Communications and Network, 2013, 5, 455-460 Published Online September 2013 (
Copyright © 2013 SciRes. CN
Schedule-Aware Power Management for Energy-Efficiency
Improvement in 802.11u WLAN*
Di Zhang1, Qinghe Du1,2, Pinyi Ren1, L i Sun1
1School of Electronic and Information Engineering, Xi’an Jiaotong University, Xi’an, China
2National Mobile Communications Research Laboratory of Southeast University, Nanjing, China
Received May 2013
Mobile stations supporting the 802.11u standard can access WLAN automatically when they are within the coverage of
the network service provided by this WLAN. To achieve this goal, the stations need to keep “on” states including idle
and active all the time. However, studies have noted that the idleness of stations often lead to considerable power con-
sumption. Although the conventional power saving mode (PSM) can provide energy saving effect to some extent, its
own disadvantage leads to lower energy efficiency when the number of stations accessing the target WLAN. In this
paper, we propose a Schedule-Aware P SM (S -PSM), which can improve the energy efficiency in 802.11u WLAN. Par-
ticularly, we use the Generic advertisement service (GAS) defined in 802.11u standard to broadcast the transmission
schedule information and all stations switch off their radios based on this information accordingly. We introduce the
Respond Contention Window to reduce the collision probability of competition channel. When there is no packet in the
access point (AP), AP broadcasts the GAS frame and actives th e Idle Timer. All stations will turn in to sleep and AP will
not send GAS frame until Idle Timer expires. Simulations have shown that our proposed scheme can significantly re-
duce power consumption compared with the conventional PSM.
Keywords: 802.11u Standard; Power Saving; Transmission Schedule; Generic Advertisement Service (GAS)
1. Introduction
The 802.11u standard [1] is an amendment of the IEEE
802.11 medium access control (MAC), which can help
users discover and select the network automatica lly. They
defined the Generic advertisement service (GAS) to pro-
vide functionality that enables stations to discover the
availability of information related to desired network ser-
vices. When a mobile station moves into a WLAN area,
the station can query and exchange information with other
external networks through GAS, and access the network
after and information authentication. Consequently, the
precondition of access is that the station’s wireless inter-
face keeps “on ” all the time. But this automatica lly access
may result in that mobile stations stay in idle state for a
long time. As mobile station is power ed by th e batter y, in
order to maximize the battery lifetime, we need an effec-
tive power management scheme for the stations to im-
prove energy utilization efficiency. 802.11u employs the
power saving mode (PSM), which was defined by IEEE
802.11 standard [2]. The PSM lets stations can spend time
in the sleep state for energy saving and switch to idle or
active state periodically to listen the information or re-
ceive the packets. This mechanism may solve the energy
wasting problem above, but the limitation of the PSM
scheme is that it doesn’t consider the impact of stations
number on energy efficiency. Because in the PSM, when
one station occupies the channel, others need to be in idle
state u ntil the chan nel free for th e next compet ition. There-
fore, most energy of each station are wasted in idle state.
In the last few years much research were dedicated to
improving the performance of the Power Saving Mode
[3-6]. In [3], the authors proposed a new AP-centric PSM
to let the AP chooses the best Beacon Interval (BI) and
Listen Interval (LI) for all clients based on their traffic
patterns. The authors of [4] proposed a dynamic wake up
period in which each client chooses its LI according to
the current round trip time of its TCP connection. The [5]
presents an analysis of the effect of intermittent connec-
tivity on minimizing energy consumption in PSM. Al-
The research reported in this paper was supported by the National
Natural Science Foundation of China under Grant No.
61102078, the
National Science and Technology Major Project under Grant No.
004, the Specialized Res earch Fund for the Doctor al
Program of Higher Education under Grant No. 20110201120014, the
Open Research Fund of National Mobile Communicat
ions Research
Laboratory, Southeast University (No.2011D10), and the Fundamental
Research Funds for the Central University.
