transmission power control in mac protocols for wireless ... - CiteSeerX

59 downloads 178 Views 306KB Size Report
Wireless Sensor Networks (WSNs) are a subclass of traditional mobile ad hoc .... piggy-backed in data packets at the “current” ideal transmission power, and ...
UFMG - ICEX DEPARTAMENTO DE CIÊNCIA DA COMPUTAÇÃO

.

UNIVERSIDADE FEDERAL DE MINAS GERAIS

TRANSMISSION POWER CONTROL IN MAC PROTOCOLS FOR WIRELESS SENSOR NETWORKS RT.DCC.011/2005

LUIZ H.A. CORREIA DANIEL F. MACEDO DANIEL A.C. SILVA ALDRI L. DOS SANTOS ANTÔNIO A.F. LOUREIRO JOSÉ MARCOS S. NOGUEIRA

JUNHO 2005

Transmission Power Control in MAC Protocols for Wireless Sensor Networks Luiz H. A. Correia, Daniel F. Macedo, Daniel A. C. Silva Aldri L. dos Santos, Antonio A. F. Loureiro, Jos´e Marcos S. Nogueira



Abstract Medium access control (MAC) protocols manage energy consumption on the network element during communication, which is the most energy-consuming event on Wireless Sensor Networks (WSNs). One method to mitigate energy consumption is to adjust transmission power. This paper presents two approaches to adjust transmission power in WSNs. The first approach employs dynamic adjustments by exchange of information among nodes, and the second one calculates the ideal transmission power according to signal attenuation in the link. The algorithms proposed were implemented and evaluated with experiments, comparing their results with B-MAC, the standard MAC protocol in the Mica Motes 2 platform. Results show that transmission power control is an effective method to decrease energy consumption, and incurs in a negligible loss in packet delivery rates. For node distances of 5m, the proposed transmission power control techniques decrease energy consumption by 27% over B-MAC.

1. Introduction Wireless Sensor Networks (WSNs) are a subclass of traditional mobile ad hoc networks (MANETs), and consist of a large number of sensor nodes, composed of processor, memory, battery, sensor devices and transceiver. These nodes send monitoring data to an access point (AP), which is responsible for forwarding data to the users [1]. Unlike traditional ad hoc networks, in general, it is not possible to replace or recharge node batteries due to the number of nodes deployed or inhospitable environmental conditions. Hence, energy conservation is a critical factor in WSNs. Severe hardware and energy constraints preclude the use of protocols developed for MANETs, which comparatively possess more resources. The strict requirements force networking protocols to be as much energy-efficient as possible. Medium access control (MAC) protocols, for example, modify transceiver parameters or even the topology of the network in order to reduce the energy consumption. One of those parameters is the transmission power that, besides reducing energy consumption, also provides higher throughput, due to the reduced number of collisions and the establishment of links with lower bit error rates [2, 3, 4]. Although an effective mechanism to reduce energy consumption, transmission power control is not implemented in any existing MAC protocol for WSNs. This occurs due to the highly imprecise nature of readings provided by the transceiver, and also due to the restricted resources found in current nodes. Those factors difficult an accurate calculation of the ideal transmission power. ∗

Computer Science Department, Federal University of Minas Gerais Belo Horizonte-MG, Brazil, {lcorreia,damacedo,daniacs,aldri,loureiro,jmarcos}@dcc.ufmg.br.

1

E-mails:

In this paper we propose a transmission power control module for WSNs, which can be employed in any existing MAC protocol. Experiments in the Mica Motes 2 platform [5] show the efficiency of the approaches proposed for power control, considering parameters such as energy and throughput. This paper is organized as follows. Section 2 presents the main sources of energy consumption in communication. Section 3 presents the related work. The methods proposed for transmission power control are described in Section 4. Section 5 describes the evaluated scenarios and presents the results. Finally, Section 6 draws the conclusions and future work.

