Cross-Layer Directional Antenna MAC Protocol for Wireless Ad Hoc Networks

Transcription

1 Cross-Layer Directional Antenna MAC Protocol for Wireless Ad Hoc Networks Hrishikesh Gossain, Carlos Cordeiro, Tarun Joshi and Dharma P. Agrawal OBR Center for Distributed and Mobile Computing Department of ECECS, University of Cincinnati Cincinnati, OH 45-3 (hgossain, joshit, Philips Research Wireless Communications and Networking Department Briarcliff Manor, New York 5, USA Abstract In this paper we propose a Directional Antenna Medium Access () protocol and its enhanced version called E (enhanced ), which takes advantage of the benefits offered by directional antennas. Both of these schemes have been inspired by the IEEE 8. MAC, with major enhancements including a new neighbor discovery scheme, an optimized circular directional transmission of RTS and CTS to prevent the hidden node problem, reduce collisions and decrease node deafness, and also a multi-buffer management scheme. A pair of communicating nodes using these schemes simultaneously transmits the circular directional RTS and CTS only to those sectors with neighbors, hence reducing overall communication delay and enhancing throughput. The main difference between and E lies in the way buffering is provided in the MAC layer. Unlike which uses a single MAC buffer for all antenna beams as in IEEE 8., E employs separate buffers for each of the antenna sectors and introduces an integrated network and MAC cross-layer design. This helps in eliminating the self induced blocking problem prominent in existing directional MAC protocols. We have compared, E, IEEE 8. and two recently proposed directional MAC protocols, and results show that and E perform better than existing MAC protocols in the majority of the scenarios investigated, while E is observed to perform best. We also point out that the performance usually depends on the network topology and traffic pattern. Keywords: 8., MAC, Directional Antenna, Spatial Reuse, Cross Layer, Wireless Ad Hoc Networks.. INTRODUCTION Most of the existing research on ad hoc networks typically assumes the use of omni directional antennas by all nodes. Such an example is the IEEE 8. medium access control (MAC) [] protocol which appears to efficiently solve the issues of this type of environment. However, due to the omni directional nature of transmissions, network capacity is considerably limited. For example, the distribution of energy in all directions other than the intended direction not only generates unnecessary interference to other nodes, but also decreases the potential range of transmissions. With directional communications, on the other hand, both range and spatial reuse can be substantially enhanced, by having nodes concentrate transmitted energy only towards their destination s direction. On the receiving side, directional antennas enable a node to selectively receive signals only from the certain desired direction, thereby increasing the signal to interference and noise ratio (SINR). Therefore, traditional MAC protocols which have been designed under the omni directional assumption [, ] are no longer suitable for use over directional antennas. The design of an efficient MAC protocol for directional antennas is then a crucial issue and needs further investigation. In directional antennas, new types of hidden node problems arise [3]. In addition, issues such as node deafness and the determination of neighbors locations have to be properly handled [4]. A detailed study that analyzes several important aspects regarding directional antennas and the factors that affect them can be found in [3]. In this paper, we initially introduce a directional antenna medium access () control protocol for use over directional antennas. addresses the hidden node problem and node deafness by employing a novel scheme of selective circular directional transmission of RTS and CTS, where these packets are transmitted only through the antennas with neighbors. For that, employs a self-learning algorithm to determine the presence or absence of nodes in given directions. Moreover, we point out the deficiencies in existing MAC protocols proposed for directional antennas and show how overcomes such problems. We then present and discuss a new and never before identified problem which limits the performance of existing directional antenna MAC protocols which all rely on a single MAC buffer, hereby defined as the self induced blocking problem (Section 6). To address this problem, we propose the enhanced (E) protocol which is based on a cross-layer MAC and routing design, wherein the MAC provides separate buffers for each antenna beam together with a simple and

2 enhanced directional routing protocol. In doing so, we not only overcome the self induced blocking problem but also increase the throughput by avoiding the time wasted due to a blocked antenna beam preventing the transmission of packets in other non-blocked directions. The rest of this paper is organized as follows. In Section we first discuss the related work on MAC protocols for directional antennas. The antenna model and a glimpse of IEEE 8. are then given in Section 3. Section 4 thoroughly describes our proposed protocol and how it overcomes the problems discussed. Comprehensive simulation study and comparison of with three other MAC protocols including IEEE 8. is given in Section 5. Next, Section 6 describes the E scheme, followed by a performance comparison of E with and other directional protocols in Section 7. Finally, this paper is concluded in Section 8 highlighting some open problems and future research plan.. RELATED WORK Most of the research in the area of directional antennas has focused broadband and cellular networks [5, 6, 7]. In the context of wireless ad hoc networks, research