IEEE SENSORS JOURNAL, VOL. 13, NO.

6, JUNE 2013

2391

An Intelligent Hybrid MAC With Trafc-Differentiation-Based QoS for Wireless Sensor Networks
Mohammad Arifuzzaman, Student Member, IEEE , Mitsuji Matsumoto, Senior Member, IEEE , and Takuro Sato, Fellow, IEEE

Abstract In this paper, we present the Intelligent Hybrid MAC (IH-MAC), a novel low power with quality of service guaranteed medium access control protocol for wireless sensor networks (WSNs). The IH-MAC achieves high energy efciency under wide range of trafc load. It ensures shorter latency to critical and delay-sensitive packets. The IH-MAC protocol achieves high channel utilization during high trafc load without compromising energy efciency. The IH-MAC does it by using the strength of CSMA and TDMA approach with intelligence. The novel idea behind the IH-MAC is that it uses both the broadcast scheduling and link scheduling. Depending on the network loads, the IH-MAC protocol dynamically switches from broadcast scheduling to link scheduling and vice versa in order to achieve better efciency. The scheduling is done in the IH-MAC with a novel decentralized approach where the nodes locally use the clock arithmetic to nd the time slot, allocated for it. Furthermore, the IH-MAC uses Request-To-Send, Clear-To-send handshakes with methods for adapting the transmit power to the minimum level necessary to reach the intended neighbor. Thus, the IH-MAC reduces energy consumption by suitably varying the transmit power. The IH-MAC also uses the concept of parallel transmission that further reduces delay. The analytical and simulation results corroborate the theoretical idea, and show the efciency of our proposed protocol. Index Terms Broadcast scheduling, Carrier Sense Multiple Access (CSMA), link scheduling, medium access control, parallel transmission, Quality of Service (QoS), wireless sensor network (WSN).

Listen

Sleep
Time Slot / Frame

Listen

Sleep

Fig. 1.

Periodic listen and sleep of a sensor node.

I. I NTRODUCTION

IRELESS sensor networks (WSNs) have become very popular in recent years. WSN consists of a large

Manuscript received December 28, 2012; accepted March 4, 2013. Date of publication March 11, 2013; date of current version April 30, 2013. This work was supported by a Grant-in-Aid from the High-Tech Research Center Project of the Ministry of Education, Culture, Sports, Science and Technology, Japan. The associate editor coordinating the review of this paper and approving it for publication was Prof. Kiseon Kim. M. Arifuzzaman is with the Graduate School of Global Information and Telecommunication Studies, Waseda University, Tokyo 169-0051, Japan (e-mail: arif@fuji.waseda.jp). M. Matsumoto was with NTT Electrical Communication Laboratories, Tokyo 180-8585, Japan. He is now with the Graduate School of Global Information and Telecommunication Studies, Waseda University, Tokyo 1690051, Japan (e-mail: mmatsumoto@waseda.jp). T. Sato was with the Research and Development Laboratories of OKI Electric Industry Corporation, Ltd., Tokyo 169-0051, Japan. He is now with the Graduate School of Global Information and Telecommunication Studies, Waseda University, Tokyo 169-0051, Japan (e-mail: t-sato@waseda.jp). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/JSEN.2013.2252163

number of wireless sensor nodes that are deployed randomly. The sensor nodes are typically small, and equipped with lowpowered battery. Unlike other wireless networks, it is generally impractical to charge or replace the exhausted battery. Since prolonging lifetime of the sensor nodes is very important, energy efciency becomes the most important attribute of design of MAC protocol of sensor networks. Other attributes are fairness, latency, delivery ratio, and bandwidth [1]. Idle listening is the major source of energy wastage for wireless sensor networks [2]. Therefore, in sensor network, nodes do not wake-up all the time rather prefer energy preservation by going to sleep time to time as explained in Fig. 1. After the sleep scheduling, nodes could operate in a low duty cycle which can signicantly save energy and extend the network lifetime at the expense of increased communication latency and synchronization overhead. In [3] different sleep scheduling schemes are analyzed and a scheduling methods that can decrease the end to end delay is proposed. But this method does not provide an interference free scheduling. One obvious approach canbe TDMA (Time Division Multiple Access) MAC, which inherently supports low duty cycle operations. Besides, TDMA has natural advantage of contention and collisions free transmission [1]. To be interference free, a straightforward approach is to assign each communication link a slot, and thus the number of slot is equal to the number of communication links of the network. However, this scheme requires much more slots than necessary, which enhance delay and reduces the channel utilization. Moreover, minimizing the number of slot assignment for producing an interference free link scheduling is a NP complete problem [4]. On the other hand performance of broadcast scheduling is worse than link scheduling in WSNs, in terms of energy conservation. Henceforth, we propose a new hybrid MAC protocol for wireless sensor network, called IH-MAC, which combines the strength of the CSMA, link scheduling andbroadcast scheduling. There are many applications of wireless sensor network where it is really needed to ensure the priority services for the critical data. One typical example can be a building equipped with a WSN where sensor nodes monitor the power