2. Energy Consumption in Communication Among the hardware components in a sensor node, the highest energy consumer is the transceiver [1]. In particular, the energy consumed by the transceiver is related to events in the communication and network organization. Communication events. Encompasses events such as overhearing (nodes listen to transmissions even if they are not the destination of the packet), idle listening (nodes listen to the medium awaiting transmissions) collision and transmission synchronization. Overhearing and idle listening are mitigated by turning the radio periodically off (called duty cycles), or when incoming transmissions are not driven to the node. Collisions and transmission synchronization are avoided with the use of backoff techniques, medium reservation and the exchange of messages [6]. Network organization. Is related to the network topology and the communication pattern (singlehop or multi-hop). The topology can be modified by altering the transmission power. With shorter communication ranges, the probability of hidden terminals [2, 7] and the number of collisions [8] is smaller, reducing energy consumption. Network organization can also be changed by topology control protocols, which turn off nodes producing redundant or unnecessary data to the application [9]. Collision

A

B

C

D

Transmission using power control Transmission without power control

Figure 1. Adjusting transmission power to avoid collisions.

Existing MAC protocols employ energy-saving techniques which operate only over communication events, ignoring network topology. Transmission power control techniques, however, can be very effective, as shown in figure 1. In this example, if nodes B and D, at the same time, transmit data at the typical transmission power (dashed lines) to nodes A and C, respectively, a collision will occur at node C. If the transmission power is reduced to the minimum necessary to reach the destination of the packet (solid lines), no collisions occur. Besides decreasing the number of collisions, transmission power control has other benefits, detailed below. Transmission Power Control in MAC Protocols for Wireless Sensor Networks - DCC/UFMG - Technical Report: RT.DCC.011/2005

2

Benefits of Transmission Power Control Transmission power control allows several improvements in the operation of WSNs, such as the establishment of links with higher reliability, communication with the minimum energy cost, and better reuse of the medium. Links with higher reliability. When used in conjunction with link reliability assessment algorithms, power control techniques can be used to improve the reliability of a link. Upon detecting that link reliability is below a certain threshold, the MAC protocol increases the transmission power, lowering the probability of receiving corrupted data. Communication at minimum energy cost. When communicating at a fixed transmission power, nodes waste energy since some links already have a high probability of a successful delivery. Hence, the transmission control algorithm could decrease the transmission power to a level where link reliability is still high, but energy consumption is lower. Better reuse of the medium. When nodes communicate at the exact power needed to ensure a successful communication, signal range is nothing broader than it was supposed to. Thus, only nodes which really must share the same space will contend to access the medium, decreasing the amount of collisions in the network. This reduced number of collisions will also enhance network utilization and lower latency times.

3. Related Work Several studies characterized channel propagation in wireless networks. Lal et al. [10] showed that it is possible to identify link reliability using an energy-efficient algorithm. Reijers et al. [11] studied the effect of obstacles and environmental changes on link quality. Also, results showed that propagation is asymmetric and directional. RSSI (Received Signal Strength Indicator) readings were found to be extremely dependent on environmental conditions, thus should be used with caution. Given the irregularity of signal propagation in wireless transmissions, Zhou et al. [12] developed a new propagation model, which closely resembles the results obtained from experimental data. Transmission power control is an active line of study in MANETs. Several MAC protocols employing this technique have been proposed. PCMA (Power Controlled Multiple Access) is a MAC protocol which provides communication at minimum propagation ranges, allowing spatial reuse [2]. Agarwal et al. proposed a distributed power control algorithm for MANETs [7]. Pires et al. improved this algorithm by adding a table in each node, which stores the transmission power used on previous transmissions [13]. In order to mitigate asymmetric links caused by transmission power variation, Jung & Vaidya proposed that transmission power should be adjusted for every transmitted byte [3]. The transmission power control techniques developed for MANETs do not apply to WSNs. Since the calculation performed is complex and imprecise in the transceivers employed in WSNs, current MAC protocols for WSNs do not implement transmission power control [14, 15, 16]. Our implementation of transmission power control improves network operation by minimizing contention and decreasing the amount of energy required for communication. Our solution employs a transmission power calculation adapted to the restrictions found in WSNs. Also, we use tables to store the minimum transmission power, as in current solutions.