is still at its infancy. In general for ad hoc networks, two models for MAC protocols for directional antennas can be identified. In the first model [8], each node is equipped with M antennas whose orientations can be maintained at any time, regardless of the node s movement. In this model, it is assumed that nodes have directional reception capability, i.e., they can activate the antenna pointing to the direction of the desired destination while deactivating antennas in all other directions. Most recent research adopts this model [3, 4, 8]. In the second model [9], antennas are always active for receiving and thus transmissions to different antennas results in collision. Some MAC proposals for directional antennas assume this model []. In this work, we consider the first model and elaborate on the same in the next section. In [8], a variation of RTS/CTS mechanism of IEEE 8. adapted for use with directional antennas is given. This protocol sends the RTS and CTS packets omni-directionally in order to enable the transmitter and receiver to locate each other, and sends the DATA and ACK packets in directional mode. A MAC protocol that sends a directional RTS and an omni directional CTS is presented in []. Here, it is assumed that the transmitter knows the receiver s location, so that it can send the RTS directionally. In case location information is not available, the RTS is transmitted in omni mode in order to find the receiver. In [] it is proposed the use of Directional Virtual Carrier Sensing in which directional RTS and CTS transmissions are employed. Here, it is assumed that the transmitter knows the receiver s location. Similarly to [], RTS are transmitted omni-directionally in case location information is not available. Finally, [3] studies the problems that appear using directional antennas and proposes a MAC protocol to take advantage of the higher gain obtained by directional antennas. This protocol employs a scheme of directional multihop RTS transmissions so as to establish directional-directional (DD) links between the transmitter and receiver. An assumption of this scheme is that the transmitter must know the entire route to the intended receiver so that the RTS packet can be routed. In addition, although [3, ] uses directional transmissions only, they do not solve the issues of increased instances of hidden terminal problem [8, 9], node deafness and the determination of neighbors location. The first two problems are thoroughly studied in [3], although a solution is not provided. The third problem originates from the fact that a node has to know through which antenna it can communicate with the intended receiver before transmitting a directional RTS. In [3, ], nodes location is assumed to be known beforehand, while [] assumes nodes location can be determined with the assistance of an additional hardware such as GPS. While this assumption simplifies the protocol design, it may render the protocol unsuitable for implementation. The concept of Directional Virtual Carrier Sense (DVCS) and Directional Network Allocation Vector (DNAV) mechanisms are proposed in [3, 4,, ], and is similar to the DNAV concept employed in our protocol discussed later. In [3, 9] a protocol called Directional MAC (DMAC) is proposed that employs directional transmission of RTS and CTS. Similar to the previous schemes, it assumes nodes locations are known a priori. This protocol also suffers from node deafness and hidden node problems as described in [4]. To overcome the shortcomings in DMAC, [4] employs a scheme of circular directional transmission of RTS with a single directional CTS packet. We refer to this scheme as Circular RTS MAC (). While does not assume prior neighbor s location availability, it does not satisfactorily prevent node deafness and collisions. First of all, only prevents node deafness in the neighborhood of the transmitter. A more serious problem with is in the design of its RTS/CTS handshake. For example, if the destination node does not reply back with a CTS (due to a collision), nodes in the neighborhood of the transmitter which correctly receive the circular RTS will not be able to initiate any transmission as their DNAV is set. Clearly, this degrades the network capacity. Another limitation in is transmission of circular directional RTS through empty sectors. By empty sectors we mean sectors which do not have any neighbors. As shown in later in our simulation studies, this overhead has an

3 increasingly larger impact as the number of antenna beams is increased and, therefore, is not an efficient and scalable solution. 3. PRELIMINARIES 3. The Antenna Model We have implemented a complete and flexible directional antenna module at the Network Simulation (NS version.6) [7]. This model possesses two separate modes: Omni and Directional. Similar to [3], this may be seen as two separate antennas: an omni-directional and a steerable single beam antenna which can point towards any specified direction. In principle, both the Omni and Directional modes may be used to transmit or receive signals. However, in our proposed protocols, the Omni mode is used only to receive signals, while the Directional mode may be used for transmission as well as reception. This way, both transmitter and receiver take advantage of the increased coverage range provided by beamforming. This feature is included in our protocol and evaluated in our implementation. In Omni mode, a node is capable of receiving signals from all directions with a gain of G O. While idle (i.e., neither transmitting nor receiving), a node stays in Omni mode when using our proposed protocols. By employing selection diversity, as soon as a signal is sensed a node can detect the antenna through which the signal is strongest and goes into the Directional mode