Printer Friendly

Design, analysis and implementation of energy-efficient broadcast MAC protocols for wireless sensor networks.

1. Introduction

In wireless sensor networks (WSNs), energy consumption is a challenging issue, since WSNs typically consist of limited battery-powered nodes. The two major sources of energy waste in WSN communications are idle listening and overhearing [1]. Idle listening is employed since the sensor node needs to keep the radio on to receive possible incoming traffic. Therefore, nodes use idle listening for the majority of time, even when there are no transmissions in the air, resulting in relatively high energy consumption. Several asynchronous MAC protocols use duty-cycling, where the nodes periodically alternate between sleep and awake states according to a polling interval, to overcome this problem. When a sender has a packet to transmit in duty-cycling, it sends a preamble longer than the polling interval. Thus, the sender can rendezvous with every receiver, which individually wakes up according to its schedule. The nodes that detect the preamble remain in the awake state, i.e., keep their radios on, so that the following data packet can be received.

However, most previously proposed WSN MAC protocols [1][2][3][4][5][6][7] are designed for unicast communications only, although it is highly desirable to support broadcast applications, such as query processing and route discovery [8]. Numerous broadcast MAC protocols have been proposed for wireless ad hoc networks. Wireless ad hoc networks are similar to WSNs in the sense that they use contention based MAC and multi-hop wireless communications. Therefore, a straightforward method of employing the broadcast operation in WSN MAC can be to adopt one of the broadcast MAC protocols devised for wireless ad hoc networks. Nevertheless, these protocols are not well-suited for WSNs, since they omit the typical characteristics of WSNs, such as asynchronous duty-cycle or energy constraints. Another solution is the naive broadcast extension of the asynchronous duty-cycle technique used in WSN unicast [2]. However, this requires the sender to conduct multiple unicast transmissions to emulate the broadcast operation, since the wake up times of its neighbors are diverse and independent, thus severely wasting energy and wireless bandwidth. Moreover, the sender needs to manage the neighbor information to conduct multiple unicasts, while the simultaneous transmissions at the neighbors increase the contention probability. In summary, we need to address key challenges, such as elimination of redundant multiple transmissions, avoidance of forwarding collisions and minimization of management overhead at the sender, to design an efficient WSN broadcast MAC protocol.

In this paper, we propose an asynchronous energy-efficient broadcast MAC protocol called Alarm Broadcast (A-CAST). A-CAST employs the strobe preambles, as in the X-MAC [2] unicast protocol and reduces energy consumption due to idle listening. However, unlike X-MAC, each preamble piggybacks the residual waiting time for the following data frame transmission. A receiver can avoid unnecessary idle listening by immediately going into sleep and scheduling a just-in-time wake up event, according to the residual waiting time information. In addition, a sequence number is embedded in each of the preambles, which can prioritize the forwarding schedules of the neighbors. A node that receives a preamble with a small sequence number has the advantage to rebroadcast the packet earlier than the node that has a larger sequence number. This reduces packet forwarding collisions.

We first evaluate the energy consumption of A-CAST via mathematical analysis. We then show the analytic results by comparing A-CAST to B-CAST, which is a simple broadcast extension of the well-known B-MAC unicast protocol [4]. We also implement A-CAST, B-CAST, and Asynchronous Duty-Cycling Broadcast (ADB) [15], which is a previously proposed broadcast protocol, on the commercial sensor mote CC2420. We compare their performance through real experiments. Our results show that A-CAST outperforms B-CAST and ADB by up to 222% in terms of energy efficiency.

The remainder of this paper is organized as follows. Section 2 summarizes the previous work of WSN low power MAC protocols. In Section 3, we describe our proposals, A-CAST and B-CAST. Section 4 presents the analytical model. Numerical results are shown in section 5. In Section 6, we discuss the performance of A-CAST in a multi hop environment and consider other factors. Section 7 describes the implementation details and experimental results. Finally, Section 8 concludes our paper.

2. Related Work

Idle listening is the most significant source of energy consumption for most sensor MAC protocols. This motivates the duty-cycling technique that alternates the sleep and awake states. MAC protocols that use duty-cycling can be categorized into the following two approaches: synchronous and asynchronous.

In synchronized protocols, such as S-MAC [1] and T-MAC [5], all nodes are time synchronized, so that they wake up and sleep at the same time. Data exchanges occur during wake up periods only, and the synchronous MACs can avoid idle listening. Synchronized schemes, however, induce severe overhead, especially in large scale WSNs, for global synchronization. Also, global synchronization may be difficult in WSNs, where links are not always reliable [9].

