SP-TPC: A Self-Protective Energy Efficient Communication Strategy for IEEE 802.11 WLANs
Youngsoo Kim+* Jeonggyun Yu+ Sunghyun Choi+
+ Multimedia & Wireless Networking Laboratory School of Electrical Engineering
Seoul National University
{yskim, jgyu}@mwnl.snu.ac.kr, [email protected]
* i-Networking Laboratory Samsung Advanced Institute of Technology
Samsung Electronics [email protected] Abstract—We propose a novel energy efficient communication
strategy, called SP-TPC, for IEEE 802.11 Wireless LANs (WLANs). SP-TPC utilizes both transmit power control (TPC) and PHY rate adaptation in order to minimize the communication energy consumption while maximizing the throughput performance. When the TPC is employed in the 802.11 WLAN, the overall performance may be degraded due to the potential increase of hidden stations, and hence the request- to-send/clear-to-send (RTS/CTS) exchange is typically used in conjunction with the TPC in order to eliminate hidden stations.
However, the RTS/CTS exchange is an expensive solution since it itself consumes the energy as well as the precious bandwidth. By taking advantage of the existing clear channel assessment (CCA) mechanism intelligently, SP-TPC does not introduce any extra hidden stations, and hence the RTS/CTS exchange is not needed if hidden stations did not exist without TPC. Through simulations, we demonstrate that SP-TPC outperforms other TPC strategies consistently in terms of both throughput and energy efficiency in small and mid-ranged WLAN environments.
Index Terms—802.11 WLAN, transmit power control (TPC), PHY rate adaptation, RTS/CTS, energy efficient communication, and hidden stations
I. INTRODUCTION
During the last decade, we have witnessed an explosive increase of the Internet and the wireless communication services. Furthermore, accessing the Internet from a mobile/portable device became a natural desire by many users.
In recent years, IEEE 802.11 Wireless LANs (WLANs) have emerged as a prevailing technology for the (indoor) broadband wireless access. The IEEE 802.11 standard defines a single Medium Access Control (MAC) and multiple Physical (PHY) layers [1]. Today’s most WLAN devices implement IEEE 802.11b PHY [3], which supports 1 to 11 Mbps raw data transmission rates at the 2.4 GHz unlicensed bands. IEEE 802.11a is an emerging high-speed PHY, supporting 8 different PHY rates, from 6 to 54 Mbps, at the 5 GHz unlicensed bands [2]. In the IEEE 802.11 WLAN, a transmitting station determines the PHY rate and the transmit power level. Most mobile/portable devices are battery-powered, and hence achieving the energy efficient communications is one of the most important requirements. Moreover, the throughput is another important parameter. Accordingly, we develop a communication strategy, which minimizes the energy consumption while maximizing the throughput performance.
The transmit power control (TPC) can be used to achieve an energy efficient communication strategy. A simple method for the TPC could be to set the transmit power level to the lowest level at which the destination station can receive the frame successfully. However, simply reducing the power level could shrink the transmission and carrier sense ranges of the station, which in turn introduces extra hidden stations where there were no hidden stations when the TPC was not used. The existence of hidden stations can severely degrade the network performance. The 802.11 MAC defines the exchange of the Request-to-Send and Clear-to-Send (RTS/CTS) frames in order to overcome the hidden station problem, and accordingly most, if not all, of the existing TPC strategies employ the RTS/CTS exchange prior to the power-controlled data frame transmission. However, the RTS/CTS exchange is an expensive solution since it itself consumes the energy as well as the precious wireless bandwidth.
In this paper, we present a novel energy efficient communication strategy, which minimizes the energy consumption without compromising the system throughput performance, by combining both TPC and PHY rate adaptation algorithms. Our scheme utilizes an inherent characteristic of the 802.11 PHYs’ clear channel assessment (CCA) mechanism.
Accordingly, by transmitting the PHY preamble/header part of a frame at the maximum power level, and then reducing the power level of the remaining part of the frame, one can achieve a self-protective power-controlled frame transmission. Our scheme does not introduce any extra hidden stations, thus eliminating the needs for the RTS/CTS exchange even if the TPC is employed. We refer to the proposed communication strategy as SP-TPC (Self-Protective Transmit Power Control).