in this particular antenna. In Directional mode, a node can point its beam towards a specified direction with gain G d (with G d typically greater than G O ), using an array of antennas called array of beams. Due to the higher gain, nodes in Directional mode have a greater range in comparison to Omni mode. In addition, the gain is proportional to number of antenna beams given that more energy can be focused on a particular direction, thus resulting in increased coverage range in this particular direction. In order to perform a broadcast, a transmitter may need to carry out as many directional transmissions as there are antenna beams so as to cover the whole region around it. This is called sweeping. In the sweeping process, we assume there is a negligible delay in beamforming in the various directions. Figure illustrates the antenna model we consider in this work. A similar model has also been considered in the existing literature [3, 4, 8, 9]. In this model, the node provides coverage around it by a total of M nonoverlapping beams. The beams are numbered from through M starting at the three o clock position and running counter clockwise. A node can receive and transmit in any of these M antenna beams. Directional coverage area 4 3 (3) Circular RTS () RTS () CTS Omni coverage area M M Antenna Elements in which RTS is sent S (4) DATA (5) ACK R (3) Circular CTS Antenna Elements in which CTS is sent Figure The antenna model Figure RTS/CTS/DATA/ACK exchange in 4. THE PROPOSED PROTOCOL The protocol aims to effectively overcome the limitations found in both DMAC and by utilizing a new combination of adaptive mechanisms. To take advantage of the increased gain obtained by directional antennas, all transmissions in are directional. Secondly, does not rely on prior availability of neighbors location, while it learns its neighbors with time as communication takes place. To prevent node deafness and the new types of hidden node problems aforementioned, employs circular directional transmission of both RTS and CTS. However, contrary to, only transmits the RTS and CTS packets through the antenna beams with neighbors. In order to accomplish that, employs an adaptive mechanism where it learns and caches information about those sectors with neighbors. Initially, performs similar to by sweeping through all antenna beams. However, as responses are received, it collects and caches neighboring information. To make the protocol simple in implementation, design has been inspired by the IEEE 8. MAC.

4 4. Determination of Neighbors Location carries out a continuous process of learning to determine through which antenna a given neighbor can be reached. For this, relies on the very basic characteristics common to the majority of routing protocols employed over ad hoc networks [4, 5, 6,, ]: the use of broadcasting. These protocols either employ a form of periodic one-hop hello packets, or at least they flood the network with route request packets before data packets can be sent. If the routing protocol employs any form of hello packets, all network nodes will eventually determine all their neighbors during the learning phase, given that hello packets are periodically transmitted. More important than determining all particular neighbors of a node, this process also allows a node to determine if some of its sectors have any neighbors at all. On the other hand, if the protocol is not based on hello packets but uses flooding as a means to discover destination nodes, proceeds in a similar manner as described while the only difference resides on the way the route discovery procedure is carried out. 4. The Optimized Circular Directional RTS and CTS In we optimize the RTS and CTS transmission by sending these control packets only through those sectors where nodes are found. This information is obtained through the neighbors location procedure described in the previous subsection. Assuming that the number of antenna beams nodes have is equal to M and that the direction D R node S uses to communicate with node R is currently idle, the sender node S and receiver node R will transmit K S and K R RTS and CTS packets, respectively, where K M and K M. s R Another important aspect in the design of is that the first RTS sent is always transmitted in the sector where its intended neighbor is located, and the circular directional RTS and CTS procedure is only initiated once the RTS/CTS handshake is successfully completed. We do this to overcome one of the limitations in that initiates the circular directional transmission of the RTS packet (thus reserving the channel) before the sender node knows if any of its RTS has or will ever be correctly received by its intended destination node. Upon reception of an RTS packet as shown by step () in Figure, the receiver proceeds similar to IEEE 8.. That is, it waits for a period of time equal to SIFS and sends back a CTS as shown by step (). Only after the RTS/CTS handshake is completed and the channel is reserved in their direction, will both sender and receiver nodes simultaneously initiate the circular directional transmission of their RTS and CTS packets, respectively, to inform their neighboring nodes. Figure illustrates the simultaneous circular transmissions through step (3), where we note that nodes S and R their RTS and CTS through sectors where neighbors can be found. To synchronize the sender and receiver to carry out DATA transmission, the sender node S includes in its RTS its value of K, that is, K S, and the receiver node R includes K R in its CTS back to node S. Through K S, node R is able to determine the exact point in time when node S will have finished its circular directional transmission of RTS and hence will start transmitting DATA. Similarly, with K R node S can precisely tell the moment node R will be ready and waiting for DATA transmission. Steps (4) and (5) in Figure depict the DATA/ACK transmission. 