Asynchronous MAC protocols [2][4] do not require synchronization and reduce idle listening by Low Power Listening (LPL) [6]. In B-MAC [4], each node wakes up periodically at each polling interval. When a node awakes, it checks the channel state to determine whether it should remain in awake state or go back to sleep. On sensing the channel as idle, the node immediately goes back to sleep mode, because there would be no packet transmissions until the next wake up time. Conversely, a busy channel indicates imminent packet transmission and the node should remain in the awake state. The sender attaches a preamble longer than the polling interval before each data packet, so that the nodes will be in the awake state when the data packet is transmitted (Fig. 1). However, B-MAC wastes power due to the long idle listening duration at the target node and overhearing at non-target nodes.

X-MAC [2] has been proposed to overcome the drawbacks of B-MAC. Instead of using one long preamble, X-MAC strobes short preambles that each contains the destination address. Therefore, X-MAC wakes up only when the destination receiver either receives or overhears the non-target nodes. In addition, X-MAC employs an immediate ACK mechanism that solves the long idle listening problem at the receiver node. However, X-MAC requires relatively longer CCA (Clear Channel Assessment) check time than in LPL, since the CCA check time at every wake-up moment must be longer than the ACK waiting duration of a sender to safely detect the short preambles.

Several broadcast protocols [10][11][12] for wireless ad hoc networks are proposed, but they are unsuitable for WSNs, since they do not consider energy consumption. Recently, Feng Wang and J. Liu [13] proposed a duty-cycle-aware broadcast protocol in WSNs. Not all nodes are active simultaneously due to the different polling intervals of sensor nodes. Under the strong assumption that the network topology and the active/dormant patterns are known, they develop a centralized optimal solution that balances between forwarding efficiency and delay with reliability guarantees. In addition, a distributed algorithm based on the above centralized solution is contrived but its operation depends on the knowledge of duty-cycles of neighbors.

This requires higher synchronization overhead. DiPippo and et al. [14] proposed a broadcast MAC protocol for WSNs. However, this method requires node synchronization that may induce high overhead.

[FIGURE 1 OMITTED]

[FIGURE 2 OMITTED]

[FIGURE 3 OMITTED]

In ADB [15], the sender uses unicast to broadcast to each neighbor individually. This technique avoids redundant transmissions, utilizing the footer beacon and allows the following hops to quickly begin forwarding the broadcast packet by disseminating the progress information. In addition, it enables the sender nodes to go to sleep again, as soon as all the neighbors have been reached. Despite these advantages, ADB may still suffer when multiple nodes wake up simultaneously. ADB uses multiple beacons, such as a wake-up beacon [16] and an ADB footer beacon. If neighbors wake up concurrently around the end of duty cycle, then ADB cannot schedule the multiple beacons in the current duty cycle (Fig. 2). Thus, some of the transmissions must be scheduled in the next duty cycle. Furthermore, if the data packet size is long, this problem becomes even more severe. In contrast, A-CAST is unaffected by the simultaneous wake up problem.

Opportunistic Flooding [17], tailored for low-duty-cycle networks with unreliable wireless links, makes an energy optimal tree to reduce delay and to avoid redundant transmissions. This protocol assumes that predetermined working schedules should be shared by all one hop neighbors. Thus, local synchronization is required to support the scheduled transmission, incurring overhead. If some neighbors have identical work schedules, simultaneous transmissions from multiple nodes can collide.

X-MAC-UPMA [18][19] is the X-MAC-based broadcast implementation in the UPMA package of TinyOS. Fig. 3 Illustrates the overall operations of X-MAC-UPMA. A sender transmits consecutive replica of a data packet over a duty cycle interval. Neighbors stay awake during the end of transmission of the sender, even if they have already received it. X-MAC-UPMA can be more reliable than A-CAST due to the redundant receptions of data packet, whereas it obviously consumes large energy.

I-Hong Hou et al. [20] proposed an efficient broadcast scheme based on network coding for WSNs to reduce broadcast traffic. Network coding [21] is a promising technique that efficiently reduces transmission count by broadcasting a XORed packet instead of unicasting two packets in sequence, but it requires higher CPU computation and memory overhead for coding and decoding. It is hard to implement network coding on the sensor nodes due to the low performance of sensor processors.

3. Proposed Broadcast Protocols

We focus on the development of asynchronous broadcast MAC protocols, since they are more suitable for broadcasting than unicasting. Conversely, synchronous protocols require a large overhead due to global synchronization. We also adopt the duty-cycling mechanism, since it is efficient in reducing power consumption.

Let us first consider B-CAST, a simple extension of the B-MAC [4] for broadcasting. In, B-CAST, the sender transmits a broadcast packet after a long preamble, instead of a unicast packet. Note that in B-CAST, like the B-MAC protocol, all receiver nodes wake up and wait in the awake state, until the packet is transmitted. Therefore, B-CAST suffers from the same long idle listening problem of the unicast B-MAC protocol.

