New CDMA-Based MAC Protocol for Ad Hoc Networks

Transcription

1 ew CDMA-Based MAC Protocol for Ad Hoc etworks Thamer Al-Meshhadany and Wessam Ajib Computer ciences Department, Université du Québec à Montréal 2, avenue du Président-Kennedy Montréal, Québec, Canada H2X 3Y7 Phone # 3227; fax thamer@labunix.uqam.ca, ajib.wessam@uqam.ca Abstract It is widely accepted that Ad Hoc networks are at the leading edge of the resear in the domain of wireless networking. These networks are not supported by infrastructure to connect the mobile hosts, thereby they have to be self configured, self organized and the resources have to be allocated in a distributed manner. The medium access control (MAC) layer is seen as the bottleneck for the throughput in wireless Ad hoc networks. Hence, we propose in this work a new Multiannel MAC protocol. The proposed protocol can be based on Code Division Multiple Access (CDMA) or Frequency Division Multiple Access (FDMA). A annel can be represented by one spreading code in CDMA systems or by one frequency band in FDMA case. In our analysis and simulations, we assume that the protocol is based on CDMA tenique. We consider one annel for control packets and multiple annels for transmitting data information. We propose that the reservation of a data cannel is done implicitly using the common annel. We show through computer simulations that our proposition of Multiannel MAC protocol improves significantly the communication performance in wireless Ad Hoc networks, even when the introduced overhead is considered. I. ITRODUCTIO Ad Hoc networks consist of terminals that are distributed in a geographical zone and connected together without infrastructure. Hence, They have to be self configured, self organized and the resources have to be allocated in a distributed manner. The medium access control (MAC) protocol in these networks encounters two major phenomena that limit the network capacity []. These phenomena are the hidden and exposed terminals. The first one causes higher rate of packet collisions; whereas, the second phenomena causes a loss in the resources. Many MAC protocols are proposed to overcome these problems. However, the proposed solutions are in general either of high cost or/and complex and they haven t completely eliminated these problems. In this paper, we propose a new multiannel MAC protocol for wireless Ad Hoc networks. The proposed protocol is based on CDMA tenique and uses spreading codes to identify different annels. However, the method considered in this protocol for annel identification allows us to utilize other multiple access tenique su as Frequency Division Multiple Access (FDMA) tenique where ea annel is represented by a frequency band in one radio transceiver systems. It can also be applied to a multi-radio systems where ea radio transceiver may correspond to one annel. In the proposed protocol, we consider one common control annel on whi RT (Request to send) and CT (clear to send) packets are exanged, and several data annels used for data packets transmission. We assume that time is divided into frames that are also divided into slots. By inserting a synronization interval at the beginning of ea frame in the control annel, the synronization among the terminals is established. The reservation of a particular data annel for one transmission is realized by the control annel. Obviously, all the terminals have to monitor the common control annel in order to identify and reserve the data annel. The rest of this paper is organized as follows. In ection II, we discuss the related works with a particular focus on the CDMA-based protocols. ection III describes our proposed multiannel MAC protocol for wireless Ad Hoc networks. imulation environments and results analysis are presented in ection IV showing the improvement of the performance realized. In terms of simplicity, efficiency and flexibility, a discussion is established in section V. Finally, we conclude this paper in ection VI. II. RELATED WORK MAC protocols based on CDMA for Ad Hoc networks return to [], in whi the authors have proposed two different protocols. According to these protocols, the codes are pre assigned to the terminals. The transmitted packets consist of data and a header used mainly to identify the destination. Hence, the packet s data is transmitted on a transmitter-based code, but the packet s header is either transmitted on a common code (the first protocol) or transmitted on the receiver code (the second protocol). Hence, the number of used spreading codes relies completely on the network size. In the receiving mode, a terminal monitors either a common code, in the first case, or a receiver-based code in the second case. ubsequently, even though the rate of occurred collisions is reduced, the collisions still happen for the packets transmitted on the common code in the first protocol or for the ones transmitted on the receiver code in the second protocol. A pairwise code assignment has been proposed in [2]. It centrally assigns codes to ea transmitter-receiver pair (link) su that no two adjacent edges have the same code. Accordingly, ea terminal must maintain many codes to connect with the neighbored terminals. The overhead, caused by the /07/$ IEEE 9