1530-437X/$31.00 2013 IEEE

2392

IEEE SENSORS JOURNAL, VOL. 13, NO. 6, JUNE 2013

consumption of the appliance in the building. Another duty of the nodes is to work as a smoke detectors, and report alarms to re monitoring hubs. So, in case of the communicating the later kind of data, it is desirable to ensure lowest possible latency. Our proposed IH-MAC protocol guarantees shorter latency for this type of critical and delay-sensitive packets. The rest of the paper is organized as follows. Section II reviews the related works. In Section III, we will elaborate the design of the IH-MAC protocol. In Section IV, we will describe the energy consumption analytical model, followed by results and discussion in Section V. And nally, Section VI concludes the paper with a concrete summary of our contributions. II. R ELATED W ORKS For sensor network, S-MAC [2] is one of the pioneering works in contention based MAC protocol. In S-MAC nodes operates in low duty cycle and energy efciency is achieved by periodic sleeping. Nodes form virtual clusters, based on common sleep schedules, to reduce control overhead and enable trafc adaptive wake-up. T-MAC [5] improves the energy efciency of S-MAC by introducing adaptive duty cycle. T-MAC reduces the idle listening by transmitting all messages in burst of variable length and sleeping between bursts and it maintains an optimal active time under variable load by dynamically determining its length. TDMA has long been dismissed as an unfeasible solution for wireless ad hoc networks for its lack of scalability and adaptability to varying environments. However, it provides a good energy efcient and collision free communication. Recently several techniques [6], [7] have been proposed for TDMA in sensor networks. Nevertheless, these protocols still are unsuccessful to address the fundamental difculties that stand-alone TDMA scheme face. In our previous work [8], we rst introduce the novel idea of combining broadcast scheduling (broadcast TDMA) and link scheduling (pair wise TDMA) together in order to gain energy efciency as well as efcient use of bandwidth. But the possibility of gaining the advantage by considering the scope of parallel transmission was not considered on that work. Never the less, the implementation of transmission power adjustment features was left as a future work. In [9], we combined parallel transmission concept with the TDMA link scheduling to achieve high energy efciency and high packet delivery ratio. But since the protocol is basically based on TDMA it suffers from the high latency and so it is not suitable for the sensor network that has to handle critical trafc by ensuring a minimum latency. In AMAC [10], each node can adjust duration of the active period depending on trafc. In [11], the performance analysis of optimized medium access control for wireless sensor network is done. B-MAC [12] is the default MAC for Mica2. B-MAC allows an application to implement its own MAC through a well-dened interface. They also adopt LPL (Low power listening) and engineer the clear channel sensing (CCA) technique to improve channel utilization. Z-MAC [13] dynamically adjusts the behavior of MAC

between CSMA and TDMA depending on the level of contention in the network. The protocol uses the knowledge of topology and loosely synchronized clocks as hints to improve MAC performance under high contention. Z-MAC uses DRAND [14], a distributed implementation of RAND [15] to assign slot to every node in the network. TIH-MAC [16] is a trafc pattern aware hybrid MAC protocol inspired from Z-MAC. It uses A-DRAND as slot assignment algorithm. A-DRAND is an improved version of DRAND for clustered wireless sensor networks where cluster heads require more slots to relay packets. BAZ-MAC [17] is another hybrid MAC protocol inspired from Z-MAC and proposed for Ad Hoc networks. Like TIH-MAC, BAZ-MAC uses a bandwidth aware slot allocation algorithm during the set-up phase, to assign slots to the nodes according to their bandwidth requirements. Our proposed IH-MAC also combines TDMA and CSMA. But IH-MAC is completely different from Z-MAC and similar hybrid MAC protocol. Because in IH-MAC, each node calculates its own slot locally and independently, this is very exible. Moreover, the hybrid concept used in case of IH-MAC does not stand for the same meaning, as the meaning used by other already proposed hybrid MAC. IH-MAC is hybrid in the sense that it combines CSMA, the broadcast scheduling and link scheduling dynamically to improve the energy efciency. Another important feature of IH-MAC is that it reduces energy consumption by suitably varying the transmit power and it reduces the latency by exploiting the concept of parallel transmission. Designing MAC protocol for wireless sensor network with the provision of QoS has also been considered as an active area of research. Recently several Medium Access Control protocols [18][20] have been proposed to address different service requirements for different trafc. In [18], authors proposed an adaptive scheme for each sensor to determine independently whether to transmit or not, so that a xed total number of transmissions occur in each slot. The protocol accomplishes its task by allowing the base station to communicate QoS information to each sensor node within the network through a broadcasting channel. In [19], authors propose Q-MAC scheme that provides quality of service by differentiating network services based on priority levels. The priority levels reect the criticality of data packets originating from different sensor nodes. The Q-MAC accomplishes its task through two steps; intra-node and inter-node scheduling. III. I NTELLIGENT H YBRID MAC P ROTOCOL D ESIGN We rst dene the terminology used in this paper. A Timeslot or Slot or a Frame is dened as the periodic interval, which consists of an active period and a sleep period. A duty cycle is the proportion or ratio of active period to the entire cycle time (frame length). A rendezvous slot is dened as a time slot explicitly dedicated to a pair of nodes to communicate with each other. During rendezvous, a node forms a channel for transmission and reception with one of its neighbor. The term channel here refers to a time period or slot as opposed to frequency or code. IH-MAC classies packets according to their importance (i.e. delay requirements) and stored the packets into the

ARIFUZZAMAN et al.: IH-MAC WITH TRAFFIC-DIFFERENTIATION-BASED QoS FOR WSNs

2393

Header

Data

Extra 1 bit
(used for priority)

Set Timer for next wakeup

End of Communication

Tx/Rx

Fig. 2.
Active Period

Format of data packet.

Sleep Period
Sleep Time Out Loose Contention Win Contention

SYNC

RTS/ CTS

Data

AcK
Time Value becomes 0 CSMA/CA

Random Access

Scheduled Access

Fig. 3.

Contents of a slot or frame for the proposed IH-MAC protocol.