In order to demonstrate the utility of the proposed SP-TPC, we limit our scope to the infrastructure mode only. Moreover, we also consider the uplink transmission, i.e., from stations to the AP, since typically the energy-efficient transmission is not a big issue to the APs, which are mains powered in most situations.
The rest of the paper is organized as follows: Section II presents the related work. A brief overview of IEEE 802.11 MAC and 802.11a PHY is presented in Section III. The proposed energy-efficient communication strategy, SP-TPC, is presented in Section IV. After evaluating our strategy via simulations in a comparative manner in Section V, we conclude in Section VI.
This work was in part supported by Samsung and Information Technology Research Center (ITRC).
II. RELATEDWORK
On top of the TPC, a power management policy can be used for an energy-efficient communication in the 802.11 WLAN [4][5]. A WLAN device can be in one of the following states: transmit state, receive state, idle state, or doze state. It consumes the highest power in the transmit state and very little energy in the doze state. Accordingly, the power management policy switches the WLAN device to the doze state adaptively at appropriate moments to save battery energy. The TPC scheme considered in this paper is considered complementary to the power management policy by reducing the communication energy consumption in the transmit state.
There have been remarkable efforts recently to develop an energy-efficient communications in the IEEE 802.11 WLAN based on the TPC [7][8][10][11]. The RTS/CTS exchange is adopted in the DCF-based WLAN environments in order to eliminate the hidden stations [8][10][11]. The authors of [10]
identified the fact that the RTS/CTS exchange does not eliminate hidden stations completely since the stations located between the carrier-sense range and the transmission range cannot receive the RTS frame correctly. In order to handle this problem, they proposed a scheme, which utilizes the Extended Inter-Frame Space (EIFS) assuming that the EIFS is used instead of Distributed Inter-Frame Space (DIFS) whenever it senses other station’s transmission, but cannot decode the transmission correctly. However, the proposed usage of the EIFS is different from that defined by the 802.11 DCF, and hence is not compatible with the standard-compliant 802.11 devices. We use the standard-compliant EIFS operation for the proposed SP-TPC. On the other hand, the authors of [8] have identified that the carrier sense range is not larger than the transmission range in the IEEE 802.11a-based WLAN.
The schemes in [7][8][11] utilize both TPC and PHY rate adaptation in order to achieve an energy efficient communication. Especially, the schemes in [7][8] are targeted to minimize the energy consumption without the throughput performance consideration. On the other hand, a number of schemes were reported to maximize the network throughput via the PHY rate adaptation [6][9]. We assume that the throughput performance is as important as the energy efficiency.
Accordingly, our strategy is developed to minimize the communication energy consumption while not sacrificing the throughput performance.
III. IEEE802.11OVERVIEW A. IEEE 802.11a PHY
In this paper, we consider the IEEE 802.11a PHY operating at 5 GHz bands. It is based on Orthogonal Frequency Division Multiplexing (OFDM), and supports 8 different transmission rates using different combinations of modulation and convoluational code schemes as shown in Table I.
Table I. Eight transmission modes of IEEE 802.11a PHY.
Mode m Modulation Code Rate r PHY Rate (Mbps)
1 BPSK 1/2 6
2 BPSK 3/4 9
3 QPSK 1/2 12
4 QPSK 3/4 18
5 16-QAM 1/2 24
6 16-QAM 3/4 36
7 64-QAM 2/3 48
8 64-QAM 3/4 54
When a MAC frame is forwarded to the underlying PHY for its transmission, the PHY generates a PHY Protocol Data Unit (PPDU) as shown in Fig. 1. Two important subfields in the PLCP header are RATE and LENGTH fields. As illustrated in Fig. 1, the SIGNAL field, which is the first part of the PLCP header, is composed of one OFDM symbol transmitted at 6 Mbps. The rest of the PPDU is transmitted at the PHY rate specified in the RATE field. The LENGTH field specifies the length of the MAC frame contained in the Data field.
RATE4 bits Reserved 1 bit LENGTH
12 bits Parity
1 bit SERVICE
16 bits PSDU Tail Pad Bits
6 bits
Data
Variable Number of OFDM Symbols SIGNAL
One OFDM Symbol PLCP Preamble
12 Symbols
Coded/ OFDM (RATE is indicated in SIGNAL) Coded/ OFDM
(BPSK,r=1/ 2) PLCP Header
Tail 6 bits
Fig. 1. PPDU format of IEEE 802.11a PHY
A major function of the PHY is to indicate the channel status, i.e., busy or idle, to the MAC, and this function is referred to as the clear channel assessment (CCA). The CCA of the 802.11a is based on both carrier sensing and energy detection. We introduce another important component of the CCA mechanism, which is defined in Clause 12.3.5.10.3 of [1].