4.3 The Directional NAV (DNAV) DNAV [3,, 3] is an extension to the NAV concept used in IEEE 8. for directional antennas. Essentially, DNAV is a table that keeps track for each direction the time during which a node must not initiate a transmission in that direction. As in IEEE 8., nodes continuously update this table upon overhearing a packet transmission. In order to ensure the correct update of DNAV, it does not suffice to only update it in the direction through which a packet has been received. When a node receives directional RTS/CTS packet, it should not only defer in the direction from which it received the packet so as to overcome the deafness problem, but also in the direction of the transmission between the sender and receiver. solves these problems by a very simple mechanism by which whenever a node receives a circular directional packet (i.e., RTS or CTS) it can reliably determine the antenna beams it should update its DNAV. More specifically, whenever a node S transmits an RTS or CTS packet to node R, it puts in the packet header the antenna beam node R will use to receive node S s packet. Node S can easily determine node R s receiving antenna, say Q RS, given that it knows its angle of arrival (AoA) [], say Q SR, it uses to communicate with node R by: Q RS (Q SR, Q ) = Q SR + 8 () Assuming the nodes are using the same number of antenna beams, Q RS can be used by Node S to determine node R s receiving antenna. For even number of antenna beams, equation () can be further simplified as a

5 function of antenna beam. For example, Node S can easily determine node R s receiving antenna, say A RS, given that it knows through which antenna, say A SR, it uses to communicate with node R by: A RS M M ASR +, if ASR < ( ASR, M ) = () M ASR, otherwise where M is the number of the antenna beams in a node, as previously defined. Basically, equation () is used to shift node S s antenna and obtain node R s receiving antenna. Now assume a neighbor of node S, say node T, receives the circular directional RTS packet through antenna beam A TS. First of all, node T updates its DNAV with the duration field contained in the RTS packet in the direction of node S, that is, it updates DNAV(A TS ). Next, node T has to determine if it needs to update its DNAV for the same duration in the direction of node R as well. For this to happen, node R has to be a neighbor of node T, and the antenna beam, say A RT, node R uses to communicate with node T is equal to the antenna beam A RS contained in the RTS packet header. To calculate A RT, node T employs the same equation () and uses A TR, the locally available antenna beam node T uses to communicate with node R, as input. We note that this new scheme reduces the overhead of in half while tackling the problem at both sender and receiver neighborhood. 5. PERFORMANCE EVALUATION We have implemented a directional antenna module in ns (version.6). This module models most of the aspects of a directional antenna system including variable number of antenna beams, different gains for different number of antenna beams among others. As for the protocol support, we have implemented DMAC,, and. For the simulations that follow, we have considered CBR traffic sources at data rates of 4 Kbps, 8 Kbps, Kbps, and 6 Kbps, and we measure the total network aggregate throughput of all flows. In addition, we evaluate DMAC, and for six, twelve, and eighteen antenna beams with transmission ranges of 46, 74 and 9 meters, respectively. For IEEE 8., the transmission range is set to 5 meters. Also, in all the scenarios we consider a Mbps network with no node mobility. 5. Linear Topology To quantitatively analyze the impact of this scenario on the network performance, we simulate the network of Figure 3 where we compare the performance of IEEE 8., DMAC,, and. In this scenario, node S transmits to node R, node A transmits to node S, and node B transmits to node R. The coverage range is such that node S s RTS does not include node B, and node R s CTS does not include node A, since nodes A and B listen to the medium omni-directionally. On the other hand, node A s RTS includes node R as node R is listens to node S (DATA transmission) directionally, and may cause collisions. Similarly, node B s RTS includes node S as node S listens to node R (CTS and ACK) directionally, and may also result in collisions. Figure 4 shows the simulation results obtained for this scenario. Similar to [3, 4], our results show that directional antennas have an inferior performance for linear topologies as compared to IEEE 8. given that the larger range is blocked in directional antennas as compared to IEEE 8.. Thus, IEEE 8. achieves a better special reuse in linear topologies. Despite of that, we see that, amongst the directional MAC protocols evaluated, performs best. This is mainly due to the optimized circular directional transmission of both RTS and CTS which informs the neighbors of a node in little time about the intended transmission, thus preventing hidden terminals., on the other hand, does not perform comparable to as it employs circular transmission of RTS only, and does it in all sectors (even the empty ones). Finally, DMAC has the poorest performance as it causes many collisions due to the hidden terminals. 5. Gain by Spatial Reuse From now on, we concentrate on the performance comparison of IEEE 8., and only, as these protocols do not assume prior knowledge of neighbors location as in DMAC. Therefore, for the sake of a fair analysis and to compare the efficiency of and neighbor discovery mechanisms, we have removed DMAC from the simulations that follow where no prior neighbors location information is available.