Wake-up Synchronozation Idle Listening

appropriate queue. The source node knows the degree of importance of the sensed data and accordingly the application layer sets the priority. Application layer does it by appending 1 extra bit at the end of the data packet. Fig. 2 shows the format of each data packet. The mechanism of IH-MAC is based on dividing the communication time into xed length slots or frames. The contents of each slot are shown in Fig. 3. Each slot begins with a SYNC period. The purpose of the SYNC packet is to maintain synchronization between the nodes within the same virtual cluster. The next part of the active period of the frame is reservation slot which is used for the data slot reservation. And the last part is used for data and ACK transmission by sensor nodes. A. Neighbor Discovery, Clustering and Synchronization Stage Frame synchronization is done by virtual clustering, as described in the S-MAC protocol [2]. When a node comes to life, it starts by waiting and listening. If it hears nothing for a certain period, it chooses a frame schedule and transmits a SYNC packet. The SYNC packet contains the time until the next frame starts. If the node during start up hears a SYNC packet from another node, it follows the schedule in that SYNC packet and transmits its own SYNC accordingly. Nodes retransmit their SYNC once in a while. When a node has a schedule but it hears SYNC with a different schedule from another node, it adopts both schedules. Adopting both schedules ensures the successful communication between the nodes of different schedule. The described synchronization scheme, which is called virtual clustering [2], urges nodes to form clusters with the same schedule. So, all the nodes in the networks need not to follow the same schedule. During this virtual cluster creation, each node creates the one hop neighbor list and with using these a node can easily constitutes the two hop neighbor list. After that each node is given an id such that within a two hop neighbor the id is unique. B. Slot Assignment Each slot in IH-MAC consists of a xed length SYNC period, a xed length data period and a sleep period that depends on the duty cycle as in Fig. 3. The duty cycle should be chosen in such a way that the sleep period of a slot is large enough to transmit a data packet along with ACK. All nodes are allowed to transmit in any slot but the

Fig. 4. State diagram (Moore machine description) of sensor nodes working in IH-MAC.

node that has a critical data (with high priority) will get the priority. If no such data is available then the owners of a slot will get the priority. Priority can be ensured by choosing contention window size which is elaborately described in the Section III-G of this paper. The owner calculation will be performed by each sensor node locally by simple clock arithmetic (modulo n ). Where n represents the number of neighbors nodes, within two hop neighbor of a node. The value of n will not be the same for all nodes in a virtual cluster. It should be noted here that using modulo might result in multiple owners of a slot, since different node IDs can map to the same slot number. In case of such situation owner will compete among each other for getting access of the medium. Now, each node can make some of its owned slot as a rendezvous slot with which it can send message to its neighbor exclusively. The rendezvous slots willalso be calculated by each node locally using clock arithmetic, as modulo m . The value of m is set according to the system requirements, i.e. network load, delay, message buffer size etc. And the value of m will be always multiple of n . For the sake of scalability, the value that we use in modulo operation i.e., m will be always larger than the value of n of a particular node. So when a new node wants to join in the network, at least there will be some slots which are not using as rendezvous and it will be used for the scalability. C. State Machine The state machine of the IH-MAC protocol is shown in the Fig. 4. During the sleep state the node turn off its radio and start a timer whose duration is predened according to the duty cycle of the protocol with consideration of the existence of rendezvous communication between any pair of nodes. When the timer expires, the node goes to wake-up state. It turns on its radio and switches to listen to the data channel and its goes to idle listening state. If the node have any data to send or receive it goes in the CSMA/CA state otherwise after time out it goes to sleep state. If the sender node wins the contention both the sender and the intended receiver go to the T x / Rx state and

2394

IEEE SENSORS JOURNAL, VOL. 13, NO. 6, JUNE 2013

A RTS RTS

RTS D

Fig. 5.

Idea of parallel transmission using four nodes.

transmit power level. Now, if a node is outside the transmission range of a sender node but within the carrier sensing range then the node cannot perform any parallel transmission while that particular sender node is communicating. It is because, the node outside the transmission range and within the carrier sensing range can sense the packet but cannot successfully decode the packet. So, with the communication range and the sensing range, the reduction of delay (i.e., increased scope of parallel transmission) depends on the topology of the particular (deployment) sensor network. By using the transmission power adjustment as described in Section III-F of this paper we can maximize the set of communication links which can transmit data simultaneously. E. Transmission Each node will sleep for some time and then periodically wakes up to see whether any other node wants to talk to it. The nodes belong to same virtual cluster will follow the same schedule, hence wake up simultaneously. During sleeping, the node turns off its radio, and sets a timer to awake later. If a node wants to send data to another node it will check whether the node itself is the owner of the slot. The node will also check the status of the data, whether it is high priority (or critical) data or not. The node which has any critical (i.e. low latency requirement data) will get the priority for transmission. If there is no such data, then the protocol will consider the ownership of the slot. The node who is the owner of the slot will get priority. If there is more than one owner of the slot then the contention will be taken place between the owners. It should be mentioned here that using modulo might result in multiple owners of a slot, since different node IDs can map to the same slot number. In case of such situation owners will compete among each other for getting access of the medium. Nowif the intended node is not owner of the slot, it will contend with other nodes (non-owner nodes) to get the slot. Broadcast packets are sent without Request-To-Send (RTS) and Clear-To-Send (CTS). Unicast packets will follow the sequence of RTS, CTS, Data, and Acknowledgement (ACK). This scheme is well recognized and used, for example in the IEEE 802.11 standard [24]. Which couple of nodes are allowed to communicate simultaneously, it can be found in the neighbors list table. So, by obeying the status of the mutually exclusiveness of the two link in the neighboring list, only those sensor nodes who has an intention to communicate will contend for successful agreement of data transfer. It should be mentioned that, since the RTS message contains the link information (both sender and the receiver) after seeing the rst RTS, the number of candidate node for desiring the access to the medium will be considerable decreased to meet the parallel transmission requirements. In order to allow more than one pair of nodes to communicate in one duty cycle there is a need for modication of the mechanism of virtual carrier sense that controlled by the Network Allocator Vector (NAV) timer which is used to prevent collision. In IEEE 802.11 and S-MAC protocols, the NAV timer used to block a node from sending and receiving