When a PHY receives and decodes the SIGNAL field (of the PLCP header) successfully, the values of the LENGTH and RATE fields are known. Using this information, the duration of the incoming frame can be calculated. Note that the channel status is set to busy at the beginning of a frame reception.
However, in the middle of a frame reception, the carrier of the incoming frame may be lost. However, even if the carrier is lost, the receiving PHY assumes that the channel is busy, and hence holds the channel busy indication to the MAC until the period indicated by the LENGTH field has expired. At the end of the frame transmission period, the PHY indicates two facts to the MAC: (1) the channel idle status, and (2) the end of a frame reception with an error, i.e., carrier lost. This is a key characteristic of the IEEE 802.11 PHY and MAC interaction, which is utilized for SP-TPC.
B. IEEE 802.11 MAC
IEEE 802.11 MAC provides two medium access control schemes, namely, Distributed Coordination Function (DCF) and Point Coordination Function (PCF). In this paper, we consider the DCF-based WLAN. The energy-efficient communication under the PCF was studied in [7]. The DCF is based on Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA). A station should determine that the medium has been idle for a Distributed Inter-frame Space (DIFS) prior to initiating a frame transmission. After a DIFS idle period, the station chooses a random back-off interval from the range [0, CW], where CW is the contention window size.
The back-off counter is decreased by one for each idle slot time. If the back-off counter becomes zero, the station transmits the frame.
There are two types of methods to determine whether the medium is idle or busy: (1) physical carrier sense, and (2) virtual carrier sense. The physical carrier sense is the CCA mechanism explained in the previous subsection. The virtual carrier sense works as follows. Inside the header of each MAC frame, there is a duration field which indicates the period (in µsec) of a subsequent frame transmission. Once a station successfully receives and decodes a frame, it sets a counter, called Network Allocation Vector (NAV), to the value found in
the duration field, at the end of the frame reception. The NAV counter value decreases every µsec regardless of the medium status. The MAC considers the medium busy while the NAV has a non-zero value irrespective of the CCA indication from the PHY, and hence it is called virtual carrier sense.
Normally, the DCF uses a DIFS interval before a backoff procedure. However, an EIFS interval shall be used instead following an unsuccessful frame reception. There are basically two different cases of an unsuccessful frame reception: (1) the PHY has indicated the erroneous reception to the MAC, e.g., carrier lost; or (2) the error is detected by the MAC via an incorrect CRC reception. The EIFS is defined to provide enough time for other stations to wait for an ACK frame of an incorrectly received frame. Accordingly, the EIFS value is determined by the sum of one SIFS, one DIFS, and the time needed to transmit an ACK frame at the underlying PHY’s lowest mandatory rate, i.e., TEIFS=TSIFS+TACK Mbps,6 +TDIFS. Fig.
2 illustrates that the stations receiving the data frame incorrectly defer for an EIFS period before starting a backoff procedure.
D ata A C K
N A V S o urce
statio n D estinatio n station
O ther statio ns receiving D ata fram e co rrectly
O ther statio ns receiving D ata fram e inco rectly
D IFS
S IFS
E IFS
B ack- off
B ack- off B ack- off
Fig. 2. DCF access operation, where the ACK frame is transmitted at 6 Mbps
IV. PROPOSEDENERGYEFFICIENT COMMUNICATIONSTRATEGY A. Protecting TPC using LENGTH Field
We have seen above that once a receiving station’s PHY gets the RATE and LENGTH fields from the PLCP header of an incoming frame, the PHY sets the CCA busy until the end of the incoming frame’s entire transmission period. Moreover, if the station happens not to receive the frame correctly, it uses the EIFS instead of the DIFS.
Now, we utilize these characteristics of the 802.11 to protect the power-controlled frame transmission. Let us assume that there are a number of stations in one BSS. The distance between each station is close enough to listen to each other.
There are no hidden stations if all frames are transmitted at the maximum power level. According to the previous lesson, we can easily imagine that we can reduce the power level after the PLCP header part without introducing any hidden stations.