6 In this section, we evaluate the performance of these protocols under scenarios where all nodes are within radio range of each other. Given that shortest transmission range is 5 meters in case of IEEE 8., the network topologies here evaluated have all nodes confined within a circle of 5 meters diameter. By doing this, we plan to evaluate the spatial reuse gain provided by directional antennas as compared to omni-directional antennas. In the next section, we focus on the gain by increased coverage range..6 A 3 4 S 3 4 R 3 4 B 3 4 Aggragate Throughput (Mbps) DMAC Figure 3 Example of a linear topology scenario Figure 4 Aggregate throughput in a linear topology 5.. Random Topology We now simulate a topology comprised of 6 nodes randomly distributed which are within the radio ranges of each other. Our results are averaged over a total of different scenarios. Figures 5(a), 5(b), and 5(c) show the simulation results when nodes possess six, twelve, and eighteen antenna beams. It is important in these figures that IEEE 8. is practically the same when all stations are within the radio range of each other, as no spatial reuse is possible. In Figure 5(a), we see that outperforms all other schemes, except under low load as nodes in a random topology may eventually have to spend more time in the circular RTS/CTS procedure. In other words, in random topologies fewer are the empty sectors. However, in high load surpasses IEEE 8.. It is interesting to note that performance is inferior to IEEE 8.. The reason is that it spends a lot more time than performing the circular transmissions. Not only this, it so happens that when a transmitter node using is over performing all its circular transmissions, the RTS happened to have collided at its intended receiver. Thus, many circular RTS transmissions end up being useless. When the number of antennas increases from six to twelve and eighteen (Figures 5(b) and 5(c)), we see that performance is further enhanced due to the increased spatial reuse. As for, it surpasses IEEE 8. in medium and high load when twelve antennas are employed. However, in eighteen antenna beams throughput is again below that of IEEE 8.. Once more, the reason is that, as the number of antenna beams increase, throughout is highly deteriorated given that it carries out far too many circular transmissions (a) 6 antenna beams (b) antenna beams (c) 8 antenna beams Figure 5 Spatial reuse gain in random topology (aggregate throughput) 5.3 Gain by Increased Coverage Range Here, we focus on the second advantage of directional antennas, namely, the increased coverage range due to directionality. Therefore, in this section we evaluate the performance under scenarios where not all pairs of source and destination nodes are within radio range of each other. Given that IEEE 8. range is 5 meters, it may

7 have to resort to the routing protocol in order to deliver a packet to a particular destination. On the other hand, it may be the case that and may or may not need to resort to routing as they can transmit for longer ranges. For the scenarios that follow, we have used the DSR routing protocol [5]. Similar to the scenario used in Section 5.., the scenario studied here reflects a total of 6 nodes randomly distributed on a two dimensional plane, where the distance between pairs of source and destination nodes is selected between [45, 89] meters. Again our results are averaged over different scenarios. The results for six, twelve, and eighteen antenna beams are depicted in Figures 6(a), 6(b), and 6(c), respectively. As expected, in all three figures the directional protocols outperform the omnidirectional IEEE 8.. Notably, is shown to provide the best performance of all protocols considered (a) 6 antenna beams (b) antenna beams (c) 8 antenna beams Figure 6 Coverage range gain in random topology (aggregate throughput) 6. THE PROPOSED E PROTOCOL All the aforementioned protocols, including, assume a traditional network layer model as shown in Figure 7(a). Here, the link layer has a single queue of packets waiting to be handed over to the MAC layer which is, in turn, a single buffered entity. Whenever the network layer has a packet to send, it determines the next hop for the packet and places it in the link layer queue. In case MAC is in IDLE state, it signals for a packet from the link queue and subsequently buffers it. It then determines the antenna beam required to transmit the packet and enters into SEND state. The MAC will only request another packet from the link layer queue when it has successfully transmitted or given up (e.g., the next hop is unreachable) on the packet it is currently handling. In existing directional MAC protocols, in the event that the packet to be transmitted is for a beam whose DNAV was set it waits for the medium to become idle. While doing so, it could so happen that other packets in the link layer queue could be transmitted over beams which are not busy at that time. In such a scenario, waiting for the medium to become idle reduces overall throughput of the system. We call this as self induced blocking phenomenon which results from using a single MAC buffer for all antenna beams. We overcome this problem in the Enhanced (E) protocol by employing a cross-layer design approach wherein the network layer is aware of the different antenna beams at the MAC layer. The MAC, in turn, has separate buffers for each the antenna beams. Accordingly, the link layer follows this approach by maintaining separate queues for each