go to sleep state after successful communication. Nodes that fail contention go to sleep state. D. Identifying the Maximal Set of Communication Link Supporting Collision Free Transmission Simultaneously Assignment Within two hop neighbors simultaneously more than one transmission is possible without collision. The idea can be explained by the following scenario as explained through Fig. 5. We use the similar approach of parallel transmission used in [22]. But we incorporate the concept of transmission power adjustment of sensor nodes in order to nd the superset of mutually collision free transmission link which further support more transmission simultaneously. In Fig. 5, we take four sensor nodes which are within three hops neighbors and belong to the same virtual cluster to explain the situation. We use the IEEE 802.11 scheme as well. Let us consider that node A wants to send a message to node B. First of all A will send a RTS packet to the sensor node B. But node C will overhear the RTS packet because the node C is within the transmission range of node A. By overhearing the RTS the node C concludes with the decision that the transmission medium is busy. Now let us consider at the same time node C has a data to send to D. In the conventional method, since node C knows that the transmission medium is busy it will not send any message to initiate any transmission. But we can see from the gure that the communication (RTS) from node C to node D will not affect the previous communication i.e., communication (RTS) from node A to node B. Since collision happens at the receiving end, and node B and node C are not within the same transmission range, transmission of C by no means will interfere with the reception of node B. Similarly if node B sends a RTS to node A, node A will reply with a CTS. This CTS will be overheard by C. But C need not to stop its transmission and it can send or receive during this time without any collision. It should be noted here that, when a node is within transmission range of a sender node, it can receive and correctly decode packets from the sender node. And when a node is within the carrier sensing range of a sender node, it can sense the senders transmission. Carrier sensing range is typically larger than the transmission range; we use the carrier sensing range as two times larger than the transmission range [23]. Again, the carrier sensing range and transmission range depend on the

ARIFUZZAMAN et al.: IH-MAC WITH TRAFFIC-DIFFERENTIATION-BASED QoS FOR WSNs

2395

Node A

DATA

Listen

Listen

Listen

Node B

DATA

DATA

Sleep

Sleep

node A and C will save energy by avoiding RTS and CTS and contention for getting the slot. Another power savings feature of IH-MAC is by adjusting transmission power, which is explained in the following Section III-F. F. Transmission Power Adjustment

Listen

Listen

Listen

Sleep

Sleep

Listen

Listen

Listen

Listen

Listen

Sleep

Sleep

Sleep

Node C

DATA

DATA

Node D

DATA

Sleep Slot i+2

Slot i

Slot i+1

Slot i+3

Fig. 6.

Timing diagram of sensor nodes working in IH-MAC.

whenever it overheard a packet transmission from other nodes. We modify the scheme in a way that after overhearing the transmission from other node the NAV timer will not directly block a node from sending or receiving rather it will judge the possibility of parallel transmission before doing so and will work accordingly. Now, if messages for a particular node queued in its buffer cross threshold value the node will make some of its owned slots as rendezvous slots. The node will rst broadcast the declaration of its making rendezvous slot. The declaration message contains how many slots will be used as rendezvous slot, and between whom the rendezvous will be done. So, remaining neighboring nodes can calculate locally about the slot so that they need not to wake up during those slots. For each rendezvous slot since all the neighbors of both sender and receiver will be in sleep mode, there will be no hidden or exposed terminal problem. So, on those cycles no RTS-CTS are required. Only DataAck will work. Part of the energy saving scheme of IH-MAC is pictorially represented in the timing diagram of Fig. 6. Consider a simple case scenario of four sensor nodes A, B, C, and D where each node is within transmission range of others and they follow same schedule. In Fig. 5 we consider four consecutive slots namely, i , i + 1, i + 2 and i + 3. Each slot is further divided into two parts, the rst one is listen part which is used for SYNC, RTS and CTS, and the next portion is sleep part, which is used for data transmission between two nodes. The proportion of listen and sleep interval depends on the duty cycle of the operation of sensor nodes. We take some arbitrary transmission to clarify the different consequences of working principle of IH-MAC. Now, during slot i , let data transmission occur between node B and C. But node A and D also need to wake up to see whether there is any data for them. But subsequently they go to sleep because either there is no data to receive or send by them or they lose in contention. Similar situation occurs in slot i + 1 where transmission is occurred between node B and D. In slot i + 2, node A and C created rendezvous between them. So, on that slot node B and D will not wake up rather they will sleep the whole slot. Therefore, during this period node B and D save energy by operating with zero duty cycle, lingering their sleep time as well as by avoiding transition from sleep to active state. And

The power adjustment features of IH-MAC allow the sensor nodes to suitably vary the transmission power to reduce energy consumption. This idea is based on power control protocol for wireless ad-hoc network proposed in [25]. IH-MAC transmits the RTS and CTS packets with maximum power Pmax . When receiver node receives an RTS packet, it responds with a CTS packet at usual maximum power level Pmax . When the source node receives this CTS packet, it calculates Pdesired based on the received power level Pr and transmitted power level Pmax as Pmax Rx thres c Pdesired = Pr where Rx thres is the minimum necessary signal strength and c is a constant. The source node uses power level Pdesired to transmit data packet. Similarly, receiver uses the signal power of received RTS packet to determine the power level to be used Pdesired , for the ACK packet. This method assumes the attenuation between sender and receiver nodes to be the same in both directions. It also assumes the noise level at nodes to be below a certain predened threshold value. Since IH-MAC allows data transmission between only one pair of nodes in a slot and all the neighbors of both sender and receiver sleep during transmission, it overcomes the shortcomings of said technique, like increased collision and degradation of network throughput. G. Contention Window Size for Critical Trafc and Owners Priority IH-MAC protocol ensures three levels of priority i.e., highest priority for critical trafc (generated by either owner or non-owner of a slot), medium priority for normal trafc generated by owner of a slot and the lowest priority for normal trafc generated by non-owner of a slot. The protocol does it by using different contention window size for critical trafc, owners trafc and non-ownerstrafc. When a node acquires a data to transmit, it rst checks whether the data is a critical data or not. If it is a critical data, the node takes a random backoff within afxed time period, Tc . When the backoff timer expires, the node runs CCA and if the channel is found clear, the node transmits data. If the channel is busy then it waits until the channel becomes free and repeats the above process. Now, if the data is not a critical data then it checks whether the node itself is an owner of the time slot or not. If it is an owner then the node waits for Tc and then performs a random backoff within a contention window [Tc , To ]. When the backoff timer expires, the node runs CCA and if the channel is clear it starts transmission. If the channel is busy then the node waits until the channel becomes clear and repeats the above process. On the other hand, having a noncritical data, a non-owner node waits for, T0 and then performs a random backoff within a contention window [To , Tno ] and follow the same procedure