2 exange and the broadcast of code assignment signalling messages, is relatively high. The number of used codes have been minimized in [3] where the authors have shown that the minimization problem is P-complete. Another way of thought is to make use of handshaking control packets to sense the annel. The handshaking packets were first introduced by MACA (Multiple Access with Collision Avoidance) protocol in [4]. The sender and receiver terminals exange two control packets, namely Ready To end (RT) and Clear To end (CT) before sending the data packet. Two protocols based on common-transmitter based (MACA/C-T) and receiver-transmitter based (MACA/R-T) are proposed in [5]. These protocols can be considered as variants to the ones presented in [] except that the new proposed protocols take into consideration the RT/CT packets. Thus, collisions occur only at the level of the control packets (RT and CT) exange. On the other hand, sending the data with the receiver s code was investigated in [6]. However, it is well-known that the big disadvantage of this kind of protocols is that it can t support multicast services. The authors of [7] have proposed a distributed algorithm of code assignment in whi the number of available codes is at least d (d )+2 where d is the maximum number of one-hop neighbours any terminal can have. An important aracteristic of this protocol is the existence of a common annel that is used to exange the code assignment information between the terminals. In [8], the authors have proposed a dual reservation MAC protocol in whi +2 codes are used ( codes for data, one common code for control messages and one code for broadcasting). Ea terminal has three lists that contain the following information: available codes list, occupied codes list and forbidden codes list. The code used for transmitting data packets between two terminals is dynamically assigned after negotiations using the control messages. A. Muqattash and M. Krunz have proposed a multiannel MAC protocol in [9]. In this proposition, the bandwidth is divided into two frequency annels. One annel is used to exange control messages and the other annel is used for the data transmission. All terminals use the common code on the control annel; whereas, on the data annel, only one code is assigned to ea terminal for data transmission. Although the proposed solution reduces the number of occurred collisions, the hardware implementation is complex and of high cost. In [0], two phases coding multiannel MAC protocol was proposed. They assume that one terminal have to take the responsibility to assign the codes for the terminals in its cell. In fact, this protocol will be more suitable to wireless LA 802. DCF. In [], Asynronous Multiannel Coordination Protocol (ACMP) is devised. It is based on 802. Distributed Coordination protocol (DCF). It uses a dedicated control annel and multiple data annels. The terminals contend on the control annel by exanging RT/CT packets in order to reserve a data annel. III. PROPOED MULTI-CHAEL CDMA-BAED MAC PROTOCOL In this work, we propose a new MAC protocol based on CDMA for Ad Hoc networks. We assume + codes: one common code for RT and CT packets transmission and codes for Data and ACK packets transmission. ubsequently, there are + virtual annels. Time is divided into frames and ea frame is divided into slots. The number of slots per frame in data annel is equal to that one in control annel. In order to ensure the synronization among the terminals, ea control frame starts by a synronization interval (Y). When a terminal wakes up or enters into a geographical zone of one Ad Hoc network, it has to wait until the next synronization interval in order to ensure the synronization with the other terminals. Ea slot in the control annel is divided into two fields. The first is used by the sender to transmit RT packet whereas the second is used by the receiver to respond with CT packet. We assume that the number of slots in one control frame is equal to the number of data annels (codes). Thereby, the slot s number during whi RT/CT packets are exanged between two terminals represents the data annel on whi the data will be transmitted. The procedure of medium access contention is described in the following. All terminals permanently tune to the common code and monitor the control annel. When an active terminal has data to transmit, it waits until the next slot in the control annel and then it transmits a RT packet during the RT field but if the terminal was out of the network coverage or it was inactive, it has to wait firstly until the next synronization interval and then it tries to transmit a RT packet on the first slot in the frame. If this terminal is the only transmitting terminal in this field, the destination terminal will receive correctly the RT packet and will respond with a CT packet. Otherwise, a collision will occur and the collided terminals will backoff for other slots. As all terminals tune to the common code, all terminals overhear the RT/CT packets transmission and they can make benefits from this exanged information. In this work, this is not used because we assume that ea connection is made only for one frame. However, if we extend this work to ensure the reservation for multiple frames, the data s size could be indicated in the RT and/or in the CT packets in order to estimate the number of required data frame. Upon receiving the CT packet, the sender terminal tunes to the corresponding data annel. When the RT/CT packets are successfully exanged and a data frame is reserved, the sender terminal starts transmitting the data at the beginning of the next data annel frame. The destination terminal receives data and replies with an acknowledgment packet on the same data annel. Thus, ea successful RT/CT packets exange reserves only one frame in the data annel. If the sender has more than one data frame, it has to aieve 92