beam. The modified protocol stack of E is shown in Figure 7(b). As we can see, in E the MAC layer has multiple buffers for each corresponding antenna beam, where each of them corresponds to a specific queue in link layer. In addition, E employs separate backoff timers for each these antenna beams to allow for simultaneous execution. In order to place the packet in the correct link layer queue, the network layer needs to determine the antenna beam which the MAC will use for transmission of this packet. We tackle this issue by augmenting the routing table with an additional entry called Antenna Beam, which corresponds to the antenna beam the MAC uses to reach the corresponding next hop. This entry in the routing table is self-learning and its computation incurs no additional overhead. Whenever the MAC receives a packet from a node, it informs the network layer the antenna beam through which it received the packet. The network layer, in turn, updates the beam entry for that destination in the routing table. As time progresses, the network layer will eventually learn about the antenna beams used to reach each of its neighbors. In case the network layer s beam entry field is empty for a given next hop, a simple broadcast is done when this entry is needed. We note, however, that this does not result in any overhead as the majority of existing routing protocols rely on some sort of broadcasting (see Section 4.). As a result, once the network layer is aware of the destination s antenna beam, it puts the packet in the link layer queue corresponding to this antenna beam. It is to be noted that broadcast packets are kept in a special dedicated queue as they are to be transmitted through all antenna beams.

8 In E, whenever the MAC enters an IDLE state it explicitly requests a packet from the link layer for which the DNAV is not set. In scheduling the next packet, we follow a simple round robin strategy for determining the next unblocked antenna beam. If no packets are available for that beam, then a packet for the next unlocked beam is sought. Upon receiving a packet, the MAC first stores it in the buffer allocated for this antenna beam and makes an attempt to transmit it. It may so happen that in the time interval between determining an antenna beam to be unblocked and actually attempting to transmit it, the beam may become locked (e.g., medium busy or DNAV set). In addition, the beam may get blocked while waiting for DIFS or backing off. In all these cases, we should first start backoff in this particular beam and then invoke the scheduler to move to next available antenna beam. This is continued until the MAC receives a packet which it successfully transmits, or all the buffers for the each antenna beam are full. In the former case, the MAC uses the round robin scheduler to select the next antenna beam, whereas in the later case the MAC simply waits for one of its backoff timers to expire before attempting transmission. In case of a virtual collision (e.g., two backoff timers expiring simultaneously), their transmissions are attempted one after the other. We note that this is different from IEEE 8. where the MAC freezes the backoff counter and waits for the medium to be idle again before resuming it. Application Layer Transport Layer Network Layer Link Layer Single Link Queue Application Layer Transport Layer Network Layer Link Layer Multiple Link Queues MAC Layer Physical Layer Physical Layer Multi buffered MAC (a) Traditional Protocol Stack (b) E Protocol Stack Figure 7 Traditional and E protocol stack It is to be noted that round robin strategy to handle scheduling in MAC have several pitfalls. It includes increased delay and possible starvation of packets in a particular direction. Comparison of different scheduling algorithm is, however, beyond the scope of this paper. 7. PERFORMANCE EVALUATION As discussed in the previous section, one of the issues in directional antennas is the phenomenon of self induced blocking problem. As a result of this, if a packet for a particular antenna beam is blocked, the node will not be able to transmit in any other directions until the blocked packet is successfully transmitted. To illustrate the effect of single MAC buffer and corresponding gain by employing separate MAC buffers for each antenna beams as in E, we redefine random topology and run the simulations for 6,, and 8 antenna sectors. For the sake of clarity, in this set of simulations we have only considered the directional protocols, E and. In this scenario we simulate a randomly distributed topology of 6 nodes. Out of these 6 nodes, 3 are chosen as traffic sources that, in turn, randomly select two of its neighbors as destinations. We have simulated a total of scenarios and the results presented here are the average of their individual results. The corresponding plots for 6, and 8 antenna sectors are shown in Figure 8. As can be seen from Figures 8(a), 8(b), and 8(c), E consistently outperforms and. The throughput of these protocols is approximately the same for the case when each traffic source generates data at 4 Kbps. This is because, in this case, the network in under loaded and the benefits of directionality are not observed. Also, the effect of the self induced blocking problem or the circular transmission of RTS on throughput is minimal. However, as the network load increases the effect of the above factors become increasingly worse and considerably affects the performance of and, while is affected more because of its circular RTS.