Copyright © 2013 SciRes. CN
though these dynamic PSM compute the best system pa-
rameters for the clients in reducing energy consumption,
they do not consider that the time-driven scheme will re-
sult in that stations be in active when they know that they
have packets. In [6], the authors proposed a power-con-
serving algorithm, which dynamically switches off the
Network Interface Card of nodes when they are neither
transmitting nor receiving a packet in Mobile Ad-Hoc
Network. In [7], the authors proposed a protocol that using
a time-packet hybrid-driven method to dynamic switch-
off the wireless interface in ad-hoc network. These meas-
ures only focus on saving energy by enlarging the sleep
time and thus ignored the ineffective of the collision proba-
bility when completing the RTS/CTS dialogues. The higher
collision probability means more time used in retransmis-
sion which will surely waste energy and decrease the
system throughput. In this paper, under the 802.11u stan-
dard [2], we propose a new scheme called Schedule-
Aware Power S aving Mod e (S-PSM), which gets a better
energy utilization efficiency and reduces the collision
probability to some extent and gets an improvement in
system throughput.
The rest of this paper is organized as follows. Section
II presents the system model and the PSM in 802.11
standard. The design and implementation details of pro-
posed S-PSM scheme are given in Section III. Section IV
makes a performance analysis between these two schemes.
In Section V, we verify the two schemes’ energy saving
ability. Finally conclusions and future works are remarked
in Section VI.
2. System Model
We suppose that there are N wireless stations and an
access point (AP) in an infrastructure WLAN. Stations
communicate with the AP. We consider the downlink
traffic that AP buffers the incoming traffic packets. The
implementation of Power Saving Mode [8-10] allows the
station to be in one of the three power states: active, idle
or sleep. In the active or idle state, the station is fully
powered and is ready to receive the packets or listen to
the information at any time. During the sleep state, the
station is not able to transmit or receive for energy saving.
When a station works in PSM, it will listen to th e beacon
frame periodically. AP buffers the packets for the station
and announces the corresponding station every beacon
beginning via Traffic Indication Map (TIM) carried in
beacon frame, then the station sends PS-Poll frame to
retrieve the packets. Until all the packets are received, it
goes back to sleep.
Figure 1 illustrates an example of the power saving
mode in an IEEE 802.11 infrastructure network. In an
infrastructure WLAN, AP sends beacon frames to mes-
sage stations which have buffered packets at the begin-
ning of a beacon period, at th e same time, all the stations
Figure 1. Procedure of Conventional PSM scheme.
wake up and listen to the buffer information which is in
the TIM field. If the bitmap for one station is not set ( i.e.,
no buffered packe ts) in the TIM, the station goes back to
sleep immediately. Otherwise, the station sends a PS-Poll
frame to the AP by means of the standard distributed
coordinate function (DCF) procedure. Upon receiving a
PS-Poll, AP sends the destined packets and receives a
corresponding acknowledgement (ACK) frame from the
station. After this transmission, the station checks the
MORE DATA bit field in the beacon frame header to
know whether the received frame is the last one in this
beacon. If the field is set to zero, the frame is the last one
and therefore the station goes to sleep mode after receiv-
ing it. If the field is not zero, station sends another PS-
Poll frame to request the next packet until there is no
buffer packets during this beacon interval.
3. Design and Implementation of S-PSM
In this section, we proposed our Schedule-Aware Power
Saving Mode ( S-PSM). Compared with c onve ntional PSM,
the main difference of S-PSM is the message-driven me-
chanism and AP determines the transmission schedule for
stations. In the following, we will describe the S-PSM
scheme in detail.
3.1. Principle of the S-PSM Scheme
In our proposed S-PSM scheme, we let the stations switch
off their interface dynamically according to the schedule
information from AP. Stations may be in two operating
states: active state and sleep state. In the active state, they
receive the packets or listen to the information. During
the sleep state, the station is powered down for energy
saving. AP buffers the incoming packets for all stations
and informs them of buffer information. In our scheme,
we define one transmission process, during which every
destination station receives one packet successfully, called
Once Transmission Cycle. AP attaches the destination
station ID, transmission schedule and packets length to
the GAS frame, and broadcasts GAS frame to all stations.