Listen

Listen

Sleep

Listen

Sleep

2396

IEEE SENSORS JOURNAL, VOL. 13, NO. 6, JUNE 2013

as described for the other two cases above. It should be noted here that when a slot is already declared as rendezvous slot for that slot without waiting for contention window the node can initiate transmission as like as a TDMA link scheduling. IV. E NERGY C ONSUMPTION A NALYTICAL M ODEL An analytical model for the energy consumption of nodes for IH-MAC is explained in this section. For simplicity we consider the case where a sensor node is either in broadcast scheduling mode or in a link scheduling mode. Let d be the duty cycle and tSIM be the simulation time and tTX , tRX , tOH , tIDLE , tSLEEP , tTRANS are denoted as the time spent for transmitting, receiving, overhearing, idle listening, sleep, and radio transitions during sleep to wakeup state of a sensor node, respectively. So, tSIM can be expressed as tSIM = tTX + tRX + tOH + tIDLE + tSLEEP + tTRANS and tSIM = tSLOT N . Here, N is total numberof slots during time tSIM . Again tSIM = tW + tR (2) (1)

where tSYNC-RTS , tDATA , tCTS and tACK are required time to send SYNC-RTS, DATA, and to receive CTS and ACK, respectively. Now, when a sensor node receives a packet, it receives SYNC, RTS, DATA and it sends CTS and ACK. So, for receiving a packet energy consumed by receiving node is eRX(W) = E RX tSYNC-RTS + E RX tDATA + E TX(max) tCTS + E TX(right) tACK (7) eRX(R) = E RX tSYNC + E RX tDATA + E TX(right) tACK (8)

Now, let the sensor nodes Poisson arrival rate of transmitting packet is TX and sensor nodes Poisson arrival rate of receiving packet is RX during time tSIM . So, the number of times the sensor node transmits and receives packet during tSIM is n TX(W) = TX tSIM(W) n TX(R) = TX tSIM(R) . Similarly, n RX(W) = RX tSIM(W) (11) (12) n RX(R) = RX tSIM(R) . (9) (10)

(3)

where tW and tR represent time period while IH-MAC operates in broadcast scheduling mode and link scheduling mode respectively. Let n H , n TX , n RX , n OH , represents the total number of times that a node hears, transmits, receives, and overhears during tSIM . A sensor node consumes energy by transmitting (eTX ), receiving (eRX ), overhearing (eOH ), and idle listening (eIDLE ) during the awake state. And during the sleep state very less energy is consumed. During transition (eTRANS) from sleep state to active state energy is also consumed. Since our IHMAC protocol operate both in broadcast scheduling and link scheduling (Rendezvous) and we have used power adjustment technique, so transmitting energy is further divided into two category, without rendezvous, eTX(W) and with rendezvous, eTX(R) . Similarly, receiving energy can be divided into eRX(W) and eRX(R) . Now energy consumption during tSIM can be expressed by e = n TX(W) eTX(W) + n TX(R) eTX(R) + n RX(W) eRX(W) + n RX(R) eRX(R) + tOH eOH + tIDLE eIDLE + tSLEEP eSLEEP + tTRANS eTRANS. (4) Since IH-MAC has the probation of adjusting transmission power we use maximum transmission power as E TX(max) and right transmission power as, E TX(right). When a sensor node transmits a packet, it sends SYNC, RTS, DATA and it receives CTS and ACK. So, for transmitting a packet energy consumed by a transmitting node is eTX(W) = E TX(max) tSYNC-RTS + E TX(right) tDATA + E RX tCTS + E RX tACK (5) eTX(R) = E TX(right) tSYNC + E TX(right) tDATA + E RX tACK (6)

The overhearing of packets and idle listening occur during listen interval. So tOH = n OH(SYNCRTS) tSYNCRTS + n OH(CTS) tCTS (13) and n OH = n H n RX (14) tIDLE = d tW n TX(W) (tSYNC-RTS + tCTS ) n RX(W) (tSYNC-RTS + tCTS ) tOH . (15) The transition from sleep mode to active mode will occur in every slot. So tTRANS = N tSA (16) where tSA represents the time required for switching radio from sleep mode to active mode. So, the energy consumption of a sensor node can be computed analytically using (4). Now, we also develop energy consumption analytical model of S-MAC, one of the fundamental MAC protocol for sensor network, to compare with IH-MAC. In fact S-MAC protocol is the most popular general purpose MAC protocol specially designed for wireless sensor network. For S-MAC the total simulation time, tSIM can be expressed as tSIM = tTX + tRX + tOH + tIDLE + tSLEEP + tTRANS and tSIM = tSLOT N . (18) S-MAC protocol operates like broadcast scheduling and no power adjustment technique is used. Therefore energy consumption during tSIM can be expressed as e = n TX eTX + n RX eRX + tOH eOH + tIDLE eIDLE + tSLEEP eSLEEP + tTRANS eTRANS . (19) (17)

