packet m1their knowledge of the interfering transmissions to mitigate that it received from B in the earlier time slot t2 . If B were tothe interference and recover the desired transmission. retain a copy of the message m1 locally, it knows theThese approaches are similar to the idea of employing message being transmitted by C in time slot t3 (assumingMUD with known IC. These works analyze the PHY that link-layer encryption is not used and any differences inaspects involved but do not address the MAC-layer the headers are ignored). If A is allowed to transmit theimplications of employing such simultaneous transmission message m2 in the time slot t3 , B can use the storedschemes in ad hoc networks. information regarding m1 to mitigate the interference In this paper, we analyze some of the fundamental limits caused by C’s transmission. We call this additionalon performing overlapped transmissions in a WANet. Our transmission, which results from the mitigation of knownanalysis provides an understanding of the performance interference, an overlapped transmission.gains of such transmissions and an insight into the PHY and A scheduling scheme that employs the idea of over-MAC interaction required for scheduling such transmis- lapped transmission for the four-node linear network issions. We then design a MAC protocol based on the IEEE depicted in Fig. 2 (also, refer to [22, Section 2]). Under this802.11 MAC protocol that exploits this feature to improve scheduling scheme, a packet is transmitted from A to B bythe spatial reuse and throughput in wireless networks. employing overlapped transmission during the time slot
    • BOPPANA AND SHEA: OVERLAPPED CARRIER-SENSE MULTIPLE ACCESS (OCSMA) IN WIRELESS AD HOC NETWORKS 371 receiver, respectively. Similarly, a transmission between two nodes is a secondary transmission if at least one of the nodes has noncausal information about the primary transmissions in the present transmission interval and performs MUD/IC to mitigate the interference. In the network in Fig. 2, the transmission of the message m2 from node A to B in time slot t3 for which B performs MUD/IC to mitigate the interference from C’s transmission is the secondary transmis- sion, and nodes A and B are called the secondary transmitter and secondary receiver, respectively.2 3 OVERLAPPED TRANSMISSION IN WIRELESSFig. 2. Four-node linear network with overlapped transmissions. AD HOC NETWORKSthat C forwards a packet to D. Since the transmission of the In this section, we analyze some fundamental limits onpacket from A to B did not involve the allocation of a performing overlapped transmission in a wireless network.separate time slot for its transmission, a packet requires on The purposes of the analysis is to provide insights into thethe average only two time slots to be transmitted from A to types of scenarios in which overlapped transmission mayD. These two time slots are required for the scheduling of be appropriate and the design of a MAC protocol totransmissions from B to C and C to D, respectively. The efficiently utilize the simulcasting capability.performance gain of this scheduling scheme can be 3.1 System Modelmeasured in terms of transmission efficiency, which isdefined as the ratio of the time taken for the transmission of Consider first a WANet with nodes distributed according toM packets under conventional scheduling scheme and the a two-dimensional homogeneous Poisson point processscheduling scheme employing overlapped transmissions, with density  nodes per unit area. Each node is equippedrespectively. The transmission efficiency À4 of this scheme with a transceiver and communicates with other nodes inis given by half-duplex mode. We assume that each node has an infinite packet buffer, and each radio retains copies of the 3M 3 packets it forwards unless that packet is transmitted to its À4 ¼ ; % ; M ) 1; ð1Þ 2ðM À 1Þ þ 3 2 final destination or until that packet has been forwarded on by one of its neighbors. To investigate some of the issueswhere M is the total number of packets transmitted by A. that will limit the performance of overlapped transmission,Note that under conventional scheduling, it takes three time we analyze the use of overlapped transmission in a systemslots for every packet from A to reach D. However, using using slotted communications. In this model, each nodethe scheduling scheme that employs overlapped transmis- transmits in a given time slot with probability p. Thissions, it takes two time slots on the average for a packet assumption is only to facilitate the analysis of overlappedfrom A to reach D.1 Similarly, in an NðN ! 4Þ node linear network, the transmissions in ad hoc networks in Section 3. However, notransmission efficiency ÀN can be shown to be such assumption is made during the development of the MAC protocol in Section 4 or the network simulations in N À 1 þ 3ðM À 1Þ 3 Section 6. We also assume that the secondary transmitter is ÀN ¼ ; % ; M ) 1: ð2Þ N À 1 þ 2ðM À 1Þ 2 informed of the corresponding primary transmission and performs overlapped transmission at the same time as theWe observe that the centralized scheduling scheme that primary transmission. The received power Pr (in the faremploys overlapped transmissions has the potential to field) can be expressed asimprove the efficiency of a linear network by up toapproximately 50 percent over the conventional scheme. Pr ¼ Kp dÀ Pt ; r ð3ÞIn Section 3, we look at some of the limitations of employingoverlapped transmissions in WANets, and in Section 4, we where Pt is the transmitted power, dr is the distancedevelop a MAC protocol that supports overlapped trans- between the transmitter and the receiver, Kp is a constant,missions in wireless networks. Since the focus of this work and is the path-loss exponent. In the absence ofis on developing a MAC protocol for overlapped transmis- interference, we assume that a transmission at the max-sion, the PHY aspects of the protocol are not evaluated here. imum power level will be received correctly if and only if We identify a transmission between two nodes as a the intended receiver is within a distance of one unit fromprimary transmission if the transmission is not predicated on the transmitter. We also assume that there is somethe use of noncausal knowledge of the interfering signals interference range, which is typically larger than theduring that transmission interval. For example, in the transmission range. Nodes within the interference rangenetwork in Fig. 2, the transmission of message m1 from C to but outside the transmission range of a transmitter canD in time slot t3 is the primary transmission, and the nodes C 2. The terms “primary” and “secondary” are also used in the cognitiveand D are called the primary transmitter and the primary radio literature to classify users according to their regulatory status and should not be confused with the terminology employed here, in which 1. The first packet requires three time slots. users are classified according to their roles in an overlapped transmission.