If the information of station ID is not set for this station,
the station turns to sleep state immediately. The destination
stations stay in active state and AP determines the
trans m issi on s c hed ule by t he Fir st Com e F ir st Ser ve (FCFS)
Copyright © 2013 SciRes. CN
algorithm. Then AP announces a GAS frame to the active
stations making clock synchronization and notifies all
stations the tran smission schedule. These sta tions according
with the schedule switch off their interfaces when current
transmission is irrelevant to them. When there is no packet
for all the stations, AP attaches the sleep time in the
packet length field and sends to all the stations and opens
the Idle Timer. All stations will wake up and AP checks
the buffer when the Idle Timer expires.
3.2. Protocol Description of the S -PSM Scheme
In our scheme, AP attaches the transmission schedule
information in the GAS frame to inform stations buffer
information [11]. These information include destination
stations ID, transmission schedule, and packets length. In
our implementation, we add one extra information ele-
ment (using one of the reserved Element IDs) of the GAS
frame for this purpose. The format of the information
element is shown in Figure 2. We can use a field, which
is reserved in the GAS Initial Request/Respond frame
body format, to carry the relative information. Hence we
use 8 bits to contain the destination station ID and queue
length information respectively. The Table 1 is the de-
scription of S-PSM process.
The overall scheme of proposed S-PSM is shown in
Figure 3. In Figure 3, station S4 receives the GAS frame
and finds that there is no its packet in the buffer, then it
turns to sleep state immediately. S1, S2, S3 will be in ac-
tive state and enter the Respond Contention Window
(RCW). They send the GAS-Respond frames to inform
AP that they are ready for receiving the data by means of
DCF procedure. This basic access procedure (DCF) has
clearly specified in 802.11u standard. When one destina-
tion station delivers the respond information su ccessfully,
it will not attend the competition anymore and get out
Respond Contention Window. Finally, after all the des-
tination stations GAS -Respond frame have been received,
the Respond Contention Window closes. Then the trans-
mission process starts. AP determines the transmission
schedule under the First Come First Serve (FCFS) algo
Figure 2. Format of the GAS frame in S-PSM.
Figure 3. Procedure of S-PSM sch em e
Table 1. The Process of S-PSM.
S-PSM process
checks the buffer;
02. IF Current Buffer Size > 0
03. AP achieves the destination stations ID;
04. AP sends GAS frame;
05. Other stations turn into sleep state;
06. Respond Content ion Wi ndow opens;
07. IF Sj sends the frame successfully;
08. Sj quits the compete process;
09. END
10. AP receives all stations’ GAS -Respond;
11. Respond Contention Window closes;
12. AP determine the transmission sequence(FCFS);
13. AP send GAS fra me ;
14. Begin transmission;
15. After transmission, turn to 1;
16. ELSEIF Current Buffer Size = 0
17. AP broadcasts GAS frame;
18. Idle Timer opens;
19. All stations switch off the interfaces;
20. AP not check buffer anymore;
21. IF Idle Time expires
22. All the stations wake up at the set time;
23. AP checks the buffer
24. Turn to 1;
25. END
26. END
rithm. In Figure 3, the transmission schedule is (S1, S3,
and S2). Then AP sends GAS frame, and stations ac-
cording with the schedule and the packets length infor-
mation to estimate the sleeping time as well as to wake
up to receive the pack ets. For each transmiss ion , AP sends
a GAS frame to make the clock synchronization. When
this transmission cycle ends, all the stations be in active
state and waiting the next GAS frame.
4. Performance Analyses
4.1. Energy Saving Efficiency Analysis
The PSM provides that station who works in PSM must
observe some provisions. First, when one station has
packets buffered in AP, this station must be active until
the last one is received during the entire beacon cycle.
Second, the station sends frame to AP by means of the
standard distributed coordinate function (DCF) procedure
[12-14]. So, when one station occupies the channel, other
stations will be in idle state until the channel becomes
free. In order to analyze the idle time of stations conve-
niently, we fail to consider the back off time and colli-
sion situation and neglect some secondary factors lik e the
transmission delay time. Let Ti is the transmission time
(including the DCF inter-frame space (DIFS), frame
transmission time t, and ACK) for one station receiving
the packets. We can write Ti as .