Moreover, the rest of the frame transmission as well as the subsequent ACK frame transmission will be protected thanks to the CCA mechanism of the PHY and the EIFS usage by the MAC. Fig. 3 shows the transmit power level adaptation during a frame transmission. The duration for the PLCP preamble and header (or SIGNAL field exactly) is 20 µsec while the actual frame transmission duration is much longer. Therefore, we can easily imagine that the energy saving due to our strategy could be considerable.
Different from typical TPC strategies, SP-TPC does not require the preceding RTS/CTS exchange. Note that SP-TPC protects a frame transmission by itself without an extra
protection mechanism, i.e., RTS/CTS, and hence was named Self-Protective TPC (SP-TPC.) The RTS/CTS exchange is apparently not free. Therefore, it is a good idea to avoid the RTS/CTS usage if possible.
D ata P C LP
H ead er P LC P P ream b le
Pm ax
Pi 0
Fig. 3. Illustration of the transmit power level adaptation during a frame transmission under SP-TPC
B. TPC and PHY Rate Adaptation Strategy
Now, we consider the algorithm to select the transmit power level to use after transmitting the PLCP preamble and header. More precisely, we consider the transmit power level selection and PHY rate adaptation jointly in order to utilize the multiple PHY rate support of the 802.11a PHY. In this paper, we pursue both the energy consumption minimization and the throughput maximization, because both are equally important.
We first consider the throughput performance as a function of the path loss for different PHY modes as shown in Fig. 4 under the following conditions: (1) there is a single transmitter and a single receiver, (2) the maximum transmit power, 23 dBm, is used, (3) the payload size of 1000 octets is used, and (4) the background noise is assumed −93 dBm.
We develop two thresholds in order to choose the transmit power and the PHY rate. These threshold values are derived from the throughput curves in Fig. 4.
1) PHY rate threshold, Th_ri, is determined such that the throughput curves of PHY modes i and i–1 cross each other at path loss Th_ri. For example, Th_r6 = 100.3 dB.
2) Transmit power level threshold, Th_pi, is determined such that the throughput of PHY mode i at path loss Th_pi is α–
percent of the maximum throughput of PHY mode i. In this paper, we choose α = 99%, and for example, Th_p6 = 99.5 dB.
Fig. 4. Throughput performance for each PHY mode vs. path loss
The smaller the value of α, the more we sacrifices the throughput performance. The transmit power level threshold values for different PHY rates are shown in Fig. 4. The following conditions hold irrespective of the frame length: (1) Th_r1 = +∞ and Th_r9 = -∞1, and (2) Th_ri > Th_pi, Th_ri >
Th_ri+1, and Th_pi > Th_pi+1. Basically, our algorithm works as follows:
1By definition,Th_r9 does not exist, but is virtually defined to be used in the algorithm
1) Before a frame transmission, the frame length is determined, and the station determines the path loss, PLr, between the AP and itself.
2) If the PLr value is in the range between Th_ri and Th_ri+1, i.e., Th_ri+1 < PLr ≤ Th_ri, the station decides to use the PHY rate of mode i for its transmission, and selects Th_pi
for the threshold of the transmit power level selection.
3) The transmit power level, Pnext, for the frame transmission is determined as follows:
( _ ), if _ ,
, otherwise,
max r i r i
next max
P PL Th p PL Th p
P P
+ − <
=
where Pnext and Pmax (= 23) are in dBm, and PLr and Th_pi are in dB.
Fig. 5. PHY rate selection vs.
path loss
Fig. 6. Transmit power selection vs. path loss The results of the PHY rate selection and the transmit power level selection are shown in Fig. 5 and Fig. 6, respectively. We observe that lower PHY rates are selected as the path loss increases. Through our algorithm, a station can reduce the transmit power level to save the communication energy while attaining 99% of the maximum throughput performance.
V. COMPARATIVEPERFORMANCEEVALUATION In this section, we present the throughput performance and the energy consumption of the proposed SP-TPC in comparison with other strategies. We compare three different strategies:
1) SP-TPC: the proposed scheme with the maximum- powered PLCP preamble/header;
2) RTS-TPC: the scheme protected by the maximum- powered RTS/CTS exchange;
3) UP-TPC (meaning Unprotected TPC): the scheme without any protection, i.e., neither the maximum- powered PLCP preamble/header nor RTS/CTS exchange.