3 Idle annel Frame successful RT/CT exange unsuccessful RT/CT exange Fig.. Data and control frame structure Idle annel Data transmission o RT/CT exange00 Y ACK successful RT/CT packets exange for ea data frame. The procedure is shown in figure. The figure illustrates an example of different successful and unsuccessful RT/CT exanges. We notice that when a RT packet is not transmitted or a RT packets collision occurs (denoted by unsuccessful RT/CT packets exange) during one slot of the common annel, the next frame in the corresponding data annel of that slot will be idle (slots and in frame M). When RT/CT packets exange is successful as in slots 2, 3 and, one data frame is reserved on the corresponding data annel for ea transmission. In this way, the data annel reservation is simply aieved. Contrary to many distributed multiannel MAC protocols for wireless Ad Hoc networks proposed in the literature, our proposed protocol is simple since it uses the slot number as an implicit indicator for the data annel. In other words, when a terminal transmits RT packet in a slot, all terminals including the destination terminal know whi data annel will be used for the transmission of data by the sender. Hence, the terminals do not need to transmit any signalling packets to reserve a data annel. Furthermore, the proposed protocol is able to be extended in order to include more complicated functions su as multiple frame reservation. The information about the annels (corresponding to different codes in CDMA systems) reserved for a particular terminal for data transmission is provided implicitly to the corresponding terminal. A. imulation environments IV. IMULATIO REULT In order to evaluate the performance of our proposed multiannel MAC protocol, our computer simulations are aieved using etwork imulator (-2). As the latter only considers the standard protocols for the medium access control like IEEE 802., Carrier ensing Multiple Access (CMA), Time Division Multiple Access (TDMA) protocols and others that all are based on one annel, we had to modify the protocol of IEEE 802. and TDMA protocols in order to aieve our model implementation. In our simulations, we assume that the maximum number of data annels is eight, hence there is up to eight flows can be simultaneously established during one frame. In other words, sixteen terminals can be connected one-to-one. We also assume there is no interference form one pair connection on another connections. When the number of flows is less than or equal to the number of slots, the maximum throughput can be aieved if we consider no RT packets transmission collision. In the simulations, we consider eight slots per frame, corresponding to eight data annels. We assume that the RT, CT and ACK packets have the same structure and the same size as in 802. standards. The maximum size of data packet is up to 800 bytes. We consider 30 terminals in geographical zone of m 2, in other word, all terminals are in same sensing zone. The traffic generated is Constant Bit Rate CBR. The implementation parameters are depicted in the table I. 93

4 TABLE I IMULATIO MODEL PARAMETER Parameter value Basic rate 2 Mbps (RT/CT) Data rate 2 Mbps (DATA/ACK) Y 44 usec (36 bytes) Packet size 000 bytes RT/CT header 20/4 bytes DATA/ACK header 28/4 bytes Channel switing delay 0 sec cenario 30 terminals slight decrease in the throughput of 802. protocol due to the increases in the rate of packet collisions. Contrary, the throughput with our MAC protocol increases when the packet arrival rate increases. However, our proposition performs the aggregate throughput for Ad Hoc networks in comparison with other protocols. In fact, our proposed protocol does not need any signalling packets to reserve a data annel, contrary to other protocols that need an exange of signaling packets packets whi introduces a waste of time and resources in order to reserve one data annel. In other word, our proposed protocol avoids the Ping-Pong phenomena between the communicated terminals in order to reserve the data annel. B. Results analysis In this section, we provide an evaluation of the performance of our proposed multiannel CDMA-based MAC protocol and we compare it by the performance of IEEE 802. based Ad Hoc network with one annel. Figure 2 depicts the aggregate throughput as a function of the number of annels. We notice from figure 2 that our proposed MAC protocol gives higher throughput in function of the number of used annels. In fact, the throughput (defined by the rate of packets correctly transmitted) of 802.