    • 372 IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. 8, NO. 3, MARCH 2009 of C. We also assume that B can perform perfect IC of C’s transmission and recover the packet transmitted by A. However, A’s transmission causes interference at node D. In order to analyze the impact of the secondary transmis- sion at the primary receiver, we evaluate the SIR at node D. We assume that the secondary transmission is the only source of interference at D. For conciseness, we introduce the following notation. Let Xij be the random variable denoting the distance between the nodes i and j. Also, let Al ðr1 ; r2 ; dÞ denote the area of the lens formed by the intersection of two circles of radii r1 andFig. 3. Ad hoc network. r2 with centers separated by a distance d. Mathematically  2 detect the presence of a transmission but will not be able to r þ d2 À r2correctly decode the packet being transmitted. Al ðr1 ; r2 ; dÞ ¼ r2 cosÀ1 1 1 2 2r1 d In this section, we consider some limitations on the  2  r þ d2 À r2ability to utilize overlapped transmissions to improve the þ r2 cosÀ1 2 2 1 2r2 d ð4Þthroughput in a WANet. These limitations come from hpffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffithe following two sources: À 0:5 ðr1 þ r2 þ dÞðr1 þ r2 À dÞ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffii . Interference due to secondary transmission. Since the  ðr1 À r2 þ dÞðÀr1 þ r2 þ dÞ : secondary receiver has noncausal knowledge of the primary transmission, it can mitigate the interfer- Let denote the ratio of the distances between nodes C and ence due to the primary transmitter and recover the D and A and D, respectively. Mathematically intended message. However, the secondary trans- XCD mission causes interference, possibly to several ¼ ; XCD < 1; XAD > 1: ð5Þ primary transmissions. In Section 3.2, we evaluate XAD the amount of interference that a secondary trans- The constraint XCD < 1 indicates that D is in the transmis- mission may cause at the primary receiver and sion range of C, and the constraint XAD > 1 reflects the fact suggest how this interference can be controlled by that the secondary transmitter A is allowed to transmit only adapting the power level of the secondary transmis- if it is not in the transmission range of D. Hence, we have sion to meet the specified signal-to-interference ratio < 1. In an exponential path-loss channel without fading, (SIR) and outage requirements or by careful selec- the ratio of the powers of the primary transmission to the tion of the secondary transmitter. secondary transmission at node D can be expressed as . Probability of secondary transmission. Overlapped   transmissions depend on the availability of suitable XCD À ¼ ¼ À ; < 1: ð6Þ secondary transmitters and the successful reception XAD of the messages at the secondary receiver. The density of can be expressed as The analytical results in Sections 3.2 and 3.3 are based on Zthe network shown in Fig. 3, which can be considered to be f ðrÞ ¼ sfXAD ;XCD ðs; rsÞds; ð7Þa part of a larger network. Nodes A and C are in the s>1transmission range of B, and B transmits packets to Dthrough C by employing multihop routing. Hence, D is in where fXAD ;XCD ðs; yÞ is the joint probability density functionthe transmission range of C but not in the transmission (pdf) of XAD and XCD . The joint pdf of XAD and XCD isrange of B. This particular region is shown in Fig. 3 with evaluated in the Appendix and is given by (22). Thedashed lines. We also assume that A has packets for B. The truncated distribution of is given bynetwork in Fig. 3 is used to simplify the analysis yet ( f ðrÞillustrate the important aspects of overlapped transmission. R1 ; 0 < 1; f ðrjr < 1Þ ¼ 0 f ðrÞdr ð8Þ3.2 Interference Due to Secondary Transmission 0; otherwise:Consider first the ad hoc network in Fig. 3 and the time slot Then, from (6) and (7), the pdf of the SINR isduring which node C forwards to D a packet that it has  1 1 received from B in an earlier transmission. The transmis- 1 1 fÀ ðÞ ¼ À1À f À j À < 1 ; ð9Þsion from C to D is a primary transmission, and a possible secondary transmission would be node A transmitting a where is the path-loss exponent.packet to node B. We assume that both nodes A and B are The distribution function FÀ ðÞ of the SIR at D for path-informed of C’s transmission to D. Node A is allowed to loss exponents ¼ 2; 4 are numerically computed andtransmit only if it is not in the transmission range of D. plotted in Fig. 4. Let 0 denote the minimum SIRThis restriction on A’s transmission reduces the amount of requirement for the successful reception of a message. Aninterference at D, but it is still non-negligible. However, A outage event occurs when < 0 . Let
    • denote the outageis allowed to transmit even if it is in the transmission range probability Prð 0 Þ ¼
    • . Since the radio locations are
    • BOPPANA AND SHEA: OVERLAPPED CARRIER-SENSE MULTIPLE ACCESS (OCSMA) IN WIRELESS AD HOC NETWORKS 373 in an upper bound on the probability of scheduling a secondary transmission in a time slot, as the secondary transmission may not be possible if it will interfere with other primary transmissions. 4. Successful reception of the overlapped data at the secondary receiver. The secondary receiver can suc- cessfully receive the message provided that no node in its interference range (with the exception of the primary transmitter) transmits. We do not consider the effect of other secondary transmissions at this secondary receiver, again yielding an upper bound on the number of successful overlapped transmis- sions that can occur in an ad hoc network. Using the example network in Fig. 3, we evaluate the probability of a successful secondary transmission from A to B while C forwards to D the packet it has received from B in an earlier transmission. Based on the above discussion,Fig. 4. Distribution of SIR . the probability of a successful secondary transmission pðSÞ can be bounded byrandom, it may not be possible to achieve
    • ¼ 0 for aparticular 0 . For example, let ¼ 4,
    • ¼ 0:05, and pðSÞ pðF ÞpðT jF Þ; ð10Þ0 ¼ 12 dB. This SIR requirement roughly translates to a where F denotes the event that there is a suitable secondaryvalue of ¼ 0:5. In Fig. 4, we have for
    • ¼ 0:05 an SIR of À1 transmitter (denoted as A in our example network), and TFÀ ð
    • Þ ¼ 4 dB, which is less than the required SIR. The denotes the event that the secondary receiver (denoted as Binterference caused by the secondary transmission can be in our example network) successfully receives the packetcontrolled by using the location information of the nodes in transmitted by the secondary transmitter.choosing the secondary transmitter. Another way to meet The probability of the event F is equivalent to finding athe target SIR requirement without increasing the inter- nontransmitting node that is in the transmission range of Bference to other nodes is to reduce the power of the but not in the transmission range of D. The area of thissecondary transmission. region is3.3 Probability of Secondary Transmission AF ðzÞ ¼  À Al ð1; 1; zÞ; 1 < z 2; ð11ÞIn this section, we evaluate the probability of a secondarytransmission given that there is a primary transmission where Al ðr1 ; r2 ; dÞ is given by (4), and z is the distancethat permits a secondary transmission. With respect to between B and D, whose pdf is given bythe network in Fig. 3, given that C successfully forwards Z ZB’s packet to D, we evaluate the probability of a fXBD