We assume that there are N stations in the network. The
number of possible transmission schedules is . Let
Mk be the kth schedule (). represents
the total idle time of N station in schedule. Then,
Copyright © 2013 SciRes. CN
the expected idle time E[t] can be expressed as
[]( )
according to the analysis results of references [13], we
can get
1 11
1( 1)!
[]( )
( 1)
k ii
== =
= =
∑ ∑∑
So, we can clearly see that the total idle time of sta-
tions in one random schedule increase as the number of
accessed stations. While considering our S-PSM, the prob-
lem of energy wasting in idleness is solved by utilizing
the transmission schedule to driven stations’ states. All
the stations switch off their interfaces by the transmission
schedule and the packets length information. During one
transmission cycle, stations turn to sleep state when the
transmission is irrelative to them. Then, we can obtain
the total idle time of stations in the schedule, which is
determined by FCFS algorithm:
[] 0ET =
In addition, the sleep time of jth station is
Therefore, with the increase of number of stations,
each station has more time to sleep. Hence, we can con-
clude that our S-PSM scheme has a better energy saving
effect than PSM.
Because of the massage-driven method, if there is no
packet in the buffer, AP will check the buffer frequently,
and stations will receive empty GAS frame and have
nothing to do. After we introduce the Idle Timer, AP will
not detect the buffer and all stations can turn to sleep
state for energy sa ving. Let Tc be the time of once trans-
mission cycle, then we can get
( 2)
= +++
where BOi is the back off time of ith stations, GAS is the
GAS frame space. The inter packet arrival time Tinter
takes on exponential distribu tions. We can get the proba-
bility of Tinter > Tc
( )1
1( |)
tT T
p TTedt
= −−=
We consider the situation that there is no packets ar-
rive during this transmission time Tc. So, the probability
of this situation can be regarded as
() c
zero interzero
= >=
Where the Pzero represents the probability of that AP
transits all the buffered packets in a transmission. In or-
der to observe the effect of stations number N on the
probability of this situation, we consider the average
transmission time and back off time
() (21)
cavg avg
Then, we can wri t e
[ ()(2 1)]
avg avg
PP e
−+ ++
We can see that with the increase of N, the value of P
will have a great decrease. So, this situation will not hap-
pen when more stations access the WLAN. Therefore, it
is necessary to introduce the Idle Timer for energy saving
when there are little stations in the WLAN. In the con-
sequently simulations, we can see the impact of Idle Ti-
mer on the S-PSM scheme.
4.2. Collision Probability Analysis
We know that the basic 802.11 MAC protocol Distri-
buted Coordination Function (DCF) is based on the Car-
rier Sense Multiple Access (CSMA). When more than
two stations whose back off time is zero detect the chan-
nel as free at the same time, a collision occurs. The initial
congestion window for all stations is CWmin, and the
back off time is BT (
0BT CW<<
). The probability of
BT is
1/p CW=
. In the conventional PSM, the prob-
ability of ith station access the channel successfully is
psmj j
=−≥≥ −+
where M is the number of stations who involve in the
channel competition.
In our scheme, we introduce the respond contention
window (RCW) for stations competing the channel. Dur-
ing the window, when a station delivers the respond frame
successfully, it will not attend the competition any more.
The probability of ith station access the channel success-
fully is
Ni jj
i Ni
P Cp
Because of the 0 < p < 1, we know that
() 1
M Ni
i iMNi
M Ni
jjj jj
jNi j
M Ni
= =
= −+=
−= −
+−> −+
= −+
Copyright © 2013 SciRes. CN
Therefore, we can get
From the Equation (13), we can see the respond Con-
tention window can reduce the probability of collision.
5. Simulation Evaluation
In this part, we have implemented the proposed S-PSM
scheme and make a comparison with the PSM. The si-
mulation environment and parameters are in Table 2. In
addition, stations need to switch-off interface frequently,
therefore we must consider the energy consumption of
switch off which is twice of active state. In the PSM si-
mulation process, we set the beacon interval (BI) and the
listen interval (LI) is 100 milliseconds. Besides that, the
minimal congestion window for all stations is 32 slots.
We design and implement the simulator based on MAT-
LAB. The set of energy consumption for each state is
reference d [5 ,15].