9 (a) 6 antenna beams (b) antenna beams (c) 8 antenna beams Figure 8- E aggregate throughput in random topology 8. CONCLUSIONS AND FUTURE WORK In this paper we have considered the problem of medium access control for ad hoc networks employing directional antennas. We have discussed the shortcomings of existing work and have proposed two schemes, called Directional Antenna Medium Access () protocol and Enhanced (E). The basic scheme implements unique mechanisms including simultaneous transmissions of RTS and CTS packets, an optimized form of sweeping. With E, we introduce the concept of multiple MAC buffers for each directional beam and introduce a cross-layer design to make the network layer aware of the various directions. By doing so, we overcome the self induced blocking problem found in and in existing directional MAC protocols. E provides an average of % performance enhancement over in all scenarios investigated. Finally, it is to be noted that the system performance depends upon the network topology as well as the traffic pattern between nodes. As future work, we plan to investigate the issue of power control over directional antennas and to compare different scheduling strategies for selecting the next free beam. ACKNOWLEDGMENTS This work has been supported by the Ohio Board of Regents Doctoral Enhancement Funds and the National Science Foundation under grant CCR-336. REFERENCES [] IEEE Std. 8-. IEEE Standard for Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specification. DOI: [] C. Fullmer, J.J. Garcia-Luna-Aceves. Floor Acquisition Multiple Access (FAMA) for packet radio networks. Computer Communication Review 995; pp:6-73. DOI:.995. [3] R. Choudhury, X. Yang, R. Ramanathan, N. Vaidya. Using Directional Antennas for Medium Access Control in Ad Hoc Networks. ACM Mobicom ; DOI: 9.. [4] T. Korakis, G. Jakllari, L. Tassiulas. A MAC protocol for full exploitation of Directional Antennas in Ad-hoc Wireless Networks. ACM Mobihoc 3; DOI: 6.3. [5] A. Chandra, V. Gummalla, J. Limb. Wireless Medium Access Control Protocols. IEEE Communications Surveys and Tutorials ; DOI: /vol, no.. [6] M. Horneffer, D. Plassmann. Directional Antennas in Mobile Broadband Systems. IEEE Infocom; DOI: [7] T. Yum, K. Hung. Design Algorithms for Multihop Packet Radio Networks with Multiple Directional Antennas. IEEE Transactions on Communications 99; 4(): DOI:.99. [8] A. Nasipuri, S. Ye, J. You, R. Hiromoto. A MAC Protocol for Mobile Ad Hoc Networks using Directional Antennas. IEEE WCNC; DOI: 9.. [9] R. Choudhury, X. Yang, R. Ramanathan, N. Vaidya. Using Directional Antennas in Ad Hoc Networks. Final report submitted by Texas A&M University to BBN technologies, July. [] Y. Wang, J.J. Garcia-Luna-Aceves. Spatial Reuse and Collision Avoidance in Ad Hoc Networks with Directional Antennas. IEEE Globecom; DOI:.. [] Y.-B. Ko, V. Shankarkumar, N. Vaidya. Medium access control protocols using directional antennas in ad hoc networks. IEEE Infocom; (3): 3-. Infocom: Tel Aviv, Israel,.