ðzÞ ¼ fXCD ;XBD ðy; zjXBC ¼ xÞfXBC ðxÞdydx; ð12Þsuccessful secondary transmission from node A to B. x yThe probability of a successful secondary transmissiondepends on the following factors: where fXCD ;XBD ðy; zjXBC ¼ xÞ and fXBC ðxÞ are given by (18) and (16), respectively. Since the nodes are Poisson dis- 1. Availability of a secondary transmitter (arbitrarily called tributed with node density , the probability pðF Þ of node A here). All the nodes that are in the transmis- finding a secondary transmitter is given by sion range of the secondary receiver (node B) but not Z X 1 in the transmission range of the primary receiver ðAF ðzÞÞn eÀAF ðzÞ pðF Þ ¼ ð1 À pn ÞfXBD ðzÞdz (node D) are identified as potential secondary n¼0 n! z transmitters. One of them is arbitrarily chosen as Z ð13Þ the secondary transmitter. In this analysis, we do not ¼ 1 À eÀAF ðzÞð1ÀpÞ fXBD ðzÞdz; address the issue of how a secondary transmitter is z chosen but investigate the factors that limit the availability of a secondary transmitter. We note that where p is the probability of transmission by a node in a identification of a secondary transmitter does not time slot. The probability pðF Þ of finding a secondary guarantee a successful secondary transmission. transmitter is shown in Fig. 5 for three different node 2. Availability of packets at the secondary receiver. In order densities . It can be seen that for a given probability of to simplify the analysis, we assume that a secondary transmission in a time slot, the probability of finding a transmitter always has packets for the correspond- secondary receiver increases with an increase in the node ing receiver. density. Also, note that for stable operation of the network, 3. Scheduling a secondary transmission. We assume that the probability of transmission p should be less than the once a secondary transmitter is identified, it trans- average number of nodes in the interference range of a mits a packet to the secondary receiver, independent node. For instance, if we assume that the interference range of the state of the medium. This assumption results is twice the transmission range, we have p ð4ÞÀ1 , where
    • 374 IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. 8, NO. 3, MARCH 2009 Fig. 7. Upper bound on the probability of a successful secondaryFig. 5. Probability of finding a secondary transmitter. transmission pðSÞ. is the node density, and the interference range of a node is transmission p ¼ ð4ÞÀ1 , the probability of node B receiv-assumed to be 2 units. With p ¼ ð4ÞÀ1 , pðF Þ is 0.51, 0.74, ing A’s message is 0.53 for all the node densities .and 0.85 for  ¼ 1, 2, and 3, respectively. The upper bound on the probability of successful The probability of successful reception at B of the secondary transmission pðSÞ (cf. (10)) is shown in Fig. 7secondary transmission from A, pðT jF Þ, can be upper for several values of node density . When the probabilitybounded by the probability that no primary transmis- of transmission p ¼ ð4ÞÀ1 , the value of the upper bound issions occur in the nonoverlapping interference regions of 0.27, 0.39, and 0.41 for  ¼ 1, 2, and 3, respectively. ForB and D. The area of this region is given by p ¼ ð8ÞÀ1 , the value of the upper bound is 0.37, 0.54, and AI ðzÞ ¼ 4 À Al ð2; 2; zÞ; ð14Þ 0.61 for  ¼ 1, 2, and 3, respectively. The preceding analysis shows that there is a highwhere Al ðr1 ; r2 ; dÞ is given by (4). Using the same approach probability of successful secondary transmission given thatas in (13), pðT jF Þ can be bounded by there is a primary transmission in a time slot. Although this Z secondary transmission causes interference to several pðT jF Þ eÀAI ðzÞp fXBD ðzÞdz: ð15Þ primary transmissions, this interference can be minimized z by either selecting secondary transmissions that are outside of the primary receiver’s interference range or by reducingThe probability of reception by node B was numerically the power of the secondary transmission. In the followingevaluated, and the pdf is plotted in Fig. 6 for three different sections, we develop a MAC protocol that supports over-node densities . The path-loss exponent is ¼ 4. As the lapped transmission and evaluate its performance undernode density increases, the probability of reception various network scenarios.decreases, which is due to the increase in the interferencearound node B. For an ad hoc network with probability of 4 OVERLAPPED CARRIER SENSE MULTIPLE ACCESS (OCSMA) PROTOCOL The OCSMA protocol is based on the distributed coordi- nated function (DCF) mode of the IEEE 802.11 MAC protocol [1, Section 9.2]. Unless stated explicitly, the terminology used in the following sections corresponds with that in the IEEE 802.11 standard. The design of the OCSMA protocol is best described with the example network in Fig. 8a. The timeline of the protocol for the example network is shown in Fig. 9, and the frame formats are shown in Fig. 10. The operation of the protocol can be divided into five phases as follows. 4.1 Primary Handshaking This phase of the OCSMA protocol is similar to the Request- To-Send (RTS)/ Clear-To-Send (CTS) exchange of the IEEE 802.11 protocol. When a node has data to transmit to another node in its transmission range, it initiates theFig. 6. Upper bound on the probability of reception by node B. handshake by sending an RTS frame. The node that receives
    • BOPPANA AND SHEA: OVERLAPPED CARRIER-SENSE MULTIPLE ACCESS (OCSMA) IN WIRELESS AD HOC NETWORKS 375Fig. 8. Typical frame exchanges in the OCSMA protocol. (a) Ad hoc network. (b) RTS. (c) CTS. (d) PTS. (e) RTT. (f) CTT. (g) DATA. (h) O-DATA.(i) ACK1. (j) ACK2.the RTS sends a CTS frame if it senses the medium to be TAVs for each frame that it receives. The medium isfree. The node initiating the handshake is the primary considered busy if any of the TAVs is set. The TAVs alsotransmitter, and the node that responds to the RTS is the store information regarding the transmitter and receiver ofprimary receiver. All the other nodes that receive the the frame if that information is available. The implementa-handshake set their transmit allocation vectors (TAVs) forthe duration of the transmission. The TAV is similar to the tion of the TAX greatly simplifies the design of the OCSMAnetwork allocation vector (NAV) defined in the IEEE 802.11 protocol, as discussed in later sections. Another importantstandard [1, Section], with a few significant differ- distinction between NAV and TAV is that a node canences as described below. transmit even if the TAV of a node is set. The conditions Each node is equipped with a transmit allocation matrix under which this is possible are discussed later.(TAX) that is responsible for the virtual carrier sense Consider the WANet in Fig. 8a, where at some point ofmechanism. The TAX is an array of several TAVs. Nodes time, node C intends to forward a packet to D that it hasreceiving a valid frame that is not destined for them update received from B in an earlier transmission. C transmits antheir TAV with the information in the Duration/ID field. RTS to D, and D responds with a CTS, as shown in Figs. 8bUnlike the NAV of IEEE 802.11, the TAX allocates a TAV foreach valid frame (not addressed to the receiving node) it and 8c, respectively. The frame formats of RTS and CTSreceives, even if the new TAV value is not greater than any (refer to Fig. 10) in OCSMA are the same as in the IEEEof the current TAVs. Thus, the TAX maintains an array of 802.11 protocol [1, Section 7.2.1].