4. Identifying the Ideal Transmission Power In order to identify the ideal transmission power, nodes must perform calculations based on several readings from the transceiver and the battery. Those are: Transmission Power Control in MAC Protocols for Wireless Sensor Networks - DCC/UFMG - Technical Report: RT.DCC.011/2005

3

• RSSI (Received Signal Strength Indicator): is the signal strength measured by the transceiver at its input interface. RSSI measurements are used to calculate the noise at the medium and the signal strength when receiving incoming data. • Sensitivity: is the least energetic power level at which the transceiver is able to detect and decode data correctly. If any transmission is received at a power level below this limit, data will be garbled. • Battery voltage: RSSI values read from the radio are calculated with battery voltage as a reference. Thus, in order to convert any RSSI reading to the actual reception power, the voltage at the moment of the reception must be known.

−60

Experimental data Nominal strength Average noise

Signal strength (dBm)

−65 −70 −75 −80 −85 −90 −95 −100

0

10 20 30 40 50 60 70 80 90 100 Distance (m)

Figure 2. Received signal strength when varying the distance among nodes.

In a successful transmission, the received signal strength is superior to the average noise at the receiver (the signal strength sampled when there are no ongoing transmissions on the medium). Communication quality also depend on factors such as distance among the receiver and the transmitter and the existence of sources of reflexion, refraction and dispersion. Figure 2 presents the received signal strength for transmissions at 5 dBm, when varying the distance among the transmitter and the receiver, in the Mica Motes 2 platform. The “Nominal strength” curve shows the expected behavior of the signal, while the “Experimental data” curve shows empirical data. We infer from those curves that reception strength is proportional to the distance between the transmitter and receiver. The average noise, however, suffered no significant alterations. Signal propagation occurs differently for indoor and outdoor environments, thus propagation models are specific to each type of environment [17, 18]. Although such models can be used to provide a fair approximation of the ideal transmission power needed to reach a node, those models are too costly to be executed in a sensor node. Thus, new algorithms suitable to the scarce resources found in WSNs must be developed. Two such methods are described below.

4.1. Assessing the Ideal Transmission Power Through Node Interaction The ideal transmission power can be dynamically determined by the interaction of nodes. Transceivers transmit data at only a few power levels. The transceiver used in the Mica Motes 2 platform, for example, provides 22 different levels, separated at roughly 1 dBm [19]. The switching between different power levels Transmission Power Control in MAC Protocols for Wireless Sensor Networks - DCC/UFMG - Technical Report: RT.DCC.011/2005

4

takes 20 µs. Since the number of possibilities is quite small, it is possible to iterate over the available power levels, increasing or decreasing the transmission power when necessary.

Figure 3. Operation of the interactive algorithm.

The algorithm proposed calculates the ideal transmission power by repeated refinements, and operates in two phases. Figure 3 shows the operation of the algorithm. In the first phase, the ideal transmission power is determined, while in the second phase the transmission power is dynamically adapted to any environmental change. Initially, the ideal transmission power is set to the maximum value allowed by the transceiver. Nodes wishing to determine the ideal transmission power send a power query message (MP Q ) piggy-backed in data packets at the “current” ideal transmission power, and await for a confirmation of reception, such as an acknowledgement (ACK) packet. If the reception is confirmed, the transmitter decreases the ideal transmission power by one level, and sends another MP Q message. When the reception of a MP Q message is not confirmed, the transmitter assumes that the ideal transmission power was found, and the second phase of operation starts. In the second phase of the algorithm, nodes use ACKs to determine if the ideal transmission power should be increased or decreased. If a number of consecutive transmissions are not confirmed with ACKs (this number is called the increase threshold, or LI ), the ideal transmission power is increased one level. Since the noise can also decrease due to environmental changes, communication can also improve, thus the transmission power is lowered if a certain number of consecutive messages are successfully received (this number is called the decrease threshold, or LD ). The values of LI and LD must be adjusted according to the typical throughput of the application, avoiding that the algorithm reacts too late to variations in link reliability when the throughput is low, or that such changes are too frequent when the throughput is high. The algorithm treats node failures and transmission failures as being the same, since the use of ACKs to assess link reliability does not allow a distinction of such events. Broadcast packets are always transmitted at a fixed power, since those packets are not acknowledged. Transmission Power Control in MAC Protocols for Wireless Sensor Networks - DCC/UFMG - Technical Report: RT.DCC.011/2005