We also extend the X-MAC for broadcast communications. However, the strobe preamble and immediate ACK mechanisms of X-MAC may not be well suited for broadcast support. If the application is straightforward, the immediate ACK leads to multiple unicasts to individual receivers, undesirable due to the energy consumption and wireless bandwidth. Another problem is that multiple unicasts require the management of neighbor node information at the sender. We modify X-MAC so that the sender transmits short preambles for duration longer than the polling interval to avoid multiple unicasts. We eliminate the immediate ACK and all receivers wait in the long idle listening state, until the data transmission is initiated. This modification, still better than multiple unicasts, negates most of the X-MAC advantages over B-MAC and brings in the long idle listening.

Note that the operation of B-CAST is almost identical to that of X-MAC-UMPA [18]. The sender of B-CAST transmits the long preamble during the entire polling interval to wake up every receiver, and sends a data packet at the end of the long preamble (see Fig. 4)). In contrast, X-MAC-UPMA transmits data packets to the receivers consecutively during the entire polling interval (see Fig. 2). Thus, the receiver operations of both protocols are the same. Furthermore, the energy consumption of both protocols is identical.

We devise a new scheme that solves the long idle listening problem of the naive X-MAC extension. In addition to the broadcast MAC address, we inserted the residual time information in the preambles. The residual time is the remaining time left for the following broadcast data packet transmission. Upon hearing a preamble, a neighbor node returns to the 'sleep' state to wake up just before data transmission begins. Note that the residual time has been previously proposed in unicast transmissions [14][22]. We term the proposed broadcast method A-CAST (Alarm CAST), because the receiver sets an alarm for the next wake up according to the residual time. Fig. 4 shows the operations of B-CAST and A-CAST with some examples. There are three neighbors, RI, R2, and R3, near sender S. A-CAST saves energy by allowing receivers to return immediately to sleep state upon hearing a preamble. In contrast, receivers remain in the awake state in B-CAST.

A-CAST transmits one data packet at the end of a duty cycle. This can cause an unsuccessful reception of the data packet due to the channel error or interference due to hidden terminals. Therefore, we embed a sequence number in each preamble that serializes the rebroadcast schedules of its neighbors. Namely, a node that receives a preamble with a small sequence number has the advantage to rebroadcast the packet earlier than the other nodes that have a large sequence number. In Fig. 5, Rl rebroadcasts the packet earlier than R2, since Rl receives a preamble earlier than R2 does. Another usage of the sequence number is resolving NACK collisions. If a node does not receive a data packet successfully, it asks for retransmission by sending a NACK. The sequence number prevents the NACK collisions by serializing the transmission time of NACK packets. The data packet is sent by unicasting in the case of retransmission.

[FIGURE 4 OMITTED]

[FIGURE 5 OMITTED]

4. Analysis

We compare the performance of A-CAST with that of B-CAST by numerical analysis. In the analysis, we only consider the energy consumption of the radio to highlight the difference between the two algorithms. We first compute the energy consumption of A-CAST and B-CAST. Then, we compute the optimal polling intervals that minimize the energy consumption. In our analysis, we assume a network consists of one sender with N neighbors. All nodes can communicate with each other directly (i.e., single hop network). The sender generates data packets at a fixed rate of [lambda] packets per second. The total energy consumption of this single hop network is composed of the energy used for sender transmission (E(s)), the receiving of N nodes (E(r)\ and the common energy (E(c)) used for polling and sleeping at both sender and receivers.

E (total ) = ([lambda](E (s ) + N[bar.E] (r ))+ (N + 1)E (c )) (1)

4.1 Energy Consumption of A-CAST

In duty-cycling, nodes alternate between four radio states: transmitting, receiving, sleeping, and polling. Let [P.sub.tx], [P.sub.rx], [P.sub.poll] and [P.sub.sleep] denote the power consumption in the transmitting, receiving, polling and sleeping states, respectively. Also, let [T.sub.tx], [T.sub.rx], [T.sub.poll] and [T.sub.sleep] denote the expected time spent in transmitting, receiving, polling and sleeping states, respectively. Table 1 lists all power values and time parameters.