10 [] M. Takai, J. Martin, A. Ren, R. Bagrodia. Directional Virtual Carrier Sensing for Directional Antennas in Mobile Ad Hoc Networks. ACM MobiHoc; DOI: 6.. [3] R Ramnathan. On the Performance of Ad Hoc Networks with Beamforming Antennas. ACM MobiHoc; DOI:. [4] M. Gerla, G. Pei, X. Hong, T.-W. Chen. Fisheye State Routing Protocol (FSR) for Ad Hoc Network. Draft-ietfmanet-fsr-.txt, Nov.. [5] B. Bellur, R. Ogier, F. Templin. Topology Broadcast Based on Reverse-Path Forwarding (TBRPF). Draft-ietfmanet-tbrpf-.txt, March. [6] D. Johnson, D. Maltz, Y.-C. Hu, J. Jetcheva. The dynamic source routing protocol for mobile ad hoc networks (DSR). IETF Internet-Draft, Nov.. [7] ns- Network Simulator, DOA:.4 [8] C. L. Fullmer, J. J. Garcia-Luna-Aceves. Solutions to Hidden Terminal Problems in Wireless Networks. ACM SIGCOMM, 997. [9] W. Moh, D. Yao, K. Makki. Wireless LAN: Study of hidden terminal effect and multimedia support. Computer Communications and Networks, pp.4-43, 998. [] E. -S. Jung, N. Vaidya. A Power Control MAC Protocol for Ad Hoc Networks. ACM Mobicom,. [] C. Perkins, E. Royer, S. Das. Ad Hoc On Demand Distance Vector Routing (AODV). Internet Draft, March [] P. Jacquet, P. Muhlethaler, A. Qayyum, A. Laouiti, L. Viennot, T. Clausen, Optimized Link State Routing Protocol, Draft-ietf-manet-olsr-4.txt. BIOGRAPHY HRISHIKESH GOSSAIN (hgossain@ececs.uc.edu) is a PhD candidate at the Department of ECECS, University of Cincinnati. He received B.E. in Electronics Engineering from Motilal Nehru Regional Engineering College, Allahabad India in 998, where he was an undergraduate Gold-Medalist of the College. He has four pending patents involving wireless MAC protocols, access network design, QoS and e-media. His research interests are in the areas of wireless and mobile computing including MAC layer protocol design and analysis, IEEE 8., power control, directional antenna systems, TCP over wireless, Mobile IP and ad hoc networks. He also has previous work experience in Nortel Networks in Richardson, Texas. CARLOS DE MORAIS CORDEIRO is a Senior Member Research Staff of Philips Research USA, Briarcliff Manor, NY. Before joining Philips Research USA, he was a Senior Research Engineer at Nokia Research Center. He received his PhD in computer science and engineering in 4 from the University of Cincinnati, OH, USA, where he won the honorable Outstanding Doctoral Dissertation Award and the prestigious 3/4 The National Dean s List Award. Earlier, he obtained a M.S. and B.Sc. in computer science in 998 and, respectively, from the Federal University of Pernambuco, Brazil. His research interests are in the broad area of wireless and mobile communication including MAC protocol analysis and design, MIMO systems, IEEE 8., IEEE 8.5, IEEE 8.6, cognitive radios, power control, spectrum management, ad hoc and sensor networks, routing, multicast, TCP over wireless, and cellular networks. Dr. Cordeiro has published numerous papers in the wireless area and holds many pending patents involving MAC protocols for WLANs and WPANs. He has delivered tutorials in areas such as directional antenna systems, wireless broadband and mobile ad hoc and sensor networks, and in the past was the recipient of best paper awards from refereed networking conferences. Dr. Cordeiro has also worked for IBM in San Jose, CA, and is a member of the IEEE.. TARUN JOSHI is a PhD student at the Department of ECECS, University of Cincinnati. He received his Bachelor of Technology in Computer Science from Indian Institute of Technology - Roorkee in. Since then he has been working as a research assistant at the Center for Distributed and Mobile Computing. His research interests include MAC protocols over directional antennas, broadcasting and routing issues in directional antennas, TCP behavior over directional antennas, scheduling in wireless networks, mobility management, and security in infrastructure wireless networks. Tarun Joshi has also worked as a Software Engineer Intern for Talisma Corporation, India. DHARMA P. AGRAWAL is the Ohio Board of Regents Distinguished Professor of Computer Science and Computer Engineering and the founding director for the OBR Research Center for Distributed and Mobile Computing in the Department of Electrical & Computer Engineering and Computer Science, University of Cincinnati, OH. His current research interests include wireless and mobile networks, distributed processing, and scheduling techniques. Dr. Agrawal is an editor for the Journal of Parallel and Distributed Systems and the International Journal of High Speed Computing. He has served as an editor of the IEEE Computer magazine, and the IEEE Transactions on Computers. He has been the Program Chair and General Chair for numerous international conferences and meetings. He was selected for the "Third Millennium Medal" by the IEEE for his outstanding contributions. Four of his patents in wireless networking area have also been approved recently. Dr. Agrawal is a fellow of the IEEE, ACM and AAAS.