Figures 4 and 5 show the comparison results between
S-PSM and PSM. We can clearly find that our scheme
can overcome the defect of PSM that stations waste ener-
gy in idle state. As shown in the Figure 4, the time being
in idle state under PSM has a conspicuous increase as the
number of stations increase, and the resident time of
Table 2. Table type styles (Table capt ion is indispensable).
Simulation Parameters Values
Number of clients
Data Transmission rate
Packet size
Beacon interval
PS-Poll/ACK frame size
average inter-frame arrival time
Slot Time
Sleeping power
Active/Idle power
AP buffer size
2 to 14
11 Mbps
1500 bytes
100 ms
14 bytes
10 ms
20*106 s
10*106 s
50*106 s
50 mW
750 mW
2 GB
Figure 4. Comparison of Average Resident Time of each
state between PSM an d S-PSM.
sleep state has a significant drop. On the other hand, the
S-PSM scheme gets longer resident time of sleep state. In
our scheme, stations utilize the transmission sequence to
dynamically switch-off their interface when they have no
data to receive, consequently, stations will be in sleep
state to save energy. For the competition time, S-PSM is
less than PSM. In S-PSM, all stations only take part in
competition during the respond contention window, which
will effectively reduce the competition time. Figure 5
plots the throughput and energy consumption for one sta-
tion during this simulation process. We can see the S-
PSM has a preferable performance in throug hput and the
energy consumption has a significant downward. With
increase of station number, the average throughput is big-
ger than PSM. This is because that the collision probabil-
ity goes down by introduction the Respond Contention
Window. So, more time will be utilized to transfer the
packets. The energy consumption of S-PSM will be around
3Joule, which is about 80% lower than that of PSM. This
is in agreement with theoretical analysis that station in
S-PSM will has more time to sleep for energy saving.
Figure 6 shows the influence of Idle Timer on the
Figure 5. Comparison of Throughput and Energy Con-
sumption between PSM and S-PSM.
Figure 6. Impact of Idle Timer for S-PSM.
2 4 6810 1214
The Number of St ations
The average time for station at each state (s)
S-PSM sleep state
PSM sleep state
S -PSM com petition state
PSM competition state
PSM idle state
05 10 15
T he Number of Stati ons
Average Throughout for Each Station (Mbps)
S-PSM scheme
PS M scheme
05 10 15
T he Number of Stati ons
Average Energy Consumption for Each Station(J)
S-PSM scheme
PSM scheme
05 1015
The Number of St ations
Average Time for Station at Each State (s)
S-PSM sle e p time
S-PSM(without Idle
Timer) idle tim e
S-PSM(without Idle
Timer) slee p time
0510 15
The Number of St ations
Average Energy Consumption for one Station (J)
(without Idle Timer)
Copyright © 2013 SciRes. CN
performance of S-PSM. First, let’s observe the perfor-
mance of S-PSM without Idle Timer. From the simula-
tion result, we can see the station will spend some time in
idleness when M < 8. And the corresponding energy
consumption is very large. When AP has no buffered
data, AP will check the buffer frequently, and stations
will receive empty GAS frame and have nothing to do.
When the number of stations is little or network load is
low, this issue will appear frequently and cause the most
energy wasting, which is consistent with the performance
analysis of Equation (9). We introduce the Idle Timer
whose value is set as 50 slots to solve this problem.
When there are no packets buffered in AP, AP sends
GAS frame to tell all the stations turn to sleep state and
opens the idle timer. After the timer expires, all stations
switch to active state and AP checks buffer information
to begin the next transmission. We can see the energy
consumption remaining about 3Joule, which is a huge
energy saving improvement.
We have verified the two mechanism’s validity and
analyze their performances from three aspects (resident
time, system throughput, and energy consumption). We
observed that our proposed scheme has a better perfor-
mance in energy saving. By introducing the Idle Timer
and using message-driven scheme to let stations know
the transmission schedule, S-PSM overcomes the short-
age of the PSM and obtains a significant improvement in
terms of energy saving.
6. Conclusions
We have proposed the S-PSM that increases energy effi-
ciency of all wireless clients in an infrastructure network.