For all three schemes, we use the algorithm presented in Section IV.B in order to select the transmit power level and the PHY rate. To evaluate these schemes, we have used the ns-2 network simulator.
A. Simulation Model and Setup
We adopt the energy consumption model of the 802.11a device presented in [7][8] for our simulations. The parameter of power characteristics used in the simulation are the common power consumption (Pcom = 500mW), the receive power (Prec = 50mW), and the transmit power (Ptx= –19 to 23 dBm). The noise level is –93dBm. In the simulation, we consider an infrastructure BSS composed of a single AP and multiple stations. We consider both star and random topologies for our simulations. Moreover, all stations transmit frames to the AP in
a greedy mode. The log-distance path loss model with the path loss exponent of four is used.
To evaluate the effectiveness of SP-TPC, we consider two performance metrics. The first metric is the throughput (in Mbps), which is calculated as the total amount of data delivered to the AP, divided by the total simulation period. The second metric is the energy efficiency (in Mbits/Joule), which is determined by the total amount of data delivered to the AP, divided by the aggregated amount of energy consumed by all stations.
B. Simulation Results
First, Fig. 7 and Fig. 8 illustrate the throughput and the energy efficiency of three strategies for the star topology with 8 stations as the radius of the circle increases. From both figures, we observe that both throughput and energy efficiency decrease as the radius increases. Further observations can be made for three different radius ranges:
1. Radius within (5 m, 15 m) range: in this range, SP-TPC and UP-TPC equally achieve the maximum throughput performance, while RTS-TPC achieves a lower throughput due to the RTS/CTS overheads. We observe a similar trend in the energy efficiency, while UP-TPC is slightly better than SP-TPC due to the maximum- powered PLCP preamble/header of SP-TPC.
2. Radius within (15 m, 23 m) range: in this middle range, the throughput of UP-TPC rapidly decreases as the radius increases, and becomes almost zero because of the hidden station effects, while SP-TPC continues to achieve the maximum throughput.
3. Radius within (23 m, >45 m) range: in this long distance, the throughput of SP-TPC becomes worse than that of RTS-TPC, because of the hidden stations. Hidden stations are introduced beyond this range since the transmitted signal cannot reach the opposite side of the circle even if the maximum power is used. One notable fact is that the energy efficiency of SP-TPC does not drop as rapidly as that of UP-TPC in the second region thanks to its self-protection mechanism.
Fig. 7. Throughput for the star
topology Fig. 8. Energy efficiency for the star topology In summary, SP-TPC is observed to achieve the best throughput and energy efficient performance up to the mid- ranged circle when the star topology is used.
In order to consider more realistic situations, we now simulate with the random topology where 8 stations communicate inside the circle of radius 20 m transmitting 1000 bytes packet. We simulate 50 scenarios (or randomly-generated topologies), where the throughput performances for different scenarios are shown in Fig. 9. We observe that SP-TPC consistently outperforms the other strategies. Depending on the simulated topology, the throughput performance fluctuates, where UP-TPC fluctuates the most heavily depending on
whether hidden stations exist or not. Note that RTS-TPC performs worse consistently than SP-TPC due to the RTS/CTS exchange overheads. We do not show the energy efficiency performance as it shows a similar trend.
Fig. 9. Throughput for randomly-generated topology scenarios We perform more simulations by varying these two parameters to see their effects. The random topologies are used for the rest of the results as well. First, Fig. 10 shows the simulation results as the number of stations varies, where the frame payload size is again fixed at 1000 octets. The numbers of stations considered are 2, 4, 6, and 8. The result of each simulation is averaged from 50 different scenarios. We understand that the throughput of SP-TPC decreases as the number of stations increases, since the collision probability increases. On the other hand, the throughput of RTS-TPC does not change much for different number of stations. Finally, UP- TPC performs worse as the number of stations decreases. We observe a different trend in general from the energy efficiency curves. That is, the energy efficiency curves decrease for all three strategies as the number of stations increases. This is due to the fact that stations consume more energy in the idle and receive modes as the number of stations increases. However, SP-TPC is till consistently better than the other strategies.