    • 376 IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. 8, NO. 3, MARCH 2009Fig. 9. Timeline of the OCSMA protocol.4.2 Secondary Handshaking receiver, and the PID field contains the unique ID of theThe secondary handshaking can be thought of as a data frame that is being transmitted to the primary receiver.secondary RTS/CTS exchange to determine the possibility The node receiving the PTS frame is called the secondaryof performing overlapped transmission with the primary receiver. Being a secondary receiver implies that the presenttransmission. Upon receipt of the CTS, the primary node has information regarding the primary transmissiontransmitter sends a Prepare-To-Send (PTS) frame to the and is capable of receiving an overlapped transmission.node from which it received the present data frame in an Upon receipt of the PTS, the secondary receiver ensuresearlier transmission. If the data is locally generated, no PTS that its TAV is set only by the primary transmitter. Noteis sent, and transmission of the data frame starts after an that the TAVs store information regarding the transmitterSIFS [1, Section 9.2.5]. If the PTS is sent, the primary and receiver of any valid frame it receives that is nottransmitter defers the transmission of the data frame until addressed to the receiving node. This is to ensure that therethe completion of the secondary handshaking. are no other transmissions occurring in the range of the Continuing our example using Fig. 8a, after the comple- secondary transmitter except for the primary transmission.tion of the RTS/CTS between C and D, C sends a PTS to B. If this is true, it identifies a suitable partner for secondary transmission as described below.The PTS frame format is shown in Fig. 10. The format is Once the secondary receiver identifies the medium to besimilar to the format of an RTS frame except for the free except for the primary transmission, it generates a listadditional fields Destination Address (DA) and Packet ID of potential partners. The nodes are identified based on the(PID). The DA field contains the address of the primary following criteria: 1. The node should not cause excessive interference to the primary transmission. In this paper, we consider only one of the two approaches described in Section 3.2, in which the secondary receiver knows the locations of the neighboring nodes and uses this information to identify potential candidates for the secondary transmitter. 2. The node should have transmitted a frame to the secondary receiver in an earlier time slot. The information regarding the receipt of frames from all the other nodes is maintained in a cache at the MAC layer. The second condition is based on the heuristic that if a node has transmitted a frame to the secondary receiver in an earlier time slot, it is very likely that there might be more frames destined for the secondary receiver. This ensures that there is a greater probability of secondary transmission for any particular partner. A node is chosen randomly3 from the potential candidates to be the secondary transmitter. The secondary receiver sends a Request-to-Transmit (RTT) frame to the selected secondary transmitter. The 3. Using other approaches such as round-robin scheduling may increaseFig. 10. Frame formats of the OCSMA protocol. the probability of choosing a node with a packet for the secondary receiver.
    • BOPPANA AND SHEA: OVERLAPPED CARRIER-SENSE MULTIPLE ACCESS (OCSMA) IN WIRELESS AD HOC NETWORKS 377format of RTT is similar to the format of RTS except that it cancels the interference and recovers the overlapped data.also contains an additional field, Primary Address (PA), This phase is illustrated in Fig. 8h, which depicts node Bwhich contains the address of the primary transmitter. The receiving an O-DATA frame while canceling out thesecondary transmitter compares the address of the primary interference caused by C’s transmission (primary transmis-transmitter against the transmitter info of the TAVs (if it is sion). Note that the secondary transmission is allowed toavailable), and all the TAVs that are set by the primary terminate Á1 seconds after the end of the primarytransmitter are reset. This ensures that the TAV of the transmission.secondary transmitter is not set by either the RTS or the PTSsent by the primary transmitter. If it finds the medium to be 4.5 Data Acknowledgmentsfree and has a suitable packet to be transmitted, it responds If the DATA and O-DATA frames are successfully received,with a Clear-to-Transmit (CTT) frame whose format is the the primary and secondary transmitters acknowledge thesame as that of CTS (cf. Fig. 10). Transmission of the CTT successful reception of the primary and overlapped dataimplies that the secondary transmitter is capable of frames, as shown in Figs. 8i and 8j, respectively. The formattransmitting overlapped data without causing interference of the ACK frames is the same as in the IEEE 802.11 protocolto any of the transmissions (including the primary [1, Section].transmission) in its communication range. How the nodes then contend for channel access is an In the example network in Fig. 8a, when B receives thePTS from C, it ensures that its TAV is set only by C’s important design consideration that significantly affects thetransmission of RTS to D (refer to the TAV0 setting of B performance of OCSMA. Consider first the primary andshown in Fig. 9). Since B is not in the transmission range secondary receivers. If the DATA and O-DATA packetsof D, it will be able to detect D’s transmission of CTS but were successful, both of these nodes have packets towill not be able to decode it. This would cause B’s TAV1 transmit and will contend for channel access. If the primaryto be set to a duration of Extended Interframe Spacing receiver sends an RTS before the secondary receiver, then it(EIFS) [1, Section], but it would expire before the will become the primary transmitter for that packet, and thePTS frame is received (refer to the TAV1 setting of B in secondary receiver from the previous overlapped transmis-Fig. 9). Based on the selection criteria for choosing a sion will have the appropriate packet to act as a secondarypartner, assume node B chooses node A to send the RTT. transmitter for an overlapped transmission. However, if theWhen A receives the RTT, it ensures that its TAV vector is secondary receiver gains access to the channel before thenot set (refer to the TAV settings of A in Fig. 9). If it senses primary receiver, then an overlapped transmission willthe medium to be free, it responds with a CTT frame. In depend on the availability of appropriate packets furtherthe present example, if we assume that A is in the back in the network. To increase the chance of the primaryinterference range of C (it can sense C’s transmission but receiver contending for the channel first, the primarynot decode it), it would have set its TAV (when C receiver acts as a successful receiver in the IEEE 802.11transmits PTS to B) to EIFS, which would have expired by protocol [1, Section]. To give the secondary receiver athe time A receives the RTT frame. high probability of choosing to defer longer than the4.3 Primary Transmission primary receiver, it will choose a random backoff value inA timer at the primary transmitter is set to expire upon a window that is twice the size of its current contentioncompletion of the secondary handshaking. Note that its window (CW) value, once it senses the channel to be idle.TAV timer will not be set during the transmission of the Next, consider the reception of acknowledgments at thesecondary handshaking (refer to the TAV settings of node C primary and secondary transmitters. Upon reception ofin Fig. 9). We note that this differs from the typical NAV ACK, the primary transmitter resets its CW parameter as inimplementation of the IEEE 802.11 protocol. When the timer the IEEE 802.11 [1, Section] protocol. If it has a packetexpires, it transmits its data frame to the primary receiver. to transmit, the channel access mechanism is the same asIn the example network in Fig. 8a, upon completion of the the mechanism in the IEEE 802.11 protocol. However, thesecondary handshaking, C starts the primary transmission secondary transmitter does not reset its CW. This ensuresto D, as shown in Fig. 8g. The frame format of the DATA that with high probability, the secondary transmitter doesframe (refer to Fig. 10) is the same as in the IEEE 802.11 not contend with the primary transmitter for channelprotocol [1, Section 7.2.2]. access. The CW parameter of the secondary transmitter is4.4 Secondary Transmission reset when it receives an ACK for any DATA frame (andThe secondary transmitter starts its overlapped transmis- not an O-DATA frame) that it transmits later. We observedsion Á0 seconds after the commencement of the primary that in networks with linear flows, this design leads to atransmission (refer to Fig. 9). This overlapped delay Á0 is greater probability of overlapped transmission.designed to allow the secondary receiver to acquire thetiming and phase of the primary transmission, which 5 DESIGN CONSIDERATIONSgreatly simplifies the IC mechanism at the PHY. It doesnot ensure perfect symbol or phase synchronization of the In this section, we discuss various design issues concerningprimary and secondary transmissions at the secondary the OCSMA protocol. In particular, we compare andreceiver. The format of the overlapped data (O-DATA) contrast the OCSMA protocol with the IEEE 802.11 MACframe is the same as the data frame. The secondary receiver protocol, on which it is based.