Let us compute the energy consumed by the sender. The energy consumption of the sender is composed of carrier sensing energy and transmitting energy. In this analysis, we assume that the sender does not perform carrier sensing, since there is only one sender. When the sender tries to send short preambles, the sender alternates the 'transmitting' and 'sleep' states to save energy. We assume that [T.sub.xListen], the gap between two consecutive short preambles, equals the duration of sending one short preamble, [T.sub.xmacSeg], since the length of them is quite similar in reality. Accordingly, one half of the [T.sub.interval] is used to transmit short preambles, and the other half is used for sleeping. Therefore, the energy consumption of the sender is given by:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.] (2)

[FIGURE 6 OMITTED]

It is important for receivers to successfully receive and decode a preamble that contains the residual time information. If a receiver wakes up in the middle of a short preamble, it may fail to synchronize with that preamble. The receiver must wake up no later than the beginning of the short preamble to decode a short preamble successfully. Therefore, on average (i.e. worst case+best case/2), the receiver must listen for the duration of two preamble times for successful interpretation (see Fig. 6). The energy consumption of the receiver is given by:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.] (3)

The common energy required for polling and sleeping is given by:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.] (4)

[P.sub.sleep] is much smaller than other power parameters. It does not affect overall energy consumption significantly. Using Eq. (1)-(4), we obtain the energy consumption of the

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.] (5)

4.2 Energy Consumption of B-CAST

As mentioned in the previous section, B-CAST can be substituted with X-MAC-UMPA [18], since they result in the same energy consumption. The transmission energy of B-CAST is the same as that of B-MAC. Therefore, we use the same result derived in [4].

E(s) = [P.sub.tx]([T.sub.interval] + [T.sub.data]). (6)

We assume that the wakes up times of N receivers are uniformly distributed. Therefore, a receiver waits in the idle listening state for half of the polling interval on average. The energy consumption of the receiver is given by:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.] (7)

The common energy of B-CAST is the same as that of A-CAST. Using Eq. (1), (4), (6), and (7), we obtain the total energy consumption, as follows.

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.] (8)

The marginal performance gain of A-CAST compared to B-CAST is derived by subtracting Eq. (8) from Eq. (5), as follows.

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.] (9)

5. Numerical Results

5.1 Energy Consumption vs. polling interval

We evaluate our energy model using the specifications of the two well-known radio modules, CC2420 radio and CC1000 radio. Note that the figures in the "Numerical Results" section are based on the numerical results from theoretical analysis. First, we show the energy consumption of B-CAST and A-CAST, as a function of polling interval. Here, there are ten receivers and the packet generation rate [lambda] is 0.01 packets per second.

The difference of the sender operation of both protocols is that the sender of A-CAST switches to RX state to wait for an early ACK. Fig. 7 shows the energy consumption of a sender. In the case of the CC2420 radio, A-CAST consumes more energy than B-CAST, because CC2420 radio uses higher energy in the RX state than in the TX state (TX:52.2mW, RX:56.4mW). Conversely, in the case of CC1000 radio, A-CAST consumes less energy than B-CAST due to the low power in the RX state (TX:31.2mW, RX:22.2mW). When the polling interval is 100msec, the polling overhead occurring in every polling interval is the main energy consumption factor. However, as the polling interval increases, the transmission overhead of the long preamble also increases.

In Fig. 8 and Fig. 9, we observe that A-CAST consumes less energy than B-CAST at all polling intervals. More specifically, the performance gap widens, as the polling interval increases, because B-CAST wastes energy during the idle listening that increases in proportion to the polling interval. A-CAST, however, eliminates the idle listening issue, since receivers can interpret the data packet transmission time exactly. That is, A-CAST is insensitive to the channel polling interval. In addition, when the polling intervals are 400msec in the CC2420 radio and 500msec in the CC1000 radio, the energy consumption of the receivers increases due to the increased overhead for listening to the long preamble. As shown in the Table 1, the CC2420 radio consumes more energy than the CC1000 radio in both TX and RX states during the unit time. In the case of B-CAST, the sender equipped with the CC2420 radio transmits a long preamble during the entire polling interval to rendezvous with a receiver and the receiver has to listen to the channel until the end of the preamble. Thus, transmission and reception costs of the CC2420 radio are much higher than that of the CC1000 radio.

[FIGURE 7 OMITTED]

[FIGURE 8 OMITTED]

[FIGURE 9 OMITTED]

5.2 Energy Consumption vs. Packet Generation Rate

Next, we evaluate the energy consumption of B-CAST and A-CAST, as a function of packet generation rate. We set the number of receivers to 10, the polling interval to 300msec, while changing the packet generation rate from 0.001 to 0.1.