-based Ad Hoc network is almost fixed since only one annel is used. On the other hand, the throughput increases linearly with the number of annels using our MAC protocol. When all annels are used and the number of active terminals (transmitting terminals) is more than the number of annel, the throughput becomes invariant. This behaviour is expected because the saturated case is reaed. Throughput(pkt/s) Our proposition umber of annels Fig. 2. Throughput as a function of number of annels Another profile of results is depicted in figure 3 where the throughput versus the packet arrival rate is provided. When the traffic is low, the behaviour of our proposition likes the behaviours of other propositions including IEEE 802. protocol. But, when the packet arrival rate increases, the performance of 802. gets saturated, even, we notice a Throughput (kbps) Fig DCF Our proposition Packet arrival rate per flow Throughput as a function of packet arrival rate per flow V. DICUIO AD FUTUR WORK This section presents our future works and provides some discussion about the flexibility, simplicity and efficiency that caracterizes our proposed multi-annel MAC protocol based on CDMA tenique. We can say that our proposed seme is efficient, simple and flexible. It is efficient because connections can be aieved at the same time where is the number of used spreading codes that in turn corresponds to the number of slots in one frame in the control annel. The seme is simple because it uses a simple algorithm of code allocation. By the the number of the slot used for control packets, the sender terminal implicitly tells the destination terminal about the data annel on whi the data will be transmitted. Moreover, the proposed seme is flexible because its structure is open for future developments. The protocol can be applied to systems that use FDMA as access tenique. Even, the protocol can be applied to the actual 802. system if we consider that one annel is used as control annel and the rest of annels as data annels. The data annel is reserved during the control annel monitored by all terminals. In this work, a terminal can establish only one connection because it has only one simple transceiver but if this terminal 94

5 has multi-transmitters and multi-users detection receiver, it can reserve up to connections during one frame under certain conditions, i.e. this terminal can use all the data codes at the same time. Our proposed MAC protocol can be improved in order that ea terminal will be able to reserve the necessary frames by using only one RT/CT packets exange. The sender terminal has to indicate the amount of transmitted data in order to estimate the number of required data frames. A data annel can be simply reserved by adding busy tone at the beginning of the corresponding slot in the control frame. The busy tone will prevent the other terminals to contend on this annel. The impact of Multiple Access Interference MAI on the resources allocation is not studied in this work. Hence, in order to complete this work, the impact of MAI and the power control must be studied. [] J. hi, T. alonidi and E. W. Knightly, tarvation Mitigation Through Multi-Channel Coordination in CMA Multi-hop Wireless etworks, MobiHoc 06, May VI. COCLUIO In this paper, we presented the Multiannel MAC Protocol and its use in wireless ad hoc networks to improve network performance. The most significant advantage of this protocol is that several connections can occur concurrently using multiple annels. The annel allocation and identification are simply aieved by transmitting RT/CT packet in one slot in the control annel. The slot s number represents the number of a corresponding data annel. By this, the hidden and exposed terminals are eliminated. We evaluated the performance using the ns-2 simulator in term of throughput when the multiannel MAC protocol in wireless ad hoc networks is used. The evaluation results show that higher network throughput can be obtained. REFERECE [] E.. OUA and J. A. ILVETER, preading Code Protocols for Distributed pread-pectrum Packet Radio etworks, IEEE Transactions on communications, Vol. 36, o. 3, Mar 988. [2] L. Hu, Distributed Code Assignements for CDMA Packet Radio etworks, IEEE/ACM Transactions on etworking, Vol., o. 6, December 993. [3] A. A. Bertossi and M. A. Bonuccelli, Code Assignment for Hidden Terminal Interference Avoidance in Multihop Packet Radio etworks, IEEE/ACM Transactions on etworking, Vol. 3, o. 4, August 995. [4] P. Karn, MACA-A ew Channel Access Method for Packet Radio, ARRL/CRRL Amateur Radio Computer etworking Conf., 990.* [5] M. Joa-g and I-Tai Lu, pread pectrum Medium Access Protocol with Collision Avoidance in Mobile Ad-Hoc Wireless etwork, IFOCOM 99. Eighteenth Annual Joint Conference of the IEEE Computer and Communications ocieties, Volume 2, pp: , 2-25 Mar 999. [6] R. Fantacci, A. Ferri and D. Tari, A MAC tenique for CDMA based Ad-Hoc network, IEEE Wireless Communications and etworking Conference, pp: , Mar [7] J. J. Garcia-Luna-Aceves and J. Raju, Distributed assignment of codes for multihop packet-radio networks, in Proc. MILCOM, pp: ,, vol., 997. [8] M. Jia, H. Chen and Y. Yuan, A Dual Reservation MAC Protocol based on CDMA for Ad hoc etworks, Asia-Pacific Conference on Communications, pp: 53-56, October [9] A. Muqattash and M. Krunz, CDMA-Based MAC Protocol for Wireless Ad Hoc etworks, MobiHoc 03:The Fourth ACM International ymposium on Mobile Ad Hoc etworking and Computing, pp: 53-64, June [0] L. Zhang, B-H oongtwo-phase Coding Multiannel MAC Protocol With MAI Mitigation for Mobile Ad Hoc etworks, IEEE Communications letters, Vol. 8, o. 9, eptember