ARIFUZZAMAN et al.: IH-MAC WITH TRAFFIC-DIFFERENTIATION-BASED QoS FOR WSNs

2397

When a node transmits a packet, it sends SYNC, RTS, DATA and it receives CTS and ACK. So, for transmitting a packet energy consumed by transmitting node is eTX = E TX (tSYNCRTS + tDATA ) + E RX (tCTS + tACK ). (20) Now, when a node receives a packet, it receives SYNC, RTS, DATA, and sends CTS and ACK. So, for receiving packet energy consumed by a receiving node is eRX = E RX (tSYNCRTS + tDATA ) + E TX (t CTS + tACK ) (21) Let, sensor nodes Poisson arrival rate of transmitting packet and receiving packet during the time tSIM are same as before. So, the number of times the node transmits and receives packet during tSIM is (22) n TX = TX tSIM Similarly n RX = RX tSIM (23) The overhearing of packets and idle listening occur during listen interval. So tOH = n OH(SYNCRTS) tSYNCRTS + n OH(CTS) tCTS and n OH = n H n RX (25) tIDLE = d tSIM n TX (tSYNC-RTS + tCTS ) n RX (tSYNC-RTS + tCTS ) tOH (26) The transition from sleep to active mode will occur in every slot. So tTRANS = N tSA (27) So, the energy consumption of a sensor node of S-MAC can be computed analytically using (19). For simplicity we avoid considering the collision both for S-MAC and IH-MAC in our analytical model. Now, if we compare (5) and (6) with the (20) we see that the consumed power for a packet transmission for the source node is less in IH-MAC than S-MAC. Similarly, if we compare (7) and (8) with (21) we see that the consumed power for a packet reception for the destination node is less in IH-MAC than S-MAC. Finally if we put these value in (4) and (19) we can conclude that the IH-MAC is more energy efcient than S-MAC. V. R ESULT AND D ISCUSSION In this section, we investigate the performance of the proposed IH-MAC protocol. We have simulated with Castalia, a simulator for Wireless Sensor Networks and Body Area Networks [26] which is developed on the discrete event simulator OMNET++ [27]. In the simulation setup, we take 100 nodes distributed in a uniformly random way on a 100 100 m area grid. The nodes are static and the radio range is chosen so that all the non-edge nodes have eight neighbors. The sink node is chosen on the bottom right corner of the network grid. The duty cycle is chosen 15 percent. The results are averaged over (24)

TABLE I PARAMETERS U SED FOR S IMULATING THE MAC P ROTOCOLS Parameter Name Channel bandwidth Data packet length Transmission power Receive power Idle power Sleep state Frame length Threshold value for the buffer size (for IH-MAC) Duty cycle Value 20 kbps 20 B 36 mW 14.4 mW 14.4 Mw 15W 1s 5 packet 15%

several simulation runs. The parameters used for simulation are listed in Table I. We compare the performance of our proposed IH-MAC protocol with the standard S-MAC protocol, T-MAC protocol and Q-MAC protocol. We took S-MAC and T-MAC protocol because they are widely accepted Medium Access Control protocol for Wireless Sensor Network and the reason behind taking the Q-MAC is that Q-MAC protocol considers the trafc with different priority like our proposed IH-MAC protocol. Performance metrics used in evaluation of IH-MAC protocol are Energy consumption, and Average Packet Latency. Energy consumption of sensor nodes for IH-MAC, S-MAC, T-MAC and Q-MAC are shown in Fig. 7(a). We vary the packet generation interval from 1 to 10 seconds. We see that energy consumption per bit of IH-MAC is less than the energy consumption of S-MAC and Q-MAC for both the heavy and light trafc. The reason behind consumption of less energy is the use of adjusted transmission power in our proposed IH-MAC which is explained in Section III-F. And during heavy trafc the performance of IH-MAC protocol in terms of energy conservation is more evident. It is because during heavy trafc IH-MAC protocol makes some rendezvous slots. Energy consumption in rendezvous slot is less than the energy consumption of a slot of S-MAC as explained with Fig. 6 of Section III-E of this paper. And with comparison of T-MAC, IH-MAC performs better with the heavy trafc. But as the trafc declines IH-MAC cannot create frequent rendezvous slot, as well as the scope of parallel transmission also decreases, hence its energy efciency deteriorates. When the messages inter arrival period increases, i.e., the trafc becomes light, the performance of T-MAC is slightly better than our proposed IH-MAC. It is because T-MAC trades off latency for energy savings. About the prioritized trafc, as Fig. 7(b) shows, the energy consumption is not much effected with the introduction of differential trafc (trafc with different priority). It should be noted that when we use the term of non-prioritized trafc, it means that all the trafc has the equal priority. And in case of prioritized trafc we use (consider) the 5% of the total trafc with critical or high priority and it is generated randomly. Average packet latency of sensor nodes for IH-MAC, S-MAC, T-MAC and Q-MAC for non-prioritized trafc is

2398

IEEE SENSORS JOURNAL, VOL. 13, NO. 6, JUNE 2013

500 450 400 350 300 250 200 150 100 50 0 1

Average Packet delay (Second)

Energy consumpon(J/bit)

S-MAC IH-MAC T-MAC Q-MAC

100 90 80 70 60 50 40 30 20 10 0 1 2 3 4 5 6 7

S-MAC IH-MAC T-MAC Q-MAC

4
(a)

9 10

9 10

Message Inter arrival period(Second)

Message Inter arrival period(Second)

(c)

500 450 400 350 300 250 200 150 100 50 0 1 2

Energy consumpon(J/bit)

Average Packet delay (Second)

S-MAC IH-MAC T-MAC Q-MAC

100 90 80 70 60 50 40 30 20 10 0 1 2 3 4 5
(d)