    • 378 IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. 8, NO. 3, MARCH 2009 TABLE 1 Simulation Setup5.1 Cross-Layer InteractionThe design of the OCSMA protocol involves a greater level Fig. 11. Throughput comparison in a 10-node linear network with TCPof cross-layer interaction compared to the IEEE 802.11 traffic.protocol. For instance, when a node receives an RTT, theMAC needs to interact with the higher layers to determine if with reduced overhead (OCSMA_RO). The performance ofa packet of suitable length can be sent to the secondary this reduced overhead protocol is simulated in the nextreceiver. It is also possible that a packet might need section. The PTS can also be eliminated by including thefragmentation such that the transmission of overlapped information of the PTS frame in the RTS. In this case, thedata is terminated within Á1 seconds of the termination of RTS format will be much different from the format of RTS ofthe primary transmission (refer to the timeline of the the IEEE 802.11 protocol. However, we did not observe anyprotocol in Fig. 9). Similarly, when the secondary receiver significant change in the throughput with this modification.receives a CTT, the MAC needs to indicate to the PHY that Finally, the frame formats of all the frames can be modifiedinterference mitigation will be needed to recover the to reduce the overhead, although we did not evaluate suchoverlapped transmission. Cross-layer interaction is also approaches in this article.needed at the secondary transmitter when identifyingpotential partners for overlapped transmission. 6 SIMULATION RESULTS5.2 Complexity of the Protocol We evaluated the performance of the OCSMA protocolThe OCSMA protocol involves greater computational under different network topologies and traffic conditionscomplexity than the IEEE 802.11 protocol. This is a result using Network Simulator (ns2) [26]. Since we evaluate onlyof employing MUD at the PHY and also increased book- the performance of the MAC protocol, we assume perfectkeeping in the MAC. However, the increase in the IC at the PHY and that the O-DATA packet can becomputational complexity at the MAC is minimal, and we recovered whenever there is an overlapped transmissionbelieve that the design of the protocol can greatly reduce the with the corresponding primary transmission being thecomputational complexity at the PHY in comparison to only source of interference. The simulations are based onother MUD approaches. We also note that the protocol the 1-Mbps DSSS mode (cf. [1, Section 15]) of IEEE 802.11; except where specified, the parameters are given in Table 1.overhead of OCSMA is more than that of IEEE 802.11 The overlapped delay Á0 of 240 s corresponds to aboutbecause of an increase in the number of control frames. 30 bytes of data, which is slightly larger than the sum ofHowever, as the results in Section 6 indicate, this overhead the lengths of the PLCP header and the PLCP preamblebecomes negligible as the size of the data frame increases. (24 bytes) [1, Section 15.2.2]. For other system parameters,5.3 Reduced Overhead the default values of the IEEE 802.11 implementation inThe design of the protocol and the frame formats are to a ns2 are used.large extent compatible with the existing IEEE 802.11 frame We first evaluate the OCSMA protocol in a fixed 10-nodeformats. Hence, they can be integrated with existing IEEE linear network, with a source and destination located at802.11-based wireless networks with minimal changes. The either end of the network. The nodes are placed at regularoverhead of the OCSMA protocol can be reduced consider- intervals, with adjacent nodes being in the communicationably if no such conformity is required. For instance, the CTT range of each other and nodes two hops apart being in thepacket can be eliminated without a significant penalty on interference range of each other. The transmission power ofthe throughput. The elimination of the CTT packet results in the secondary transmission is the same as that of thereduced protocol overhead but increases the power primary transmission. The traffic model is based on FTPconsumption at the PHY of the secondary receiver since “simulated application,” in which the TCP queue is neverIC has to be turned on more often. In addition, the DA and empty. TCP is used for flow control, with a maximumPA fields of the PTS and RTT frames, respectively, can be window size of 32. The end-to-end throughputs of theeliminated without any significant performance penalty network under the OCSMA, OCSMA_RO, and IEEE 802.11(refer to Fig. 10). We call this protocol the OCSMA protocol MAC protocols are shown in Fig. 11.