5

4.2. Assessing the Ideal Transmission Power Through Signal Attenuation The ideal transmission power can also be calculated as a function of signal attenuation. The ideal power is such that, given the attenuation in the link, data is transmitted at a signal strength that allows the reception at a signal slightly higher than both radio sensitivity and the noise on the receiver. This method works as follows. Nodes periodically sample the signal strength when no transmissions occur, in order to determine the base noise (NB ). If node A wishes to communicate with node B, it transmits a packet to B at the standard transmission power (PT X ). When B receives the packet from A, it determines the received signal strength (PRX , or reception power) and calculates the ideal transmission power (PT Xideal )1 from A to B using equation 1. Next, B sends the calculated power to A, which will transmit subsequent messages to B at this power level. In order to dynamically adjust the ideal power transmission, node A always sends in its packets to B the current transmission power. B, in turn, recalculates the ideal transmission power, and sends this value to A piggy-backed in the ACK messages. ½ PT Xideal = max

RXthreshold SIN Rthreshold × NB , GA→B GA→B

¾ (1)

The ideal transmission power is directional, that is, it depends on the direction of the communication, and must compensate the attenuation imposed by the link ( RXGthreshold ). Also, the reception power must A→B SIN Rthreshold × NB ) at the receiving node. The attenuation be higher than the noise and radio sensitivity ( GA→B from A to B (GA→B , or gain) is the relation of the reception and transmission power ( PPRX ), and is TX considered to be symmetric in our calculations. The signal to noise ratio (SIN R) is the ratio of the reception signal when compared to the noise, and NB is the noise in node B. The transmission power must ensure that the signal is received in B without errors. In order to do so, some values must be determined empirically, such as radio sensitivity (RXthreshold ) and the SINR threshold, since they vary for each transceiver. Finally, the calculated power must lie within the maximum and minimum transmission power allowed by the radio. Sensor nodes provide integer values as output for RSSI readings, which must be converted to values in dBm. Since current sensor nodes do not perform floating point arithmetics, the calculation must be made with integers, compromising its precision. Besides, readings from the transceiver and battery vary over time, thus the calculation must be adjusted to avoid subtle variations. The challenge of calculating the transmission power through attenuation resides in defining a precise, stable and efficient algorithm, which can be implemented with the operations provided by the micro-controller.

4.3. Storing the Ideal Transmission Power To communicate at the ideal transmission power without requiring a calculation before every packet transmission, the protocol stores the current ideal transmission for each neighbor node [13]. Thus, nodes first query the table in order to detect if the ideal transmission power was already calculated. If it was, then data will be sent at this power. If it was not, the power transmission calculation is executed, and the result is stored on the table for future use. Table 1 shows the fields stored in the table. Each node stores the ideal transmission power, coded as the bit configuration that must be fed into the radio in order to transmit at the ideal power (P otT x). A control variable (N oReduce) indicates if the ideal power has been calculated, while the Addr field stores the MAC address of the neighbor. Since the noise is dynamic, and nodes may move or leave the network, entries on the table are invalidated if no transmissions occur after some time. This avoids that nodes transmit data at the wrong 1

The relation in the equation has its terms in mW.

Transmission Power Control in MAC Protocols for Wireless Sensor Networks - DCC/UFMG - Technical Report: RT.DCC.011/2005

6

Field P otT x N oReduce Addr

Size 1 byte 1 byte 2 bytes

Description Ideal Tx. power Ideal Tx. power already calculated? MAC address of the neighbor

Table 1. Fields stored and their memory consumption in the Mica Motes 2 platform.

power after extended periods of silence.