S-MAC IH-MAC Low priority T-MAC Q-MAC Low Pririty Q-MAC High Priority IH-MAC High Pririty

4
(b)

9 10

9 10

Message Inter arrival period(Second)

Message Inter arrival period(Second)

Fig. 7. (a) Average energy consumption under different trafc loads (nonprioritized trafc). (b) Average energy consumption under different trafc loads (prioritized trafc). (c) Average packet latency under different trafc loads (nonprioritized trafc). (d) Average packet latency under different trafc loads (prioritized-trafc).

shown in Fig. 7(c). We see that, IH-MAC achieve better delay performance compared to other three protocol. It is because of using parallel transmission. And link scheduling feature of IH-MAC further minimizes control signal during transmission. Besides, when the protocol works in link scheduling it also avoid the contention phase which minimizes the overall delay. Again as the messages inter arrival rate increase (the network is running with a light trafc) the delay performance of IH-MAC deteriorates and become almost similar to other three protocols. Average packet latency of sensor nodes for IH-MAC, S-MAC, T-MAC and Q-MAC for prioritized trafc is shown in Fig. 7(d). We use the 5% of the total trafc as the critical trafc of high priority. So though the packet generation is high (message inter arrival rate is low) the number of priority trafc is not high. On the other hand, IH-MAC protocol gives the service to the priority trafc by using small contention window size compared to the normal trafc (trafc with less priority). And for using the parallel transmission the delay will be further minimized. Another most important achievement of our proposed IH-MAC protocol that needed to be mentioned that in the Fig. 7(d) we can notice that, though the IH-MAC is serving the high priority trafc with the lowest minimum delay, still the normal trafc are not experiencing

high delay. Where in case of Q-MAC protocol the normal trafc experience a greater delay in order to achieve the lower delay for trafc with high priority. VI. C ONCLUSION This paper presents IH-MAC; a novel energy efcient hybrid based medium access control protocol for wireless sensor networks. There are three novel contributions in this paper. Firstly, our proposed IH-MAC protocol introduces the use of the concept of link scheduling and broadcast scheduling together. To the best of our knowledge, in a single protocol, IH-MAC is the pioneer which exploits these two concepts (pairwise TDMA and broadcast TDMA) together to obtain optimum resource utilization. Secondly, we successfully identied (and achieved) the possibility of enhancement of the scope of parallel transmission by transmitting a signal (wireless) with the appropriate power (adjusted power). The third novel contribution is the introducing the idea and realization of a decentralized TDMA. We successfully showed that without any centralized scheduling how TDMA can run smoothly (we use clock arithmetic here). We believe these novel ideas will signicantly contribute not only in the area of protocol design of WSNs but also the protocol design of any wireless Ad-hoc network.

ARIFUZZAMAN et al.: IH-MAC WITH TRAFFIC-DIFFERENTIATION-BASED QoS FOR WSNs

2399

R EFERENCES
[1] I. Akyildiz, W. Su, Y. Sankarasubramaniam, and E. Cayirci, A survey on sensor networks, IEEE Commun. Mag., vol. 40, no. 8, pp. 102114, Aug. 2002. [2] W. Ye. J. Heidemann, and D. Estrin, Medium access control with coordinated adaptive sleeping for wireless sensor networks, in Proc. IEEE/ACM Trans. Netw., vol. 12, no. 3, pp. 493506, Jun. 2004. [3] A. Keshavarzian, H. Lee, and L. Venkatraman, Wakeup scheduling in wireless sensor networks, in Proc. 7th ACM Int. Symp. Mobile Hoc Netw. Comput., 2006, pp. 322333. [4] E. Arikan, Some complexity results about packet radio networks, IEEE Trans. Inf. Theory, vol. 30, no. 4, pp. 681685, Jul. 1984. [5] T. Van Dam and K. Langendoen, An adaptive energy-efcient MAC protocol for wireless sensor networks, in Proc. 1st Int. Conf. Embedded Netw. Sensor Syst., 2003, pp. 171180. [6] J. Li and G. Y. Lazarou, A bit-map-assisted energy efcient MAC scheme for wireless sensor networks, in Proc. 3rd Int. Symp. Inf. Process. Sensor Netw., New York, USA, 2004, pp. 5560. [7] V. Rajendran, K. Obraczka, and J. J. Garcia-Luna-Aceves, Energyefcient, collision-free medium access control for wireless sensor networks, Wireless Netw., vol. 12, no. 1, pp. 6378, 2006. [8] M. Arifuzzaman, M. S. Alam, and M. Matsumoto, An efcient hybrid MAC with intelligent sleep scheduling for wireless sensor network, in Proc. ITU-T Kaleidoscope Conf., Dec. 2011, pp. 17. [9] M. Arifuzzaman and M. Matsumoto An efcient medium access control protocol with parallel transmission for wireless sensor networks, J. Sensor Actuat. Netw.. vol. 1, no. 2, pp. 111122, 2012. [10] S. H. Lee, J. H. Park, and L. Choi, AMAC: Trafc-adaptive sensor network MAC protocol through variable duty-cycle operations, in Proc. IEEE Int. Conf. Commun., Jun. 2007, pp. 32593264. [11] R. Yadav, S. Varma, and N. Malaviya, Performance analysis of optimized medium access control for wireless sensor networks, IEEE Sensors J., vol. 10, no. 12, pp. 18631868, Dec. 2002. [12] J. Polstre, J. Hill, and D. Culler, Versatile low power media access for wireless sensor networks, in Proc. 2nd Int. Conf. Embedded Netw. Sensor Syst., New York, USA, 2004, pp. 95107. [13] I. Rhee, A. Warrier, M. Aia, and J. Min, Z-MAC: A hybrid MAC for wireless sensor networks, in Proc. 3rd Int. Conf. Embedded Netw. Sensor Syst., Nov. 2005, pp. 90101. [14] I. Rhee, A. Warrier, J. Min, and L. Xu, DRAND: Distributedrandomized TDMA scheduling for wireless ad hoc networks, in Proc. 7th ACM Int. Symp. Mobile Ad Hoc Netw. Comput., 2006, pp. 190201. [15] S. Ramanathan, A unied framework and algorithm for (T/F/C) DMA channel assignment in wireless networks, in Proc. IEEE 16th Annu. Joint Conf. IEEE Comput. Commun. Soc., Driving Inf. Revolut., Apr. 1997, pp. 900907. [16] S. Li, D. Qian, Y. Liu, and J. Tong, Adaptive distributed randomized TDMA scheduling for clustered wireless sensor networks, in Proc. IEEE Int. Conf. Wireless Commun., Netw. Mobile Comput., Sep. 2007, pp. 26882691. [17] Y. K. Rana, B. H. Ajfandika, and S. Jha, Bandwidth aware slot allocation in hybrid MAC, in Proc. 31st IEEE Conf. Local Comput. Netw., Nov. 2006, pp. 8996. [18] R. Iyer and L. Kleinrock, QoS control for sensor networks, in Proc. IEEE Int. Conf. Commun., May 2003, pp. 517521. [19] Y. Liu, I. Elhanany, and H. Qi, An energy-efcient QoS-aware media access control protocol for wireless sensor networks, in Proc. IEEE Int. Conf. Mobile Adhoc Sensor Syst. Conf., Nov. 2005, pp. 191193. [20] Q. Zhao and L. Tong, QoS specic medium access control for wireless sensor networks with fading, School Electr. Comput. Eng., Cornell Univ., Ithaca, NY, USA, ACSP Tech. Rep. TR-06-03-01, Jun. 2003. [21] V. Bharghavan, A. Demers, S. Shenker, and L. Zhang, MACAW: A media access protocol for wireless LANS, in Proc. ACM SIGCOMM Comput. Commun. Rev., vol. 24, no. 4, pp. 212225, 1994. [22] I. Ammar, I. Awan, and G. Min, An improved S-MAC protocol based on parallel transmission for wireless sensor networks, in Proc. Int. Conf. Netw.-Based Inf. Syst., Sep. 2010, pp. 4854. [23] A. Kamerman and L. Monteban, WaveLAN-II: A high-performance wireless LAN for the unilicensed band, Bell Labs Tech. J., vol. 2, no. 3, pp. 118133, 1997. [24] Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specications, IEEE Standard 802.11-1999, 1999.