    • BOPPANA AND SHEA: OVERLAPPED CARRIER-SENSE MULTIPLE ACCESS (OCSMA) IN WIRELESS AD HOC NETWORKS 379 TABLE 2 Comparison of Events at the MAC-Layer for a 10-Node Linear Network with Packet Size 400 Bytes We observe that the throughput of the IEEE 802.11 MACprotocol increases until the data packet length reaches1,000 bytes, beyond which it starts decreasing. However, Fig. 12. Effect of TCP maximum CW on throughput gain of OCSMA andthe throughput of both OCSMA and OCSMA_RO increases OCSMA_RO in a 10-node linear network.until the packet length reaches 1,400 bytes, beyond whichthe throughput decreases. The OCSMA protocol provides busy by the secondary transmitter. To investigate the reasonthroughput gains of 4 percent to 39 percent over the range for the low CTT-to-RTT ratio, we created a No-Packet-to-of packet lengths shown in Fig. 11. The maximum Transmit (NPT) frame. The secondary receiver transmits anthroughput under OCSMA is achieved for a packet length NPT in response to an RTT if it finds the medium to be freeof 1,400 bytes, at which point it provides 21 percent but does not have a suitable packet for the secondarythroughput gain over IEEE 802.11. Similarly, the reduced receiver. The NPT frame was introduced only for simula-overhead version of OCSMA (OCSMA_RO) provides tion purposes and is not a part of the OCSMA protocol. Wethroughput gains of 11-40 percent over the packet lengths did not observe any adverse effect on the system through-simulated and provides a throughput gain of 28 percent put from the inclusion of the NPT frame.over IEEE 802.11 for a packet length of 1,400 bytes. As can be seen in Tables 2 and 3, the ratio of CTT to The MAC-layer events across the network for all three NPT is about 54 percent and 41 percent for packet lengthsprotocols are tabulated in Tables 2 and 3 for data packet of 400 bytes and 1,800 bytes, respectively. This indicateslengths of 400 and 1,800 bytes, respectively. The average that the full potential of the overlapped transmissions israte of RTS frames received for OCSMA and OCSMA_RO is not realized due to lack of suitable packets. The ratio ofhigher than the rate of RTS frames received in the case of overlapped data (O-DATA) packets received to that ofIEEE 802.11. We observe that the proportion of the average data (DATA) packets is 19.5 percent and 17.9 percent forrate of reception of PTS to that of RTS is very high, OCSMA and OCSMA_RO, respectively, for a packet sizeindicating that there is a very high probability of an of 400 bytes. The ratio is 17.9 percent and 17.3 percent,overlapped transmission from the perspective of the respectively, when the packet size is increased toprimary transmitter. However, the ratio of the reception 1,800 bytes. It is also worth noting that the averageof CTT to that of RTT is significantly lower, which indicates number of collisions at the MAC layer in the case of boththat the actual number of O-DATA transmissions is OCSMA protocols is higher than that of the IEEE 802.11significantly less than the potential overlapped transmis- protocol. We observed that these collisions are mainly duesions. This might be due to the lack of suitable packets at the to the control frames during the secondary handshakingsecondary transmitter or the medium being perceived as (RTT and CTT), causing collisions in the vicinity of the secondary transmitter. However, these collisions are offset by the increase in throughput due to overlapped TABLE 3 transmissions. Comparison of Events at the MAC-Layer for a 10-Node As TCP may significantly affect the availability of Linear Network with Packet Size 1,800 Bytes packets for secondary transmission, we consider the impact of the maximum TCP window size on the throughput of OCSMA and OCSMA_RO protocols. The throughput gains of OCSMA and OCSMA_RO protocols over IEEE 802.11 as a function of the TCP window size is shown in Fig. 12. The throughput of OCSMA protocols is less than that of IEEE 802.11 for TCP window sizes two and four. This is due to the unavailability of packets at the secondary transmitters to perform overlapped transmissions. As the TCP window size increases, the throughput gains of both OCSMA and OCSMA_RO increase, providing a maximum gain of
    • 380 IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. 8, NO. 3, MARCH 2009 TABLE 4 Events at the MAC Layer in a 10-Node Linear Network under OCSMA Protocol21 percent and 30 percent, respectively, for window sizesgreater than 16. The MAC-layer events for the OCSMA Fig. 14. Throughput comparison in a 10-node linear network with CBRprotocol for several window sizes are shown in Table 4. It traffic.can be seen that the ratio of the CTT to NPT framesincreases with an increase in the TCP window size; of packets delivered by IEEE 802.11. However, underhowever, the collision rate also increases. OCSMA and OCSMA_RO, the decline in the throughput Since the collision rate for OCSMA protocols is higher is more gradual, and the throughput gains provided bythan that of IEEE 802.11, we next analyze the impact of the OCSMA protocols over IEEE 802.11 are significant.STA Short Retry Count (ssrc) and STA Long Retry Count Next, we consider the effect of multiple flows in a linear(slrc) limits [1, Section] on the throughput of OCSMA network. Three sources and three destinations are placed atand OCSMA_RO. Fig. 13 shows the throughput gains of either end of a 10-node linear network, and the traffic typeOCSMA and OCSMA_RO over IEEE 802.11 for various is CBR. The throughput gains of OCSMA and OCSMA_ROvalues of ssrc and slrc limits. The TCP window size is 32, over IEEE 802.11 with CBR traffic and multiples flows in aand the packet size is 1,400 bytes. The values of the slrc and linear network are shown in Fig. 15. The packet arrival ratessrc limits used are shown in parenthesis. We observe that indicates the common rate at which packets arrive at each ofthe throughput gain of OCSMA and OCSMA_RO increasemonotonically with an increase in ssrc and slrc limits. The the sources. It can be observed that even in the presence ofincrease in the collisions in the case of OCSMA protocols multiple flows, OCSMA and OCSMA_RO provide signifi-are offset by the increase in the ssrc and slrc limits. cant gains over IEEE 802.11 in a 10-node linear network. The throughputs for the 10-node linear network with Next, we vary the number of nodes in the linear network.constant bit rate (CBR) traffic is shown in Fig. 14 for several FTP “simulated application” traffic with a packet size ofpacket arrival rates. The packet size is 1,000 bytes. We 1,400 bytes was simulated. The end-to-end throughputobserve that the throughput of all three protocols is the gains of the OCSMA and OCSMA_RO protocols oversame until the packet arrival rate reaches 20 packets/s. As IEEE 802.11 are shown in Fig. 16 as a function of thethe packet rate increases, there is a dramatic fall in the rate number of nodes in the linear network. It can be seen thatFig. 13. Effect of short and long retry counts on throughput gains of Fig. 15. Throughput comparison in linear network with multiple CBROCSMA and OCSMA_RO in a 10-node linear network. flows.