The energy consumption of the sender in both protocols is quite similar, as shown in Fig. 10. In Fig. 11, in the case of A-CAST, the slope of the energy consumption at the receiver side in CC2420 radio is smaller than that in CC1000 radio due to the high data rate of CC2420 radio (Table 1). The data transmission time of CC2420 is 416 [micro]s per byte and that of CC1000 is 32 [micro]s per byte. That is, the data rate of CC2420 is 1byte/32 [micro]s = 250kbps (kilo bits per second) and that of CC1000 is 1byte/416 fss = 19.20 kbps. This means that CC2420 may send a data packet 13 times faster than CC1000, thus CC2420 consumes much less energy than CC1000 to transmit the same amount of data. Note that CC2420 consumes much more energy than CC1000 for the same amount of time at the saturation condition, since CC2420 can send 13 times more data packets than CC1000. From Fig. 12, as the packet generation rate increases, we see that A-CAST significantly improves energy efficiency, because the reception cost is affected by receiving a short preamble and a data packet except for long idle listening.

[FIGURE 10 OMITTED]

[FIGURE 11 OMITTED]

[FIGURE 12 OMITTED]

5.3 Energy Consumption vs. Number of Nodes

Fig. 13 shows the effect of the number of receivers on the energy consumption of B-CAST and A-CAST. We set the packet generation rate as 0.03 packets per second, the polling interval as 100msec, while varying the number of receivers from 1 to 20 nodes. As the number of receivers increases, the energy consumption of both protocols is proportional to the number of receivers. However, the slope of A-CAST is smaller than that of B-CAST, because receivers of A-CAST are free from idle listening.

We observe that the energy consumption of B-CAST and A-CAST is affected by three factors: packet generation rate, polling interval and number of receivers. We derive the optimal polling interval in the next sub section, considering these factors.

[FIGURE 13 OMITTED]

5.4 Optimal Polling Interval

The optimal polling interval that minimizes the energy consumption of each proposed scheme is computed by differentiating E(total) with [T.sub.interval].

* The optimal polling interval of A-CAST:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.] (10)

* The optimal polling interval of B-CAST:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.] (11)

Fig. 14 shows the optimal polling intervals of the proposed schemes for N=10, when the packet generation rate varies from 0.001 to 0.1. We observe that the optimal polling interval decreases, as more packets are generated. Over the whole range of [lambda], the optimal polling interval of A-CAST is more than twice as large as that of B-CAST for both CC2420 and CC1000.

Fig. 15 shows the energy consumption of A-CAST and B-CAST, when their optimal polling intervals for N=10 are applied. A-CAST conserves energy by almost one-half over the whole range of [lambda]. For instance, when the packet generation interval is 0.1 seconds, A-CAST outperforms B-CAST by 222%.

[FIGURE 14 OMITTED]

[FIGURE 15 OMITTED]

6. Discussion on multi hop environment

We analysed the performance of the two proposed protocols in a single hop environment. Now we discuss the performance of both protocols in a multi hop environment. Note that we should account for the flooding packets from its one-hop neighbors and its two-hop neighbors in a multi hop topology. A-CAST is free from the idle listening problem of receivers due to the residual time information in each preamble. Therefore, the reception energy of A-CAST operating in a multi hop environment is the same as that of A-CAST operating in a single hop network. In contrast, the energy consumption of B-CAST is affected by the number of neighbors.

Next, we compare A-CAST to X-MAC-UPMA to investigate the trade off among various factors, such as throughput, energy efficiency, and delivery delay. In X-MAC-UPMA, a sender transmits consecutive replica of a data packet over a duty cycle interval. Neighbors stay awake during the entire duty cycle, even if they have already received the data packet. Receiving redundant packets is helpful in terms of reliability, whereas it is disadvantageous in terms of energy consumption. The transmission energy consumption of A-CAST and that of X-MAC-UPMA are equal, since the senders of both protocols transmit preambles (data packets) persistently during the entire duty cycle. Receivers of X-MAC-UPMA, however, waste energy during the half duty cycle interval on average. Conversely, the receivers of A-CAST go back to sleep state immediately after receiving the preamble. Consequently, the idle listening of A-CAST converges almost to zero when the polling interval increases. In a single hop network, X-MAC-UPMA could outperform A-CAST in terms of delivery delay. However, in a multi hop environment, the delay of both protocols are almost the same, because the neighbors of X-MAC-UPMA should wait until the start time of the next duty cycle to rebroadcast a data packet.

A trade off exists between reliability (delivery ratio) and energy efficiency. The simple way to increase reliability is to receive redundant packets several times. However, the protocol then wastes more energy, as explained above. We can shorten the delay, using some beacons to notify the wake up event, while these beacons also increase the collision probability and result in wasting energy. There are numerous wireless sensor applications, where each application has a particular objective. Energy efficiency is the most important criterion in some surveillance applications, whereas delivery delay is critical to emergency applications. A-CAST focuses on the energy efficiency with reasonable delay bound.