5. Evaluation To evaluate the efficiency of the proposed transmission power control techniques, we conducted experiments in the Mica Motes 2 platform, modifying its standard MAC protocol, called B-MAC, to transmit packets at the ideal transmission power. The version employing the iterative method is called B-MACPCI, while the version employing the attenuation method is called B-MAC-PCA.

5.1. B-MAC Protocol The B-MAC protocol was tailored to event-driven applications [14], and aims to be energy-efficient, avoid collisions and be simple, reducing code size. In order to broaden its applicability, B-MAC provides interfaces to reconfigure most of its parameters. Since B-MAC does not employ channel reservation (RTS/CTS messages), the protocol mitigates collisions with an heuristic called CCA (Clear Channel Assessment), which is used to identify transmissions in the medium. This heuristic periodically samples the signal strength when there are no ongoing transmissions, in order to determine the maximum noise level (the base noise). If the sampled signal strength is higher than the base noise, the protocol detects an ongoing transmission. Idle-listening is minimized with the use of a duty cycle. Nodes periodically sample the channel, using the CCA heuristic, to check for transmissions. If a transmission is not identified, nodes enter the reception mode. To ensure that every packet sent is received by all nodes, preambles must be as long as the inactive period of the duty cycle. This asynchronous channel listening method is called LPL (Low Power Listening).

5.2. B-MAC-PC Protocols The B-MAC-PC protocol was implemented in the TinyOS operating system [20], over the B-MAC protocol. In order to further increase energy savings, power control information is piggy-backed in data and acknowledgement (ACK) packets. The power control information increased the data packet size by 3 bytes (one byte for the transmitted power, and 2 bytes for the sender address). ACK packets are increased in 5 bytes (2 bytes for the sender address, another 2 bytes for the receiver address, and one byte for the ideal transmission power). Besides transmitting additional fields at each packet, B-MAC-PC employs a transmission power table, described in section 4.3. This table stores information of up to 20 neighbors. As nodes might not send unicast messages to every neighbor, the size of the table can be reduced according to the needs of the application. The addition of the interactive transmission power control module increased the code size of B-MAC from 7650 to 8440 bytes, and the RAM memory consumed increased from 242 to 340 bytes. B-MAC-PC using the attenuation method consumed 9600 and 385 bytes in code memory and RAM, respectively. We conducted empirical measurements to determine some parameters used in B-MAC-PC. Those measurements are briefly described below. Transmission Power Control in MAC Protocols for Wireless Sensor Networks - DCC/UFMG - Technical Report: RT.DCC.011/2005

7

Interactive method (B-MAC-PCI). The ideal values for LI and LD were adjusted to minimize packet losses, as shown in figure 4. When increasing LD , less packets are lost, but the method responds slowly to environmental changes. For LI , the behavior is the opposite. LI should be set to a small value, since it responds rapidly to variations in the noise. Figures 5 and 6 show the behavior of the transmission power when varying LD and LI , respectively. Figure 5 shows that higher values of LD keep the transmission power more stable, while smaller values of LI increase energy consumption, since the transmission power is more easily and frequently increased. Figure 6 shows that the value of LI also defines the amplitude of the variation. For small values of LI , errors in bursts might significantly increase the transmission power. 70%

LI = 1 LI = 2 LI = 3 LI = 4

Packets lost (%)

60% 50% 40% 30% 20% 10% 0%

1

2

3

4

8 LD

Figure 4. Average packet losses when varying LD and LI .

Transmission power (dBm)

−6

LD = 1 LD = 8

−8 −10 −12 −14 −16 −18 −20

20

40

60

80

100

120

Packets sent

Figure 5. Transmission power behavior when varying LD .

Attenuation method (B-MAC-PCA). The values of RXthreshold and SIN Rthreshold were defined in experiments where we varied the distance between nodes until the receiver could not decode the transmitted data. The experiments were made in an outdoor area free of obstacles, using two nodes elevated 1.5m from the ground. The transmission power used was 5 dBm. From this experiment we determined SIN Rthreshold as 10 dBm, and RXthreshold as -85 dBm. Transmission Power Control in MAC Protocols for Wireless Sensor Networks - DCC/UFMG - Technical Report: RT.DCC.011/2005

8

Transmission power (dBm)

−6

LI = 1 LI = 4

−8 −10 −12 −14 −16 −18 −20

20

40

60 80 Packets sent

100

120

Figure 6. Transmission power behavior when varying LI .