    • BOPPANA AND SHEA: OVERLAPPED CARRIER-SENSE MULTIPLE ACCESS (OCSMA) IN WIRELESS AD HOC NETWORKS 381 Fig. 17. Throughput gain in a random network with mobility.Fig. 16. Effect of varying the number of nodes in a linear network on transmissions is small when there is high mobility in thethroughput gain of OCSMA and OCSMA_RO. network. However, in certain scenarios, OCSMA provides a significant gain over IEEE 802.11.OCSMA and OCSMA_RO provide maximum throughput We also observed that in certain scenarios, the perfor-gains of 72 percent and 77 percent, respectively, when the mance of the OCSMA protocol is much worse than that ofnetwork consists of six nodes. The gain decreases with an the IEEE 802.11 protocol. In these scenarios, the increase inincrease in the number of nodes in the network. In a 30- the spatial reuse due to overlapped transmissions in-node network, the throughput gains of OCSMA and creased the number of interflow contentions, whichOCSMA_RO are 16 percent and 10 percent, respectively. affected the throughput of the network. We are currentlyFor a fixed TCP window size, under OCSMA and investigating the interflow contention issues associatedOCSMA_RO, we observed that as the size of the linear with the OCSMA protocol.network increases, the ratio of O-DATA to DATA framesdecreases, and the collision rate increases. The increase in 7 CONCLUSIONthe spatial reuse provided by OCSMA (and OCSMA_RO) isoffset by the increase in the collisions in the network. In this paper, we developed overlapped transmission We next consider a random topology with 50 nodes schemes to enhance the spatial reuse and throughput ofdistributed in a 1,500 m  1,500 m square. This scenario wireless networks. By taking advantage of a priori knowl-corresponds to an average node density of four nodes in a edge of the interfering packet, the receiver can employ a simplified IC scheme to receive a packet in the presence ofcircle of radius equal to the transmission range of a node interference. We analyzed some of the factors that limit the(set to 250 m). The mobility model chosen is the random use of overlapped transmissions in an ad hoc network. Wewaypoint model, which is the default model in ns2. The developed the OCSMA protocol based on the IEEE 802.11nodes move with a speed that is uniformly distributed in MAC protocol to support overlapped transmissions in athe interval ½0; max speedŠ, where we consider different wireless network. Network simulations employing OCSMAvalues of max_speed. Twenty TCP connections were protocol and its reduced overhead variant, OCSMA_RO,randomly generated with packet size 1,400 bytes, and the show that the end-to-end throughput can be improved byrest of the system parameters are given in Table 1. The as much as 77 percent over the IEEE 802.11 MAC protocolthroughput gains of OCSMA and OCSMA_RO over IEEE in a linear network with TCP traffic. Under CBR traffic,802.11 are averaged over 500 instantiations of the random OCSMA and OCSMA_RO are more robust to the trafficnetwork. The performance gain of OCSMA protocols over load and multiple flows than the IEEE 802.11 protocol. In aIEEE 802.11 as a function of the maximum speed of the random network with 50 nodes and 20 TCP connections, thenodes in the network is shown in Fig. 17. We observe that OCSMA and OCSMA_RO protocols provide an averagethe throughput gains of the OCSMA protocols decrease as throughput gain of 13 percent and 17 percent, respectively,the mobility in the network increases. When there is no when there is no mobility in the network. The throughputmobility in the network, OCSMA provides an average gain of the OCSMA protocols decreases with an increase inthroughput gain of about 13 percent with a standard the mobility in the network. Although the average gaindeviation of 0.11. The high standard deviation indicates that provided by the OCSMA protocols in high-mobility condi-in certain scenarios, OCSMA provides significant gains over tions is only 5 percent to 7 percent, the throughput gain inIEEE 802.11. Similarly, OCSMA_RO provides an average certain scenarios can be much higher.throughput gain of 17 percent with a standard deviation of0.12. Under high mobility ðmax speed ¼ 20 m=sÞ, OCSMA APPENDIXand OCSMA_RO provide average throughput gains of5 percent ðstandard deviation ¼ 0:05Þ and 7 percent DERIVATION OF THE JOINT PDF OF XAD , XCDðstandard deviation ¼ 0:05Þ, respectively. The results indi- In order to evaluate the joint distribution of XAD and XCD ,cate that in general, the throughput gain from overlapped we first look at the relative positions of nodes A and D with
    • 382 IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. 8, NO. 3, MARCH 2009 The joint distribution of XAD and XCD is given by (22). ZZ fXAD XCD ðs; yÞ ¼ fXAD ðsjx; y; zÞfXCD ;XBD ðy; zjxÞfBC ðxÞdzdx: x z ð22Þ ACKNOWLEDGMENTS The material in this paper was presented in part at the IEEE Military Communications Conference held on October 2007 in Orlando, Florida. This work was supported by the US National Science Foundation under Grant CNS-0626863 andFig. 18. Circle-circle intersection for analysis. by the US Air Force Office of Scientific Research under Grant FA9550-07-10456.respect to node B. Note that A is uniformly distributed in aunit circle with B at the center. The density function of XAB ,the distance between A and B, is given by REFERENCES [1] IEEE 802.11 WLAN Committee, IEEE Standard 802.11-2007, 2x; 0 x 1; IEEE Standard for Information Technology-Telecommunications and fXAB ðxÞ ¼ ð16Þ Information Exchange between Systems-Local and Metropolitan Area 0; otherwise: Networks-Specific Requirements—Part 11: Wireless LAN MediumSimilarly, node C is also uniformly distributed within the Access Control (MAC) and Physical Layer (PHY) Specifications, IEEE, June 2007.transmission range of B, and hence, the pdf of XBC is the [2] J.Q. Bao and L. Tong, “Performance Analysis of Slotted Alohasame as that of XAB . Node D is in the transmission range of Random Access Ad-Hoc Networks with Multipacket Recep- tion,” Proc. IEEE Military Comm. Conf., vol. 1, pp. 251-255, Nov.C but not in the transmission range of B. Hence, it is 1999.uniformly distributed in the shaded region in Fig. 3. The [3] Q. Zhao and L. Tong, “Semi-Blind Collision Resolution in Randomjoint conditional distribution of XCD and XBD given XBC Access Wireless Ad Hoc Networks,” IEEE Trans. Signal Processing, vol. 48, pp. 2910-2920, Oct. 2000.can be derived in a similar fashion and is given by (refer to [4] M.M.K. Howlader and B.D. Woerner, “System Architecture forFig. 