7. Protocol Implementation

We implemented B-CAST, A-CAST and ADB on the TelosB sensor motes, which follow the IEEE 802.15.4 standard, to validate the analytical model. We inserted an energy measure component into the sensor motes to measure how much time it resides in sleep state and awake state, respectively. The TelosB sensor mote consists of MSP430 Micro Control Unit (MCU) [25] and CC2420 radio chipset [24]. The CC2420 chipset is an 802.15.4 compliant device that operates in the 2.4GHz ISM band with a data rate of 250kbps. The Chipcon CC2420 packetizing radio inserts its own preamble in the front of a payload.

We transmit consecutive short packets with a small gap during the entire polling interval for a TelosB sensor mote to support the long preamble of A-CAST and B-CAST. In the case of A-CAST, the transmission time of one short preamble is around 2.8msec that contains the packet processing overhead of our implementation on TinyOS 1.x. Therefore, the sender transmits 28 short preambles in the case of the 100msec polling interval. B-CAST sends consecutive packets more efficiently when it transmits the same short packets during the polling interval. If we push a packet to the RF buffer, we can continuously send the packet without any processing overhead. Thus, the gap between two consecutive packets is reduced.

In our implementation of A-CAST, each preamble is composed of two bytes of residual time and a one byte sequence number. As mentioned before, a sender transmits short preambles until the end of the duty cycle to wake up neighbors. Thus, the effect of embedding 3 bytes of additional information in a short preamble is negligible in terms of bandwidth utilization. In the case of ADB, we modified the short preamble to be used as the wake-up beacon and the ADB footer beacon.

In our experiment, we deploy 11 sensor motes on a star topology: a sender in the center and ten receivers surrounding the sender. All nodes conduct the duty-cycling and the sender generates broadcast traffic every 5 seconds periodically.

Fig. 16 shows the energy consumption of B-CAST, A-CAST and ADB protocols at different polling intervals, varying from 100 msec to 500 msec. We observe that, as the polling interval increases, B-CAST suffers from long preamble overhead. This implies that the network consumes more energy at 500 msec than at 300 msec to transmit and receive the longer preamble. In the case of A-CAST, however, energy consumption decreases, as the polling interval increases, because receivers can stay in the dormant state for the residual time. We observe that ADB consumes large energy at 100 msec compared to the other polling intervals, since it suffers from the contention among the wake-up beacons. The contending nodes must stay in the receiving state during the contention, thus consuming a large amount of energy for the CC2420. As the polling interval increases, however, ADB consumes less energy, since the contention opportunity among the nodes sending the wake-up beacon was decreased by the long polling interval. The overall performance of A-CAST, ADB, and B-CAST are similar to our analytical model, verifying its accuracy. We believe that the small gap between the experimental result and the numerical analysis stems from the overhead of TinyOS implementation.

[FIGURE 16 OMITTED]

8. Conclusion

Although the unicast MAC protocols have been widely studied and developed, the broadcast MAC protocols for WSNs have not yet been actively investigated, even though their applications have various benefits when used in WSNs. In this paper, we proposed an asynchronous energy-efficient broadcast MAC protocol called A-CAST. A-CAST employs the strobe preamble that informs the residual waiting time of the start of the following data frame transmission. In turn, the receivers reduce the energy consumption, by avoiding the unnecessary idle listening. In addition, a sequence number is added into each preamble to reduce forwarding collisions. Through both mathematical analysis and TelosB sensor mote implementation, we show that A-CAST achieves the best performance, in terms of energy consumption. We believe that A-CAST can be easily deployed in current devices, since it only requires a simple modification at the preambles.

DOI: 10.3837/tiis.2011.06.002

Received February 24, 2011; revised May 5, 2011; accepted May 24, 2011; published June 28, 2011

References

[1] W. Ye, J. Heidemann and D. Estrin, "An energy efficient mac protocol for wireless sensor networks," in Proc. of IEEE INFOCOM, New York, NY, USA, pp. 1567-1576, 2002. Article (CrossRef Link)

[2] M. Buettner, Gary V. Yee, E. Anderson and R. Han, "X-MAC: a short preamble MAC protocol for duty-cycled wireless sensor networks," in Proc. of ACM SenSys, Boulder, Colorado, USA, Nov. 2006. Article (CrossRef Link)

[3] M. Awenuti, P. Corsini, P. Masci and A. Vecchio, "Increasing the efficiency of preamble sampling protocols for wireless sensor networks," in Proc. of IEEE Mobile Computing and Wireless Communications (MCWC), Amman, Jordan, pp. 117-122, Sep. 2006. Article (CrossRef Link)

[4] J. Polastre, J. Hill and D. Culler, "Versatile low power media access for wireless sensor networks," in Proc. of ACM SenSys, Baltimore, MD, USA, Nov. 2004. Article (CrossRef Link)