(a) Throughput (b) Energy efficiency Fig. 10. Comparison with various numbers of stations
(a) Throughput (b) Energy efficiency Fig. 11. Comparison with various frame payload sizes Finally, Fig. 11 shows the simulation results for various sizes of the frame payload where 8 stations are randomly located. We consider the payload sizes of 100, 500, 1000, and 1500 octets. We clearly observe that SP-TPC outperforms the other two strategies for every frame payload size in terms of
both throughput and energy efficiency. One interesting observation is that UP-TPC performs almost the same as SP- TPC for small payload sizes, and becomes relatively worse as the payload size increases. This is because of two reasons.
First, if the payload size is large, the penalty of a collision becomes relatively high. Second, when the payload size is small, the chance of a collision becomes slim.
VI. CONCLUSIONANDFUTUREWORK In this paper, we present a novel transmit power control and PHY rate adaptation strategy, called SP-TPC, which can save the communication energy consumption without compromising the throughput performance. Under SP-TPC, a PHY rate is selected in order to maximize the throughput performance, and then a lower transmit power is selected, where the power level is the lowest level at which the nearly-maximum throughput can be achieved. For the frame transmission, the maximum transmit power is used for the PLCP preamble/header part, and then the selected lower power level is used for the rest of the frame transmission. Transmitting the PLCP preamble/header at the maximum power protects the entire frame due to the inherent characteristics of the CCA mechanism and the EIFS deference operation. SP-TPC does not need an external protection, i.e., RTS/CTS exchange, of the power-controlled frames, and hence is named Self-Protective TPC (SP-TPC).
The simulation results demonstrate that SP-TPC outperforms the other TPC strategies consistently in terms of both throughput and energy efficiency in the small and mid- ranged WLAN environments. As a future work, we will work on a mechanism to detect hidden stations since we understand that the RTS/CTS exchange should be used in the large-ranged WLAN environments irrespective of a TPC employment due to the inherent existence of hidden stations. With such a hidden station detection mechanism in place, we should be able to use the proposed SP-TPC and a typical RTS/CTS-aided TPC scheme in an adaptive manner
REFERENCES
[1] IEEE, “Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) specifications,” IEEE 802.11-1999, 1999.
[2] IEEE, “Supplement to Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) specifications: High-speed Physical Layer in the 5 GHz Band,” IEEE Std. 802.11a-1999, 1999.
[3] IEEE, “Supplement to Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) specifications: Higher Speed Physical Layer Extension in the 2.4 GHz Band,” IEEE Std. 802.11b-1999, 1999.
[4] Tajana Simunic, Luca Benini, Peter Glynn, and Giovanni De Micheli,
“Dynamic Power Management for Portable Systems,” in Proc. ACM MobiCom’00, Boston, MA, Aug. 2000, pp. 11–19.
[5] Manish Anand, Edmund B. Nightingale, and Jason Flinn, “Self-Tuning Wireless Network Power Management,” in Proc. ACM MobiCom’03, San Diego, CA, Sept. 2003.
[6] Daji Qiao, Sunghyun Choi, and Kang G. Shin, "Goodput Analysis and Link Adaptation for IEEE 802.11a Wireless LANs," IEEE Trans. on Mobile Computing (TMC), vol. 1, no. 4, pp. 278-292, Oct.-Dec. 2002.
[7] Daji Qiao, Sunghyun Choi, Amjad Soomro, and Kang G. Shin, "Energy- Efficient PCF Operation of IEEE 802.11a Wireless LAN," in Proc.
IEEE INFOCOM’02, New York, June 2002.
[8] Daji Qiao, Sunghyun Choi, Amit Jain, and Kang G. Shin, “MiSer: An Optimal Low-Energy Transmission Strategy for IEEE 802.11a/h," in Proc. ACM MobiCom'03, San Diego, CA, Sept. 2003.
[9] Javier del Prado and Sunghyun Choi, “Link Adaptation Strategy for IEEE 802.11 WLAN via Received Signal Strength Measurement,” in Proc. IEEE ICC’03, May 2003.
[10] Eun-Sun Jung and Nitin Vaidya, “A Power Control MAC Protocol for Ad Hoc Networks,” in Proc. ACM MOBICOM’02, September 2002.
[11] Jens Jelitto, Andre Noll Barreto, and Hong Linh Truong, “Power and Rate Adaptation in IEEE 802.11a Wireless LANs,” in Proc. IEEE VTC’03-Spring, Jeju, Korea, April 2003.