18) Implementing Multiuser Detector within an Ad-Hoc Network,” Proc. IEEE Military Comm. Conf., vol. 2, pp. 1119-1123, Oct. ( y2 ÀA ð1;y;xÞ 2001. l ÀAl ð1;1;xÞ ; z > x þ y; [5] R.M. de Moraes, R. Sadjadpour, and J.J. Garcia-Luna-Aceves, “A FXCD ;XBD ðy; zjXBC ¼ xÞ ¼ Al ðz;y;xÞÀAl ð1;y;xÞ New Communication Scheme for MANETs,” Proc. Int’l Conf. ÀAl ð1;1;xÞ ; z < x þ y; Wireless Networks, Comm. and Mobile Computing, vol. 2, pp. 1331- ð17Þ 1336, June 2005. [6] C. Comaniciu, N.B. Mandayam, and H.V. Poor, Wireless Networks:where Al ð1; y; xÞ is given by (4). The conditional joint Multiuser Detection in Cross-Layer Design. Springer, 2005. [7] V. Naware, G. Mergen, and L. Tong, “Stability and Delay of Finite-density function is given by (18). User Slotted ALOHA with Multipacket Reception,” IEEE Trans. Information Theory, vol. 51, no. 7, pp. 2636-2656, July 2005.fXCD ;XBD ðy; zjXBC ¼ xÞ [8] G. Mergen and L. Tong, “Receiver Controlled Medium Access in 8 Multihop Ad Hoc Networks with Multipacket Reception,” Proc. > ÀA 1 4yz pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ; > < l ð1;1;xÞ 0 x; y 1; IEEE Military Comm. Conf. (MILCOM ’01), vol. 2, pp. 1014-1018, ðxþyþzÞðxþyÀzÞðxÀyþzÞðÀxþyþzÞ Oct. 2001. ¼ 1<z x þ y; > > [9] L.L. Xie and P.R. Kumar, “A Network Information Theory for : Wireless Communication: Scaling Laws and Optimal Opera- 0; otherwise: tion,” IEEE Trans. Information Theory, vol. 50, no. 5, pp. 748-767, ð18Þ May 2004. [10] S. Boppana and J.M. Shea, “Overlapped Transmission in Wireless Since A is uniformly distributed in a unit circle with B at Ad Hoc Networks,” Proc. Int’l Conf. Comm. Circuits and Systems (ICCCAS ’06), vol. 2, pp. 1309-1314, June 2006.the center, the conditional distribution of XAD , the distance ´ [11] D.S. Lun, M. Medard, R. Koetter, and M. Effros, On Coding forbetween nodes A and D, is given by Reliable Communication over Packet Networks, Arxiv cs.IT/0510070, 1 preprint, 2005. A ð1; s; zÞ; s þ 1 > z; [12] S. Chachulski, M. Jennings, S. Katti, and D. Katabi, “Trading FXAD ðsjXBD ¼ zÞ ¼  l ð19Þ Structure for Randomness in Wireless Opportunistic Routing,” 0; otherwise; Proc. ACM SIGCOMM ’07, pp. 169-180, Aug. 2007. [13] C. Fragouli and E. Soljanin, “Information Flow Decomposition forand the pdf is given by Network Coding,” IEEE Trans. Information Theory, vol. 52, no. 3, (   pp. 829-848, Mar. 2006. 2s À1 s2 þz2 À1  cos ; s þ 1 > z; [14] E. Fasolo, M. Rossi, J. Widmer, and M. Zorzi, “On MAC fXAD ðsjXBD ¼ zÞ ¼ 2sz ð20Þ Scheduling and Packet Combination Strategies for Practical 0; otherwise: Random Network Coding,” Proc. IEEE Int’l Conf. Comm. (ICC ’07), pp. 3582-3589, June 2007.Note that XAD is conditionally independent of XCD and [15] ´ S. Katti, H. Rahul, W. Hu, D. Katabi, M. Medard, and J. Crowcroft,XBC given XBD . Hence, “XORs in the Air: Practical Wireless Network Coding,” Proc. ACM SIGCOMM ’06, pp. 243-254, Sept. 2006. fXAD ðsjXCD ¼ x; XBC ¼ y; XBD ¼ zÞ ¼ fXAD ðsjXBD ¼ zÞ: ð21Þ [16] R. Koetter, Network Coding Bibliography, http://tesla.csl.uiuc.edu/ ~koetter/NWC/Bibliography.html, 2007.
    • BOPPANA AND SHEA: OVERLAPPED CARRIER-SENSE MULTIPLE ACCESS (OCSMA) IN WIRELESS AD HOC NETWORKS 383[17] Y. Wu, P.A. Chou, Q. Zhang, K. Jain, W. Zhu, and S.-Y. Kung, John M. Shea received the BS degree (with “Network Planning in Wireless Ad Hoc Networks: A Cross- highest honors) in computer engineering from Layer Approach,” IEEE J. Selected Areas in Comm., vol. 23, Clemson University in 1993 and the MS and no. 1, pp. 136-150, Jan. 2005. PhD degrees in electrical engineering from[18] S. Deb, M. Effros, T. Ho, D.R. Karger, R. Koetter, D.S. Lun, Clemson University in 1995 and 1998, re- ´ M. Medard, and N. Ratnakar, “Network Coding for Wireless spectively. He is currently an associate Applications: A Brief Tutorial,” Proc. Int’l Workshop Wireless professor of electrical and computer engineer- Ad-Hoc Networks (IWWAN ’05), May 2005. ing, in the Department of Electrical and[19] Y.E. Sagduyu and A. Ephremides, “On Joint MAC and Network Computer Engineering, University of Florida, Coding in Wireless Ad Hoc Networks,” IEEE Trans. Information Gainesville. Prior to that, he was an assistant Theory, vol. 53, no. 10, pp. 3697-3713, Oct. 2007. professor at the University of Florida from July 1999 to August 2005[20] P.A. Chou and Y. Wu, “Network Coding for the Internet and and a postdoctoral research fellow at Clemson University from Wireless Networks,” IEEE Trans. Signal Processing, vol. 24, no. 5, January 1999 to August 1999. He was a research assistant in the pp. 77-85, Sept. 2007. Wireless Communications Program, Clemson University, from 1993[21] Y.E. Sagduyu and A. Ephremides, “Cross-Layer Optimization to 1998. He is currently engaged in research on wireless commu- of MAC and Network Coding in Wireless Queueing Tandem nications with emphasis on error-control coding, cross-layer protocol Networks,” IEEE Trans. Information Theory, vol. 54, no. 2, design, cooperative diversity techniques, and hybrid ARQ. Dr. Shea pp. 554-571, Feb. 2008. is an editor for the IEEE Transactions on Wireless Communications.[22] S. Zhang, S.C. Liew, and P. Lam, “Physical-Layer Network He was an associate editor for the IEEE Transactions on Vehicular Coding,” Proc. ACM MobiCom ’06, pp. 358-365, Sept. 2006. Technology from 2002 to 2007. He was selected as a finalist for the[23] J. Zhang, K. Cai, K.B. Letaief, and P. Fan, “A Network Coding 2004 Eta Kappa Nu Outstanding Young Electrical Engineer Award. Unicast Strategy for Wireless Multi-Hop Networks,” Proc. IEEE He was a US National Science Foundation fellow from 1994 to Wireless Comm. and Networking Conf. (WCNC ’07), pp. 4221-4226, 1998. He received the Ellersick Award from the IEEE Communica- Mar. 2007. tions Society for the Best Paper in the Unclassified Program of the[24] S. Katti, S. Gollakota, and D. Katabi, “Embracing Wireless IEEE Military Communications Conference in 1996. He is a member Interference: Analog Network Coding,” Proc. ACM SIGCOMM ’07, of the IEEE. pp. 397-408, Aug. 2007.[25] Y. Hao, D. Goeckel, Z. Ding, D. Towsley, and K.K. Leung, “Achievable Rates of Physical Layer Network Coding Schemes on . For more information on this or any other computing topic, the Exchange Channel,” Proc. IEEE Military Comm. Conf. please visit our Digital Library at www.computer.org/publications/dlib. (MILCOM ’07), pp. 1-7, Oct. 2007.[26] The ns Manual, K. Fall and K. Varadhan, eds., chapter 16, The VINT Project, 2003. Surendra Boppana received the BTech degree in electronics and communication engineering from the Indian Institute of Technology (IIT), Guwahati, India, in 2003 and the MS and PhD degrees in electrical and computer engineering from the University of Florida, Gainesville, in 2005 and 2008, respectively. From May 2006 until January 2007, he was a graduate technical intern with the Communications Circuit Labora- tory, Intel Corp., Hillsboro, Oregon. He iscurrently a senior engineer at Qualcomm, Inc., San Diego. His researchinterests include wireless communications, information theory, andcross-layer design. He is a student member of the IEEE.