[5] T. van Dam and K. Langendoen, "An adaptive energy-efficient MAC protocol for wireless sensor networks," in Proc. of ACM SenSys, Los Angeles, CA, Nov. 2003. Article (CrossRef Link)

[6] A. El-Hoiydi, "Aloha with preamble sampling for sporadic traffic in ad hoc wireless sensor networks," in Proc. of IEEE ICC, New York, USA, pp. 3418-3423, Apr. 2002. Article (CrossRef Link)

[7] Injong Rhee, Ajit Warrier, Mahesh Aia and Jeongki Min, "Z-MAC: a hybrid MAC for wireless sensor networks," in Proc. of ACMSynSys, San Diego, California, pp. 511-524, Nov. 2005. Article (CrossRef Link)

[8] Akyildiz, W. Su, Y. Sankarasubramaniam and E. Cayirci, "A survey on sensor networks," IEEE Communications Magazine, vol. 40, no. 8, pp. 102-114, Aug. 2002. Article (CrossRef Link).

[9] Q. Li and D. Rus, "Global clock synchronization in sensor networks," in Proc. of IEEE INFOCOM, Hong Kong, pp. 214-226, Mar. 2004. Article (CrossRef Link).

[10] Yao-Win Hong and A. Scaglione, "Energy-efficient broadcasting with cooperative transmissions in wireless sensor networks," IEEE Transaction on Wireless Communications, vol. 5, no. 10, pp. 2844-2855, Oct. 2006. Article (CrossRef Link).

[11] P. Kyasanur, R. R. Choudhury and I. Gupta, "Smart gossip: an adaptive gossip-based broadcasting service for sensor networks," in Proc. of IEEE Mobile Adhoc and Sensor Systems (MASS), Vancouver, Canada, pp. 91-100, Oct. 2006. Article (CrossRef Link).

[12] F. Stann, J. Heidemann, R. Shroff and M. Z. Murtaza, "RBP: robust broadcast propagation in wireless networks," in Proc. of ACM SenSys, Boulder, Colorado, USA, Nov. 2006. Article (CrossRef Link).

[13] F. Wang and J. Liu, "Duty-cycle-aware broadcast in wireless sensor networks," in Proc. of IEEE INFOCOM, Rio de Janeiro, Brazil, pp. 467-476, Apr. 2009. Article (CrossRef Link).

[14] L. DiPippo, D. Tucker, V. Fay-Wolfe, K. L Bryan, T. Ren, W. Day, M. Murphy, T. Henry, and S. Joseph, "Energy-efficient MAC for broadcast problems in wireless sensor networks," in Proc. of 3rd International conference on Networked Sensing System, Chicago, IL, Jun. 2006. Article (CrossRef Link).

[15] Y. Sun, O. Gurewitz, S. Du, L. Tang and D. B. Johnson, "ADB: an efficient multi-hop broadcast protocol based on asynchronous duty-cycling in wireless sensor networks," in Proc. of ACM SenSys, Berkeley, California, Nov. 2009. Article (CrossRef Link).

[16] Y. Sun, O. Gurewitz and D. B. Johnson. "RI-MAC: a receiver initiated asynchronous duty cycle MAC protocol for dynamic traffic loads in wireless sensor networks," in Proc. of ACM SenSys, Raleigh, NC, Nov. 2008. Article (CrossRef Link).

[17] S. Guo, Y. Gu, B. Jiang and T. He, "Opportunistic flooding in low-duty-cycle wireless sensor networks with unreliable links," in Proc. of ACMMOBICOM, Beijing, China, Sep. 2009. Article (CrossRef Link).

[18] K. Klues, G. Hackmann, O. Chipara and C. Lu, "A component-based architecture for power-efficient media access control in wireless sensor networks," in Proc. of ACM SenSys, Sydney, Australia, Nov. 2007. Article (CrossRef Link).

[19] UPMA Package: Unified Power Management Architecture for Wireless Sensor Networks. Article (CrossRef Link).

[20] I-Hong Hou, Yu-En Tsai, T. F. Abdelzaher and I. Gupta, "AdapCode: adaptive network coding for code updates in wireless sensor networks," in Proc. of IEEE INFOCOM, Phoenix, AZ, Apr. 2008. Article (CrossRef Link).

[21] R. Ahlswede, N. Cai, S. Li and R. Yeung, "Network information flow," IEEE Transactions on

Information Theory, vol. 46, no. 42, pp. 1204-1216, Jul. 2000. Article (CrossRef Link).

[22] S. Lim, Y. Ji, J. Cho and S. An, "An ultra low power medium access control protocol with the divided preamble sampling," in Proc. of UCS 2006, LNCS vol. 4239, pp. 210-224, 2006. Article (CrossRef Link).