5.3. Experimental Results We measured the behavior of B-MAC and B-MAC-PC when varying the distance between the receiver and the transmitter. Our experiment consisted of two Mica Motes 2 nodes. The first node sent 400 messages destined to the second node, at a rate of four messages per second. The experiments were made in an outdoor area, free of any obstacles. We chose to use only two nodes in order to avoid interference with other nodes. Nodes were placed 71cm above the ground to avoid reflexion and absorption phenomena. Node distance was varied from 5 to 20 meters. LI was set to one, and LD was set to eight. In B-MAC, we employed the standard transmission power, which is 0 dBm. All the results are presented as an average of five independent experiments, with confidence interval of 95%. Figure 7 shows the delivery rates for the protocols. Both the attenuation method (B-MAC-PCA) and the iterative method (B-MAC-PCI) sustained nearly constant delivery rates when varying the distance among nodes. The iterative method was the most stable, yielding approximately 87.5% of packets delivered. The attenuation method delivered from 79% up to 86% of the packets. B-MAC results, on the other hand, are dependent on the distance among nodes. It ranges from an outstanding 98.5%, when nodes are 5m apart from each other, to 14.2% for 20m of separation. This is mostly due to the reception power, which increases the bit error rate of the channel. For distances of 25 meters (not shown), the transmission power in B-MAC is insufficient, and no packets are received, while both B-MAC-PC protocols kept their delivery rates intact. Figure 8 presents the transmission power. For distances of 5 and 10 meters, B-MAC-PCI and B-MACPCA decrease the standard transmission power from 0 dBm to -5.7 dBm and -4.7 dBm, respectively. Meanwhile, B-MAC still transmits at the standard power, consuming more energy. The use of lower transmission power allows energy savings of 27% and 21% over B-MAC when using B-MAC-PCI and BMAC-PCA, for a distance of 5m. For a distance of 10m, this economy drops to 6.9% and 8%, respectively. For distances of 15m or more, the standard transmission power is barely enough to reach the nodes (as shown in Figure 7), thus BMAC’s link quality decreases. In those distances, B-MAC-PC increases the transmission power in order to maintain an acceptable link quality, also increasing the energy consumed. However, this increase is compensated by a much higher packet delivery rate. The causes of packet losses in B-MAC and in B-MAC-PC are shown in Figures 9 and 10, respectively. In B-MAC, for distances up to 10m, half of the packets lost are due to CRC errors, and the other half is lost because the preamble is not detected. When distance increases, the most important cause of packet Transmission Power Control in MAC Protocols for Wireless Sensor Networks - DCC/UFMG - Technical Report: RT.DCC.011/2005

9

100 Average delivery rate (%)

90 80 70 60 50 40 30

B−MAC−PCI B−MAC−PCA B−MAC

20 10

4

6

8

10

12

14

16

18

20

Distance among receiver and transmitter nodes (m)

Figure 7. Average delivery rate. 4 Transmission power (dBm)

3 2 1 0 −1 −2 −3 −4

B−MAC−PCI B−MAC−PCA B−MAC

−5 −6

4

6

8

10

12

14

16

18

20

Distance among receiver and transmitter nodes (m)

Figure 8. Average transmission power.

losses is preamble decoding. When analysing the results for B-MAC-PC (Figure 10), we identified that a lost preamble is the most significant cause of packet losses, accounting for roughly 76% of the total in the iterative method (“I” bars), and 66% of all losses in the attenuation method (the “A” bars). Thus, both B-MAC and B-MAC-PC would benefit from a preamble seek algorithm more resilient to bit errors. The attenuation and iterative methods showed frequent variations in the transmission power, as shown in Figure 11. This is caused by noise variations, and also due to imprecisions in transceiver readings. Ideally, the transmission power would change only when a significant variation in noise occurs, that is, the methods should be less susceptible to ephemerous variations. We plan to incorporate concepts from signal filtering disciplines in the algorithm, in order to avoid frequent transmission power variations. This is a challenging task, as signal filter algorithms must be small and efficient, in order to run in the restricted processors found in current sensor nodes.

