Efficient Power Management based on a ... - Semantic Scholar

Efficient Power Management based on a Distributed Queuing MAC for Wireless Sensor Networks B.Otal, C.Verikoukis, L. Alonso

Publication: Vol.: No.: Date:

in Proc. IEEE Vehicular Technology Conference (VTC-Spring07) Spring 2007

This publication has been included here just to facilitate downloads to those people asking for personal use copies. This material may be published at copyrighted journals or conference proceedings, so personal use of the download is required. In particular, publications from IEEE have to be downloaded according to the following IEEE note: c °2007 IEEE. Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works must be obtained from the IEEE.

Efficient Power Management based on a Distributed Queuing MAC for Wireless Sensor Networks Begonya Otal, Christos Verikoukis

Luis Alonso

Centre Tecnològic de Telecomunicacions de Catalunya Castelldefels, Barcelona (Spain) Email: [email protected] , [email protected]

Universitat Politècnica de Catalunya, UPC Castelldefels, Barcelona (Spain) Email: [email protected]

Abstract—Significant power is consumed at a sensor node when it either transmits a packet or when it receives a packet. In this paper, we study energy efficiency as a function of the number of sensors in the network and the payload length. Additionally, we propose power management solutions based on a distributed queuing MAC protocol for wireless sensor networks. Analytical values for the energy consumption performance are derived as a function of the system parameters. The obtained results show that our proposed MAC scheme outperforms that of IEEE 802.15.4. These benefits are obtained from eliminating back-off periods and collisions in data packet transmissions while minimizing the control overhead and the overall energy consumption. Keywords— Distributed queuing, MAC, IEEE 802.15.4, energy efficiency, power management, sensor networks

I. INTRODUCTION The key concern in wireless sensor network applications is that of extremely low power consumption, since it is often infeasible to replace or recharge batteries for the devices on a regular basis. Medium Access Control (MAC) protocols play a significant role in determining the efficiency of wireless channel bandwidth sharing an energy cost of communication. Therefore, energy consumption is a major metric for IEEE 802.15.4. MAC [1]. In [2], the authors propose a Markov chain model of IEEE 802.15.4 to analyse the throughput and energy consumption in saturation conditions as a function of the number of nodes (N) and the payload length (L). They show that the energy needed to send a utile bit steadily increases with the number of nodes in the network. That is mainly because of increased sensing and collided transmission probabilities. From the results in [2] we consider that IEEE 802.15.4 MAC jeopardizes energy consumption of wireless sensor network applications that require a high number of nodes (e.g. more than 20 sensors). Thus, we introduce hereby another family of MAC protocols that outperforms IEEE 802.15.4 MAC in all the same scenarios. The Distributed Queuing Random Access Protocol (DQRAP) is a random access protocol based on a queuing system shared among nodes. It was proposed for the first time in 1992 by Xu and Campbell [3]. They developed the DQRAP protocol for a TDMA environment proposing an analytical

This work has been funded by the Spanish Government ETECLAS (TEC2005-0736-C02-01/TCM) and PERSEO (TEC2006-10459/TCM) projects

model and showing also by means of computer simulations, how the protocol approaches the performance of the theoretical optimum system M/M/1. DQRAP divides the TDMA slot into a “reservation subslot”, that is further divided into minislots, and a “data subslot”. The basic idea is to concentrate user accesses in the control subslot, while the data subslot is devoted to collision-free data transmission following determined rules. It provides a collision resolution tree algorithm that results stable for every traffic load even over the system transmission capacity. One of the most interesting features of DQRAP is its capacity to behave like an ALOHAtype protocol for light traffic load and to smoothly switch to a reservation system as the traffic load increases, reducing automatically collisions. Alonso et al. [4] presented a version of DQRAP adapted for a CDMA environment in 2000. They showed how the protocol approaches the performance of the optimum queuing system M/M/K, where K is the number of spreading codes being used. In 2003, based on their previous research works, Alonso et al. [5] presented the Distributed Queuing Collision Avoidance (DQCA), which is a distributed high-performance medium access protocol designed for Wireless Local Area Networks (WLAN) environments. We would like here in this paper to further analyze this DQ MAC family from the energy consumption perspective in order to prove its performance under new low-rate wireless personal area networks (LR-WPAN) and low-power sensor network scenarios. Special emphasis in energy saving. The remaining of the paper is organised as follows. We first present a brief overview of IEEE 802.15.4 in section II. Section III is reserved to our proposed new frame structure for wireless sensor networks. In section IV, we present our energy consumption analysis considering IEEE 802.15.4 standard release [1]. Section V evaluates the power management of the new proposed MAC scheme for wireless sensor networks. Finally, section VI concludes the paper. II. IEEE 802.15.4 OVERVIEW Similar to all IEEE 802 wireless standards, the IEEE 802.15.4 release standardizes only the physical (PHY) and medium access control (MAC) layers [1]. A. Physical Layer The IEEE 802.15.4 PHY standard incorporates two physical layers; i) the lower band, 868.0-868.6 MHz (for Europe), plus the 902-928 MHz (for much of the Americas and