The AP in S-PSM determines the transmission sequence
and stations change their own states by the information
attached in GAS frames. AP plays a centralized role that
controls the transmission process. The stations convert
the state according to the transmission sequence which
can reduce unnecessary wake ups and maximize energy
saving. The Respond Contention Window (RCW) could
reduce collision probability effectively, which is helpful
to improve the system throughput, and the transmission
cycle assures fairness among stations. However, how the
AP communicates with sleep stations through broadcast
needs to be paid more attentions. We utilize the Idle Ti-
mer to solve this problem in our model, but cause unne-
cessary packets delay. Our needs to switch the interface
frequently, and that does harm to the interfaces. So, in
future work, we can improve S-PSM performance via
these aspects.
[1] IEEE 802.11u, Part 11, Wireless LAN Medium Access
Control and Physical Layer Specifications, Amendment 9:
Interworking with External Networks,” 2011.
[2] IEEE 802.11, Part 11, “Wireless LAN Medium Access
Control and Physical Layer Specifications,” 1999.
[3] Yi X., X. Luo and R. K. C. Chang, “Centralized PSM: An
AP-Centric Power Saving Mode for 802.11 Infrastructure
Networks,” IEEE Sarnoff Symposium, SARNOFF’09, 30
March 2009-1 April 2009, pp. 1-5.
[4] S. Nath, Z. Anderson and S. Seshan. “Choosing Beacon
Periods to Improve Response Times for Wireless HTTP
Clients,” Proceedings of MobiWac, 2004.
[5] R. Krashinsky and H. Balakrishnan. “Minimizing Energy
for Wireless Web Access with Bounded Slowdown,”
Wireless Networks, Vol. 11, 2005.
[6] J.-C. Cano and P. Manzoni, “Evaluating the Energy-
Consumption Reduction in a MANET by Dynamically
Switching-Off Network Interfaces,” Proceedings of the
6th IEEE symposium on Computers and Communications,
2001, pp. 186-191.
[7] N. Li, Y. Xu and S. Xie, “A Modified Version of IEEE
802.11 Power-Saving Protocol,” Journal of System Simu-
lation, Vol. 17, No. 1, 2005.
[8] D. Ning, A. Pathak, D. Koutsonikolas, C. Shepard, Y. C.
Hu and Z. Lin, “Realizing the Full Potential of PSM Us-
ing Proxying,” INFOCOM, 2012 Proceedings IEEE,
25-30 March 2012, pp. 2821-2825.
[9] D. Qiao, “Smart Power-Saving Mode for IEEE 802.11
Wireless LANs,” Proceedings of IEEE Infocom’05,
March 2005.
[10] H. Lin, S. Huang and R. Jan, “A Power-Saving Schedul-
ing for Infrastructure Mode 802.11 Wireless LANs,”
Computer Communications, Vol. 29, 2006, pp. 3483-
[11] X. Chen, S. G. Jin and D. J. Qiao, “M-PSM: Mobility-
Aware Power Save Mode for IEEE 802.11 WLANs,”
Distributed Computing Systems (ICDCS), 2011 31st In-
ternational Conference on, 20-24 June 2011, pp. 77-86.
[12] S. Baek and B. D. Choi, “Performance Analysis of Power
Save Mode in IEEE 802.11 Infrastructure WLAN,” ICT
2008 IEEE International Conference on Telecommunica-
tions, Piscat away, 16-19 June 2008, pp. 1-4.
[13] S. Pack, H. Park, S. Min and I. Jang, “Energy Efficiency
Analysis of IEEE 802.11 PSM in Multi-Rate Environ-
ments,” 2012 IEEE Consumer Communications and
Networking Conference (CCNC), 14-17 Jan uary 2012, pp.
[14] Y. He, R. Yuan, X. Ma, J. Li and C. Wang, “Scheduled
PSM for Minimizing Energy in Wireless LANs,” Pro-
ceeings of IEEE ICNP, 2007.
[15] S. Jin, K. Han and S. Choi, “Idle Mode for Deep Power
Save in IEEE 802.11 WLANs,” Journal of Communica-
tions and Networks, Vol. 12, No. 5, 2010, pp. 480-491.