5.4. Simulation Results The precision of the attenuation calculation was evaluated in a PC, using logs from previous experiments. Table 2 shows that results calculated in the node are very close to the expected value, and the Transmission Power Control in MAC Protocols for Wireless Sensor Networks - DCC/UFMG - Technical Report: RT.DCC.011/2005

10

90

No Preamble CRC

80 Packets lost (%)

70 60 50 40 30 20 10 0

0

5

10

15

20

25

Distance among receiver and transmitter nodes (m)

Packets lost (%)

Figure 9. Cause of packet losses in B-MAC. 22 20 18 16 14 12 10 8 6 4

0

5 10 15 20 25 Distance among receiver and transmitter nodes (m) No Preamble (I) CRC (I) No Preamble (A) CRC (A)

Figure 10. Cause of packet losses in B-MAC-PC (interactive and attenuation).

average error is close to the error incurred in rounding. We also ran the calculation in a simulator for the processor found in the Mica Motes 2 platform. Results for 20,480 executions of the code with no compiler optimizations showed that the calculation takes 834.17 cycles (208.54µs) on average. We chose not to use any compiler optimization, since initial testing showed that the compiler was artificially speeding up the calculation by storing frequently used values. Compiler optimizations, however, may be valuable in a real implementation.

6. Conclusion and Future Work Among the tasks performed by sensor nodes, communication consumes most of the energy. Thus, medium access control protocols for Wireless Sensor Networks (WSNs) must employ energy-saving algorithms. Transmission power control is one such mechanism, however it was not used in WSNs due to the limitations imposed by the sensor nodes. This article proposed and evaluated, through experiments and simulations, two transmission control algorithms developed for WSNs. Those methods allow the definition of new protocols, which improve network lifetime and performance by adjusting the transmission

Transmission Power Control in MAC Protocols for Wireless Sensor Networks - DCC/UFMG - Technical Report: RT.DCC.011/2005

11

Transmission power (dBm)

5 4 3 2 1 0 −1 −2

Iterative method Attenuation method

−3 −4

0

50

100

150 200 250 Packet number

300

350

400

Figure 11. Transmission power variation in time. Property Number of executions Average error Maximum error Minimum error First quartile of the error Median of the error

Value 7929 0.2887 dBm 0.7526 dBm 0.0013 dBm 0.1178 dBm 0.2342 dBm

Table 2. Precision of the transmission power calculation.

power to spend the minimum amount of energy needed to reach the receiver with a good communication quality. Results showed that transmission power control is an effective method to decrease energy consumption, and incurs in a negligible loss in packet delivery rates. As future work, we intend to perfect the transmission power algorithms in order to avoid frequent transmission power variations. Also, results showed that the use of different preamble sizes and various preamble identification algorithms might decrease packet losses. Finally, the transmission power control technique must be evaluated in MAC protocols which employ channel reservation, exploring RTS and CTS messages to improve the performance of the transmission power calculation.

References [1] I. F. Akyildiz, W. Su, Y. Sankarasubramaniam, and E. Cayirci. A Survey on Sensor Networks. IEEE Communications, 40(8):102–114, 2002.

Transmission Power Control in MAC Protocols for Wireless Sensor Networks - DCC/UFMG - Technical Report: RT.DCC.011/2005