the Pacific Rim), and ii) the upper band, 2.400-2.485 GHz (substantially worldwide). Both lower and upper bands employ a form of direct sequence spread spectrum (DSSS). In the lower band, binary phase shift keying (BPSK) with raisedcosine pulse shaping is employed. In the 868-MHz band, a data rate of 20 kb/s and a chip rate of 300 kc/s are used, while in the 902-928-MHz band, a data rate of 40 kb/s and a chip rate of 600 kc/s are used. In the upper band, offset quadrature phase shift keying (O-QPSK) with half-sine pulse shaping is employed at a chip rate of 2 Mc/s, along with a 16-ary orthogonal symbol scheme sent at 62.5 ksymbols/s, resulting in a data rate of 250 kb/s. B. Medium Access Control Layer In IEEE 802.15.4 networks a central controller, called the Personal Area Network (PAN) coordinator, builds the network in its personal operating space. The standard supports three topologies: star, peer-to-peer and cluster-tree. The star topology communication is established between devices and the PAN coordinator; in the peer-to-peer topology any device can communicate with each other device within its range; and in the cluster-tree topology, most devices can communicate with each other within the cluster, but only some of them may connect to the infrastructure. The standard identifies two channel access mechanisms. Beacon-enabled networks use a slotted Carrier Sense Multiple Access mechanism with Collision Avoidance (CSMA-CA), and the slot boundaries of each device are aligned with the slot boundaries of the PAN coordinator. The communication is then controlled by the PAN coordinator, which transmits regular beacons for device synchronization and network association control. The PAN coordinator defines the start and the end of the superframe by transmitting a periodic beacon. The length of the beacon period and hence the duty cycle of the system can be defined by the user between certain limits as specified in the standard [1]. The advantage of this mode is that the coordinator can communicate at will with all nodes. The disadvantage is that nodes must wake up to receive the beacon. In non-beacon mode a network node can send data to the coordinator at will using a simpler unslotted CSMA-CA, if required. However, to receive data from the coordinator the node must power up and poll the coordinator. To achieve the required node lifetime the polling frequency must be predetermined by power reserves and expected data quantity. The advantage of non-beacon mode is that the node’s receiver does not have to regularly power-up to receive the beacon. The disadvantage is that the coordinator cannot communicate at will with the node but must wait to be invited by the node to communicate. III. ENERGY SAVING NEW FRAME STRUCTURE Herewith we present the new adapted frame structure of a distributed queuing (DQ) protocol for wireless sensor networks. Like other DQ family protocols, back-off periods and collisions in data packets are eliminated. DQ protocol performance is independent of the number of transmitting sensors in the system and it is stable under all traffic conditions [3]. Moreover, the adapted DQ protocol to wireless sensor

networks is an energy efficient mechanism that scale well to networks with a large number of sensors. Fig. 1 shows the frame format structure of our adapted proposal for wireless sensor networks, suitable for an evolved 802.15.4 LR-WPAN.

UPLINK SUPERFRAME

Contention Access Window (CAW) T access ARS

ARS

DOWLINK FRAME

Contention Free Window (CFW)

Feedback Frame (FF)

TDATA ARS

m contention periods

DATA

New Frame Twa

ACK

Pr

FBP

IFS

ARS

Variable Lenght

Figure 1. New energy-saving frame format for wireless sensor networks

The whole frame structure comprises an uplink superframe and a downlink frame, seen from th transmitting sensor node perspective. The uplink superframe consists of a Contention Access Window (CAW), with fixed length TACCESS , divided into m contention periods. Within these contention periods Access Requests Sequences (ARS) are sent to gather a position for transmitting data into the Contention Free Window (CFW) of variable length TDATA . ARS are the minimum signal required for the central sensor to detect channel access. That means, the PHY level only needs to detect three different states (empty, success, collision), but no information bits have to be carried through [6]. The downlink frame starts in the worst case after Taw, which corresponds to the maximum time to wait for an acknowledgment frame to arrive following a transmitted data frame. Taw is followed by the acknowledgement (ACK) and the feedback packet (FBP), the latter preceded by a preamble (Pr). The FBP contains some subfields, such as the length of the superframe, in order to indicate the non-transmitting sensors, when the next FBP packet will be sent. FBP is also used to actualize the distributed queues inherent of this DQ family protocols (see section I). Pr enables power management between the CAW and FF for the non-transmitting sensors (see section V). At the very end of the downlink frame an Inter Frame Space (IFS) is added to allow the MAC layer to process the data received from the PHY. The main differences of this DQ frame format with respect to the other dis tributed queuing family protocols are the following. Firstly, the ACK may contain link quality information as in IEEE 802.15.4. Secondly, a new field is introduced in form of a preamble (Pr) to enable synchronization after turning off the radio chip for energysaving purposes. Both these new aspects intent to turn this DQ MAC scheme into a better power management protocol. That is, the coordinating sensor sends, on the one hand the ACK to a specific transmitting node and broadcasts the FBP to all associated hearing sensors, including the ones that have just awaken and synchronised via the preamble. Further, there is now the possibility to transmit data packets of variable length (TDATA ), using the same frame structure, at the same time that energy-saving benefits are maintained. In our evaluations, we have in mind wireless sensor networks with a star-base topology. However, its application to a peer-to-peer scenario is

straightforward as shown in [7]. There, a master-slave configuration is used, where each new master assumes the coordinating role for a period of time.

consumption of transmitting ARS (Pt_ARS). Additionally, Pr is the power consumption in receiving mode, which may also differ from the power consumption of the reception of the FBP (Pr_FBP.).

IV. ENERGY CONSUMPTION ANALYSIS Sensor nodes are limited in stored energy, computational capacity and memory. New protocols and algorithms must be designed with special attention to these differences and above all to their limited and sometimes non-renewable power storage. Significant power is consumed at a sensor node when it either transmits a packet or when it receives a packet.

In (1) in order to get the energy consumption, we compute the average time the transceiver is in each of its four states (transmission, reception, ARS transmission, FBP reception), and multiply this time by the power consumption listed in TABLE 1. Typically, in sensor networks the output power is limited. As a result, the power consumed during receiving mode may be larger than in transmitting mode [7]. Note that for simplicity we consider that both power transmissions and both power receptions have the same value, although they could differ from each other based on hardware specifications.

In order to verify our proposed approach and to get a figure of the obtainable gain, reference [2] energy analytical model has been implemented and will be compared to our energy consumption analysis as a function of payload length (L) and number of sensors (N). Fig. 2 portrays the energy consumption per utile bit in saturation as a function of N derived from [2], using default parameters values defined for 2.4 GHz frequency channels (i.e. minimum back-off exponent, BEmin=3; payload length, L=68 bytes) at its corresponding data rate of 250 kb/s. It can be seen that the energy needed to send a utile bit steadily increases with the number of nodes in the network. That is mainly because of increased sensing and collided transmissions probabilities.

TABLE 1

802.15.4 T RANSCEIVER P OWER CONSUMPTION

Pt

Pr

Pt_ARS.

Pr_FBP

15 mW

20 mW

15 mW

20 mW

The total energy per information bit or utile bit for a data frame transmission can be written as follows,

Ebit =

E frame

(2)

L(bits)

where

E frame = E DATA + E ARS + EACK + EFBP

(3)

and L corresponds to the payload length. Here Ebit units are Joules per utile bit (J/bit). The average DATA time (TDATA) used in (1) can be computed as follows:

T D A T A = t P H Y header + t M A C header + T payload

Figure 2. Energy consumption per utile bit of LR-WPAN 802.15.4 MAC

Here we want to validate our here proposed DQ mechanism versus IEEE 802.15.4 analytical model in [2]. To calculate the sensor energy consumption for the whole DQ data frame duration, we first define the following; EDATA and EARS as the energy consumption of data frames and ARS transmissions respectively, EACK and EFBP as the energy consumption for ACK and FBP frame receptions respectively. These energies can be obtained as,

EDATA = TDATA .Pt EARS = m.t A R S. Pt _ ARS

(1)

EACK = tACK . Pr EFBP = tFBP .Pr_ FBP where Pt is the power consumption in transmitting mode for a data frame, which may differ fro m the power

(4)

Next, we study DQ MAC energy consumption, as a function of the network size (number of sensors) N and payload length L. To get example figures, a scenario with N always-active sensors in saturation has been selected. The reference scenario is defined by a set of parameters provided in TABLE 2, whose fields correspond to IEEE 802.15.4 default values and in the upper frequency band of 2.4 Ghz [1]. TABLE 2 PHY header MAC header Payload Taw

IEEE 802.15.4 SCENARIO PARAMETER VALUES 6 bytes 9 to 25 bytes 8 to 127 bytes 864 µs

ACK Pr FBP SIFS

11 bytes 4 bytes 13 bytes 192 µs

Note that for our DQ protocol the number of contention minislots m is also 3 (i.e. 3 ARS packets). As previously said, the duration of these ARS packets (tARS) could be reduced to a very small value (i.e. between 2 µs and 100 µs), since no data information is needed to be carried through [6]. For our calculations, we will use 100 µs as ARS value, in order to consider the worst case scenario. Fig. 3 characterizes the energy consumption per utile bit of DQ mechanism for

wireless sensor networks derived from (2) and the parameter values of TABLE 1 & TABLE 2. As expected, DQ energy consumption decreases rapidly with increasing packet length L, due to the smaller relative overhead. Note that the DQ mechanism maximum consumption is obtained for small-sized packets and reaches its maximum of 2.5% of the overall normalized energy consumption.

All data transmissions will be successful, except from the ones that would fail anyway due to channel conditions, such as fast fading or Doppler.Thus, although the collision resolution mechanism requires some energy consumption, the complete elimination of data collisions represents a remarkable enhancement in the overall network energy efficiency. A notable feature is that DQ MAC saves more than 50% of energy consumption with respect to IEEE 802.15.4 for a high number of sensors (N>40). V. POWER M ANAGEMENT EVALUATION We have already shown DQ MAC remarkable energysaving gain with respect to IEEE 802.15.4 MAC. Here we want to evaluate DQ MAC additional power management benefits. As already explained in [7], to be able to asses the average power consumption of a sensor in a network, we must characterize the instantaneous power consumption of the transceiver when operating in and switching between states. The transceiver may support four states:

Figure 3. Energy consumption per utile bit of DQ mechanism for wireless sensor networks

Once again, the here presented DQ mechanism main feature is that, not only its throughput model is independent of the number of transmitting sensors in the network [3], but also its energy consumption evaluation. Fig. 4 portrays the achievable estimated energy consumption improvement per utile bit of DQ mechanism versus IEEE 802.15.4 MAC protocol based on [2] and aforementioned parameter values. In saturation conditions, the IEEE 802.15.4 MAC shall deal with a certain level of data collisions, which steadily increases with a high number of sensors in the network (e.g. 20 or 40 sensors). This results in a progressive reduction of the energy consumption efficiency of IEEE 802.15.4 MAC. In contrast, when applying DQ MAC protocol in saturation, no collisions will be produced in the data part of the frame [3] and therefore no energy is wasted.

1.

Shutdown: The clock is switched off and the chip is completely deactivated waiting for a startup strobe.

2.

Idle: The clock is turned on and the chip can receive commands (for example, to turn on the radio circuitry).

3.

Transmit

4.

Receive

As sketched for the uplink in Fig. 5, the IEEE 802.15.4 medium access control procedures introduce a significant overhead, which has consequential impact on the overall energy consumption. In the following, we assume that a sensor will attempt to transmit a single packet per superframe. To do so, it will first listen the beacon, after having preemptively turned on its radio in receive mode. After the beacon is received, the sensor can enter in idle mode. The contention procedure requires at least two channel senses for clear channel assessment (CCA), which requires turning the receiver on. Between the CCAs, the receiver can return to the idle state. The sensor must stay in idle rather than shutdown because of the 1 ms delay to recover form the shutdown state [7]. Once the channel is assessed clear twice, the transmission can start. If the packet is well received, a short ACK is fed back to the transmitter after a minimum time Taw (see section III), when the receiver will be in idle state. contention

Twa

ACK

BEACON

802.15.4 MAC Uplink packet

IFS

time

Pr

Twa

DQ MAC FBP

Uplink packet

ACK

ARS contention

Chip wake-up Idle Radio wake-up

IFS

time

Figure 4. Achievable estimated energy consumption improvement per utile bit of DQ mechanism (LOG)

Receive Transmit

Figure 5. IEEE 802.15.4 MAC vs. DQ MAC overhead for the uplink

Fig. 5 also shows our here proposed DQ medium access control procedure, whose relative overhead compared to that of 802.15.4 is not as appreciable in saturation conditions. In the following, we have assumed the worst case scenario from the energy-saving perspective. That is, a sensor takes part in the access procedure first and must transmit a packet immediately thereafter, following DQ MAC protocol rules [3]. Thus, the sensor may start by sending an ARS within the contention minislots, after having preemptively turned on its radio in transmitting mode, and immediately afterwards the transmission of a previously granted packet can start. If the packet is well received, a short ACK is fed back to the transmitter together with the FBP after a minimum time Taw, when the receiver will be in idle mode. The different power states of a sensor for this example are depicted in Fig. 6. The radio chip of sensor a) switches from shutdown to idle, then transmit, idle and receive mode. Sensor b) is not transmitting any packet, but turns on its radio into transmitting mode (after idle) to make an access request. Then, switches back to idle until FBP shall come, when its radio will be turned into receive mode. Sensor c) remains in shutdown until FBP arrives. Thanks to the introduced preamble, sensor b) and sensor c) will be able to synchronize with the coordinator to get the FBP, which gives all necessary information to update its distributed queues in DQ MAC. power state receive

sensor a)

transmit idle shutdown receive transmit

IFS FBP

Ack 17%

Contention 14%

Transmission 52%

Figure 7. Breakdown of energy per bit

VI.

CONCLUSIONS

In this paper, we have presented a better conditioned energy-saving frame format of an enhanced distributed queuing medium access protocol (DQ MAC) for wireless sensor network scenarios. We have shown how the here proposed DQ mechanism outperforms IEEE 802.15.4 MAC in terms of energy efficiency and power management aspects. A notable feature is that DQ MAC saves more than 50% of energy consumption with respect to IEEE 802.15.4 for a high number of sensors (N>40). Thus, we conclude that our proposal outperforms IEEE 802.15.4 for any number of sensors, but it is especially suitable for wireless sensor networks with a high number of nodes and strong energy constraints.

sensor b)

idle shutdown

REFERENCES

receive

[1]

transmit idle shutdown

6%

11%

sensor c) time

[2]

Figure 6. Power management scenario in DQ MAC

To lower power consumption in future designs, it is valuable to know the power breakdown of the sensor. Fig. 7 presents the power breakdown between the different phases of the DQ MAC protocol in our scenario for sensor a) in Fig. 6. We notice that the effective transmission uses already more than 50% of the total energy, which is an improvement with respect to IEEE 802.15.4, as analyzed in [7]. Less than 15% is spend during contention, which takes chip and radio wake up polices into account. The acknowledgement mechanism uses more than 15% of the energy, mainly because of the necessity of activating the receiver during the acknowledgement waitingtime. For listening to the FBP, around 10% of the energy is used. The rest is used for IFS.

[3]

[4]

[5]

[6] [7]

[8]

IEEE 802.15.4, Wireless Medium Access Control (MAC) and Physical Layer (PHY) Specification for Low-Rate Wireless Personal Area Networks (LR-WPANs), IEEE, October 1 2003. T.R. Park, T.H. Kim, J.Y. Choi, S. Choi, and W.H. Kwon, “Throughput and Energy Consumption Analysis of IEEE 802.15.4 slotted CSMA/CA”, Electronic Letters, Vol. 41, No.18, 1st September 2005. W. Xu, G. Campbell, "A Near Perfect Stable Random Access Protocol for a Broadcast Channel", IEEE Proceedings of ICC'92, Vol. 1, pp. 0370-0374. L. Alonso, R. Agusti, O. Sallent, “A near-optimum MAC protocol based on the distributed queueing random access protocol (DQRAP) for a CDMA mobile communication system”, IEEE J. Sel. Areas Commun., vol. 18, pp. 1701-1718, September 2000. L. Alonso, R. Ferrús, R. Agusti, “MAC-PHY Enhancement for 802.11b WLAN Systems via Cross-layering” IEEE VTC 2003-Fall, Orlando, October 2003. Cambell et. al, Patent Application No. US6,408,009 B1, June 18, 2002. J. Alonso, L. Alonso, “A novel MAC protocol for dynamic ad hoc wireless networks with dynamic self-configurable master-slave architecture ” PIMRC, 5-8 September, Barcelona, Spain. B. Bourgard, F. Catthoor, D. C. Daly, A. Chandrakasam and W. Dehaene, “Energy Efficiency of the IEEE 802.15.4 Standard in Dense Wireless Microsensor Networks: Modeling and Improvement Perspectives, Proc.”, Design Automation and Test in Europe Conference and Exhibition, pp. 196-201, March 2005.