[25] J. Gomez, A. T. Campbell, M. Naghshineh, and C. Bisdikian, Conserving transmission power in wireless ad hoc networks, in Proc. 9th Int. Conf. Netw. Protocols, Nov. 2001, pp. 1114. [26] Castalia, a Simulator for Wireless Sensor Networks and Body Area Networks, Version 3.2, National ICT Australia, Sydney, Australia, 2011. [27] OMNET ++ Discrete Event Simulation, Version 4.1, 2010.

Mohammad Arifuzzaman (S13) received the B.Sc. degree in computer science and engineering from the Bangladesh University of Engineering and Technology (BUET), Dhaka, Bangladesh, in 2001, and the Masters degree in Global Information and Telecommunication, Waseda University, Tokyo, Japan, in 2012, where he is currently pursuing the Ph.D. degree. He was an Assistant Professor at IBAIS University, Dhaka, from 2001 to 2005. After that, he joined the Bangladesh Civil Service in 2006 and worked as an Assistant Secretary to the Government of Bangladesh till 2010. Mr. Arifuzzaman was the recipient of many awards including the Best Paper Award in the ITU Kaleidoscope Conference, Cape Town, South Africa, in 2011. His current research interests include communication protocols, wireless ad-hoc networks, sensor networks, 4G Mobile communication systems and Information Centric Networks.

Mitsuji Matsumoto (SM10) received the Ph.D. degree from Waseda University, Tokyo, Japan, in 1994. He has been a Professor with the Graduate School of Global Information and Telecommunication Studies (GITS), Waseda University, Tokyo, Japan, since 2000. He was with NTT Laboratories, Japan, from 1970, where he was engaged in research and development of terminal design for telematics and multimedia systems. In 1996, he joined the Faculty of Science and Engineering, Waseda University, as a Professor. From 2001 to 2004, he was a Vice Chairman of ITU-T SG 16. Since 1993, he has been active in Infrared Data Association (IrDA) activities and was a Vice President in 2006. His current research interests include the engineering design for the next-generation wireless communication systems using infrared, visible light, and radio waves. Dr. Matsumoto is a member of IET, IEICE, IPSJ, and IIEEJ in Japan.

Takuro Sato (F13) received the B.E. and Ph.D. degrees in electronics engineering from Niigata University, Niigata, Japan, in 1973 and 1993, respectively. He joined the Research and Development Laboratories, OKI Electric Industry Co., Ltd., Tokyo, Japan, in 1973. He has been engaged in research on PCM transmission equipment, mobile communications, data transmission technology, and digital signal processing technology. He developed wideband CDMA system for personal communications system and joined the PCS standardization committee in the USA and Japan. He contributed in high speed cellular modem standardization for ITU, 2.4GHz PCS standardization for ITA and wireless LAN standardization for IEEE 802.11. He was a Senior Research Manager and Research Director with Communication Systems Laboratory, OKI Electric Industry Co., Ltd. He served as a Professor with the Niigata Institute of Technology from 1995 and he researched on CDMA, OFDM, personal communication systems and related area. In 2004, he joined as a Professor of GITS, Waseda University, Tokyo, and currently serving as a Dean. His current research interests include wireless sensor network, Mobile IP Network, ICT in Smart Grid, 4G mobile communication systems. Dr. Sato is a Senior Member of IEICE.