12

[2] Jeffrey Philip Monks. Transmission Power Control for Enhancing the performance of wireless packet data networks. Doctor of philosophy, University of Illinois at Urbana-Champaign, 2001. [3] Eun-Sun Jung and Nitin H. Vaidya. A power control MAC protocol for ad hoc networks. In Proceedings of the 8th annual international conference on Mobile computing and networking, pages 36–47. ACM Press, 2002. [4] Phill Karn. A New Channel Access Protocol for Packet Radio. In American Radio Relay League – 9th Computer Networking Conference, April 1990. [5] Jason Hill, Robert Szewczyk, Alec Woo, Seth Hollar, David Culler, and Kristofer Pister. System architecture directions for networked sensors. In ASPLOS-IX: Proceedings of the ninth international conference on Architectural support for programming languages and operating systems, pages 93–104. ACM Press, 2000. [6] Luiz H. A. Correia, Daniel F. Macedo, Aldri L. dos Santos, Jos´e M. Nogueira, and Antonio A. F. Loureiro. A taxonomy for medium access control protocols in wireless sensor network. Annales des t´el´ecommunications, 2005. to appear. [7] S. Agarwal, S. Krishnamurthy, R. Katz, and S.K. Dao. Distributed power control in ad hoc wireless networks. In Personal and Indoor Mobile Radio Communication – (PIMRC), volume 2, pages 59–66. IEEE, October 2001. [8] Sameer Tilak, Nael B. Abu-Ghazaleh, and Wendi Heinzelman. Infrastructure Trade-offs Sensor Networks. In First International Workshop on Wireless Sensor Networks and Applications, Electrical and Computer Engineering, 2002. [9] Michaela Cardei and Jie Wu. Energy-Efficient Coverage Problems in Wireless Ad Hoc Sensor Networks. Journal of Computer Communications on Sensor Networks, 2004. [10] D. Lal, A. Manjeshwar, F. Herrmann, E. Uysal-Biyikoglu, and A. Keshavarzian. Measurement and characterization of link quality metrics in energy constrained wireless sensor networks. In IEEE GLOBECOM, pages 172–187, December 2003. [11] Niels Reijers, Gertjan Halkes, and Koen Langendoen. Link layer measurements in sensor networks. In 1st IEEE Int. Conference on Mobile Ad hoc and Sensor Systems (MASS ’04), Oct 2004. [12] Gang Zhou, Tian He, Sudha Krishnamurthy, and John A. Stankovic. Impact of radio irregularity on wireless sensor networks. In Proceedings of the 2nd international conference on Mobile systems, applications, and services, pages 125–138. ACM Press, 2004. [13] Alexandre Andrade Pires, Marcelo Fernandes Fontes, and Jos´e Ferreira de Rezende. Proposta e avalia¸c˜ao de um esquema de controle de potˆencia com mem´oria em redes ad hoc 802.11. In Simp´ osio Brasileiro de Redes de Computadores, Gramado, RS, Maio 2004. [14] Joseph Polastre, Jason Hill, and David Culler. Versatile low power media access for wireless sensor networks. In Proceedings of the 2nd international conference on Embedded networked sensor systems, pages 95–107. ACM Press, 2004. [15] Wei Ye, John Heidemann, and Deborah Estrin. An energy-efficient mac protocol for wireless sensor networks. In Proceedings of the IEEE Infocom, pages 1567–1576, New York, NY, USA, June 2002.

Transmission Power Control in MAC Protocols for Wireless Sensor Networks - DCC/UFMG - Technical Report: RT.DCC.011/2005

13

[16] Koen Langendoen and Gertjan Halkes. Energy-efficient medium access control. In R. Zurawski, editor, Embedded Systems Handbook. CRC Press, 2005. [17] H. T. Friis. A note on a simple transmission formula. Proc. IRE, 34, 1946. [18] William C. Lee. Mobile Communications Engineering. McGraw-Hill Professional, 1982. [19] CC 1000. Chipcon corporation. CC1000 low power FSK transceiver. http://www.chipcon.com, March 2005. [20] Philip Levis, Sam Madden, Joseph Polastre, Robert Szewczyk, Kamin Whitehouse, Alec Woo, David Gay, Jason Hill, Matt Welsh, Eric Brewer, and David Culler. TinyOS: An operating system for wireless sensor networks. In W. Weber, J. Rabaey, and E. Aarts, editors, Ambient Intelligence. Springer-Verlag, New York, NY, 2004.

Transmission Power Control in MAC Protocols for Wireless Sensor Networks - DCC/UFMG - Technical Report: RT.DCC.011/2005

14