[23] Chipcon Inc. CC1000 data sheet. Article (CrossRef Link).

[24] Chipcon Inc. CC2420 data sheet. Article (CrossRef Link).

[25] Texas Instruments Inc., MSP430 MCU, Article (CrossRef Link).

This research was supported in part by MKE (The Ministry of Knowledge Economy), Korea, under the ITRC (Information Technology Research Center) support program supervised by the NIPA (National IT Industry Promotion Agency (NIPA-2010-(C1090-1011-0004)) and by the Korean Science and Engineering Foundation (KOSEF) grant funded by the Korean government (MEST) (No.20100027410) and by the Seoul R&BD Program (WR080951). Seoul National University ICT provided research facilities for this study.

Young-myoung Kang received his B.S. in Computer Science from Gyeongsang National University in Aug 2000. He received his M.S. in Computer Science and Engineering from Seoul National University in 2003. He worked as a mobile handset software engineer at LG Electronics from 2003 to 2006. Since September 2006, he has been a Ph.D. student in Computer Science and Engineering from Seoul National University. His research interests include wireless sensor networks, mesh networks and IEEE 802.11.

Sangsoon Lim received the BS degree in computer science from Sungkyul University, the MS degree in the department of electronics and computer engineering from Korea University in 2005, 2007 respectively. He is currently a PhD degree candidate in the School of Computer Science and Engineering at Seoul National University, Seoul, Korea. His current research interests are in the area of wireless networks including wireless LAN, Wireless Sensor Networks, Cognitive radio networks.

Joon Yoo received his B.S. in Mechanical Engineering from Korea Advanced Institute of Science and Technology (KAIST), and Ph.D. in Computer Science and Engineering from Seoul National University in 1997 and 2009, respectively. He worked as a research assistant professor at University of Seoul in 2009. From 2009 to 2010, he was a post-doctoral researcher at the University of California, Los Angeles (UCLA). Since December 2010, he has been with Bell Labs, Alcatel-Lucent as a Member of Technical Staff. His research interests include vehicular networks, mesh networks, LTE, cognitive radio and IEEE 802.11 and 802.16.

Chong-kwon Kim received the B.S. degree in industrial engineering from Seoul National University, the M.S. degree in operations research from Georgia Institute of Technology, and the Ph.D. degree in Computer Science from University of Illinois at Urbana-Champaign in 1981, 1982, and 1987, respectively. In 1987, he joined Bellcore as a Member of Technical Staff and worked on Broadband ISDN and ATM. Since 1991, he has been with Seoul National University as a Professor in the School of Computer Science and Engineering. His research interests include wireless and mobile networking, high speed network control, distributed processing, and performance evaluation.

Young-myoung Kang (1), Sangsoon Lim (1), Joon Yoo (2) and Chong-kwon Kim (1)

(1) Department of Computer Science and Engineering, Seoul National University Seoul, Korea [e-mail: {ymkang, slim}@popeye.snu.ac.kr, ckim@snu.ac.kr]

(2) Bell Labs, Alcatel-Lucent, Seoul, Korea [e-mail:joon.yoo@alcatel-lucent.com]

* Corresponding Author:Young-myoung Kang
Table 1. Symbols used in analysis and typical values for the CC1000
[23] and a CC2420 [24]

symbol             Meaning                           CC1000    CC2420

[P.sub.tx]         Power in transmitting             31.2mW    52.2mW
[P.sub.rx]         Power in receiving                22.2mW    56.4mW
[P.sub.sleep]      Power in sleeping                 3 ffW     3 fW
[P.sub.poll]       Power in channel polling          7.4mW     12.3mW
[T.sub.interval]   Channel polling period            varying   varying
[T.sub.data]       Time to send one data packet      varying   varying
[T.sub.apl]        Avg. time to poll channel         3 ms      2.5ms
[T.sub.b]          Time to send/receive a byte       416 fis   32 fs
[T.xmacSeg]        Time to send one short preamble   varying   varying
[T.sub.xListen]    Time to receive one ACK packet    varying   varying
[L.sub.data]       Data packet length                50byte    50byte
[L.sub.xmacSeg]    Short preamble length             20byte    20byte
COPYRIGHT 2011 KSII, the Korean Society for Internet Information
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2011 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:medium access control
Author:Kang, Young-myoung; Lim, Sangsoon; Yoo, Joon; Kim, Chong-kwon
Publication:KSII Transactions on Internet and Information Systems
Article Type:Report
Date:Jun 1, 2011
Words:6379
Previous Article:On the fairness of the Multiuser Eigenmode Transmission system.
Next Article:Malicious user suppression based on Kullback-Leibler divergence for cognitive radio.
Topics:

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters