Académique Documents
Professionnel Documents
Culture Documents
4, DECEMBER 2012
473
AbstractIt is crucial to save power and prolong the runtime of mobile stations (STAs) in wireless local area networks
(WLANs). In an infrastructure WLAN, a STA cannot be connected until it is associated with an access point (AP), which
is responsible for buffering frames for all the associated STAs
operating in the power saving mode. Hence, efficient memory
utilization is critical for an AP to accommodate as many powersaving STAs as possible. The basic power management (BPM)
scheme introduced in the IEEE 802.11 standard achieves power
saving by allowing STAs not engaging in data delivery to operate
in doze mode, but it does not consider the efficient use of
the memory in the AP. To tradeoff power consumption for
memory usage, we present an AP-priority timer-based power
management (APP-TPM) scheme and develop a novel model for
stochastic analysis of the proposed scheme. Based on this model,
the probability distributions of the numbers of frames buffered
at the AP and the average numbers of frames buffered at the
AP are obtained. Moreover, a power-aware buffer management
scheme (PBMS), which is based on the derived statistics, is
proposed to accommodate as many STAs as possible given a
fixed amount of memory in the AP while maintaining low power
consumption. Simulation results show that the proposed scheme
performs better than BPM in terms of memory usage in the AP.
Index TermsPower management, WLAN, IEEE 802.11,
power saving.
I. I NTRODUCTION
c 2012 IEEE
1932-4537/12/$31.00
474
IEEE TRANSACTIONS ON NETWORK AND SERVICE MANAGEMENT, VOL. 9, NO. 4, DECEMBER 2012
Fig. 1.
ZHU et al.: ACCESS POINT BUFFER MANAGEMENT FOR POWER SAVING IN IEEE 802.11 WLANS
TABLE I
ACRONYMS AND NOTATIONS
Definition
AP
Access Point
AR
Association Request
AID
Association ID
APP-TPM
LI
Listen Interval
BI
Beacon Interval
BPM
NBF-AP
ANBF-AP
PS-poll
Power-Saving-poll
PFD
TPM
TIM
TI
TD
MIT
PIT
475
(2)
(n)
n
k=1
(1)
w.r.t. x
s.t.
n
k=1
(2)
,x
x(k)
1
E (k)
(1)
(2)
(n)
; TI , TI , ..., TI
N
(i)
LAP ;
i=n+1
(n)
, ..., x
(k)
x(k) LAP
476
Fig. 2.
IEEE TRANSACTIONS ON NETWORK AND SERVICE MANAGEMENT, VOL. 9, NO. 4, DECEMBER 2012
Flowchart of APP-TPM-AP.
Fig. 3.
Flowchart of APP-TPM-STA.
At any time, each STA can operate in transmission, reception, idle, or doze mode. As in [2], we combine transmission
and reception into one state, called active state.
In APP-TPM, the doze period of the k-th STA is deter(k)
ministic, i.e., it is a constant TD set to LI that is equal to a
multiple of BIs [1]. In order to reduce the difficulty of deriving
the statistics by stochastic analysis, we first consider the doze
period as a random variable (assumption (i) below) and then
(k)
let the variable take a deterministic value TD to derive the
statistics for the APP-TPM.
A. Assumptions
(k)
(x)
ZHU et al.: ACCESS POINT BUFFER MANAGEMENT FOR POWER SAVING IN IEEE 802.11 WLANS
(k)
(k)
(k)
(2)
(k)
t
(k)
(k)
(k)
Besides, we use PI , PA , and PD to represent the probability that the k-th STA is in the idle state, active state,
(k)
(k)
= PI , PA =
and doze state, respectively. Thus, PI
(k)
i=0
j=0 qAi,j and PD =
i=0
j=0 PDi,j . Further, we
introduce z-transforms:
Qi (z)
qAi,j z j (i = 0, 1, ...; 0 < z < 1) ;
(k)
(k)
(3)
From (3), we have PDi,j (t) = 0 pDi,j (t, x)dx. Let PDi,j
lim PDi,j (t), qAi,j lim qAi,j (t), and PI lim PI (t).
j=0
Pr{x D < x + t}
Pr{D x}
477
K(y, z)
i=0
Let I be the idle time interval from the instant when the
STA becomes idle to the instant when an incoming frame
arrives at the AP or an outgoing frame is generated at the STA.
Thus, from assumption (iii), I is exponentially distributed
(k)
(k)
(k)
with rate 1,2 [6], i.e., the pdf and CDF of I are 1,2 e1,2 x
(k)
(k)
(k)
P r{I +t TI |I < TI } =
(k)
(k)
1,2 e1,2 TI t
(k)
(k)
1 e1,2 TI
+o(t)
(4)
where o(t) is an infinitesimal satisfying lim o(t)/t =
t0
0.
Considering the state transitions in Fig. 4, we obtain the
following set of equations (Refer to Appendix I for the detailed
derivations).
PI (t) =
(k)
1,2
1e
(k)
(k)
1,2 TI
(5)
(k)
qA
(t) = qA0,0 (t)((k) + 1,2 ) + (k) qA0,1 (t)
0,0
(k)
pD0,1 (t, x) (k) (x)dx
+ qA1,0 (t) +
0
(k)
+
pD1,0 (t, x) (k) (x)dx + 1,2 PI (t)
(6)
C. Model of APP-TPM
Clearly, (k) , defined in assumption (iii), is the rate of
transmitting/receiving frames between the AP and the STA,
i.e., the average number of frames transmitted/received per
unit time. From assumption (ii), the combined rate of incoming
(k)
and outgoing frames arriving at the AP and the STA is 1,2 .
(k)
Hence, we assume 1,2 < (k) , i.e., (k) < 1 to prevent
the number of frames to be transmitted from growing without
bound.
Let (t) be the state of the STA at time t, and Y (t) be the
accumulated doze time (ADT) at time t, which is the time
period from the instant when the STA starts dozing to time
t. Use pDi,j (t, x) to represent the probability density of the
(k)
(t) = qA0,j (t)((k) + 1,2 ) + (k) qA0,j+1 (t)
qA
0,j
(k)
pD0,j+1 (t, x) (k) (x)dx,
+ 1 qA0,j1 (t) +
j = 1, 2, ...
(7)
(k)
(k)
qA
(t) = qAi,0 (t)((k) + 1,2 ) + 2 qAi1,0 (t)
i,0
(8)
478
IEEE TRANSACTIONS ON NETWORK AND SERVICE MANAGEMENT, VOL. 9, NO. 4, DECEMBER 2012
Fig. 4.
(k)
(k)
qA
(t) = qAi,j (t)((k) + 1,2 ) + 2 qAi1,j (t)
i,j
(k)
+ qAi,j+1 (t) +
pDi,j+1 (t, x) (k) (x)dx,
i, j = 1, 2, ...
(9)
(k)
(k)
pD0,j (t, x) +
pD (t, x)
t
x 0,j
(k)
(k)
= [1,2 + (k) (x)]pD0,j (t, x) + 1 pD0,j1 (t, x),
(k)
1,2 e1,2 TI
(12)
(13)
pD (t, x) +
pD (t, x)
t i,0
x i,0
(k)
(k)
= [1,2 + (k) (x)]pDi,0 (t, x) + 2 pDi1,0 (t, x),
pD0,0 (t, x) +
pD (t, x)
t
x 0,0
(k)
= [1,2 + (k) (x)]pD0,0 (t, x), t > 0, x > 0
(11)
(14)
ZHU et al.: ACCESS POINT BUFFER MANAGEMENT FOR POWER SAVING IN IEEE 802.11 WLANS
pDi,j (t, x) +
pD (t, x)
t
x i,j
(k)
(k)
= [1,2 + (k) (x)]pDi,j (t, x) + 1 pDi,j1 (t, x)
+
(k)
2 pDi1,j (t, x),
(k)
(15)
[1 z0 + z0 e
(16)
D. Statistics of APP-TPM
From (4)-(16), we obtain the following equations (Refer to
Appendix II for the detailed derivations).
(k)
yzy
1
]
[1
(k) (k)
1,2 TI
y zy
z0 (1 e
)
(k)
(k)
2 y(zy )2 (z y)
zy [(1 + (k) )z 1]
Q
(z
)
+
Q0 (z)
0
y
z zy
(k) (y zy )(z zy )
(k)
(k)
(k) e1,2 TI
(k)
(k)
zy ye(1,2 1 z0 )TD
+ PI
{
(k) (k)
(k) (k)
z0 (y zy )
[1 e1,2 TI ][1 e1,2 TD ]
(k)
(k)
(k)
(1,2 1 zy )TD
zy (y z)e
(z zy )(y zy )
(k) (k)
1,2 TD
zy e
(k)
(k)
(k)
(1,2 1 z)TD
z zy
z0
1
(k)
21
(k)
(k)
+1,2
(k)
(k)
(k)
(k)
(23)
}.
(k)
(k)
(k)
E (k) = PI EI + PA EA + PD ED
(24)
(k)
P AP (j) =
PDi,j
i=0
(k) (k)
(k) (k) j
(1 TD )s
(k) (k)
PI 1,2 e1,2 TI [1 e1 TD
]
s=0
s!
(k) (k)
(k) (k)
(k)
1,2 TI
1,2 TD
1 (1 e
)(1 e
)
(k)
(k)
LAP =
(k)
jP AP (j)
j=1
}
(17)
where
(k)
(25)
zy (z y)e
zy ye
+
z0 (y zy )
(y zy )(z zy )
zy e
(k)
{(1 z0 ) + e1,2 TI
(k)
2 z0
(k)
(k)
(k)
(k) (k)
(1,2 1 z0 )TD
2 TD
z zy
(k)
(k)
(k)
(k)
(1,2 1 zy 2 y)TD
(k)
(k)
(k)
(k)
(1,2 1 z2 y)TD
(k)
1,2
K(y, z) = PI
(k)
PA = K(1, 1) = PI
479
(k)
(k)
((k) + 1,2 )2 41 (k) ) (18)
(k)
PI
(k)
(k)
(k)
(k)
(26)
(k)
(k)
(k)
(k)
zy
(19) next beacon comes [1]. We refer to the time period from the
(k)
(k)
(k) + 1,2 2 y
instant when the STA becomes idle to the instant when the
next beacon arrives as the mandatory idle time (MIT). As
(k) (k)
(k)
(k)
(k)
(k)
Q0 (z) = PI 1,2 {(z0 z)[1 e1,2 TD + e1,2 (TI +TD ) ] mentioned in Section I, the main difference between BPM and
(k) (k)
(k)
(k)
(k)
(k)
(k)
(k)
APP-TPM is that, when a STA becomes idle, BPM allows the
+ e1,2 TI [ze(1,2 1 z0 )TD z0 e(1,2 1 z)TD ]}
STA to operate in doze mode after MIT expires whereas APP(k) (k)
(k) (k)
TPM requires the STA to stay in the idle state for a period
/{z0 (1 e1,2 TI )(1 e1,2 TD )
of TI , which is equal to MIT plus a period of time called the
(k) 2
(k)
(k)
(k)
[1 z ( + 1,2 )z + ]}
prolonged idle time (PIT) that is equal to a multiple of BIs.
(20)
Obviously, APP-TPM is reduced to BPM if PIT is set to 0.
and
As MIT is a random variable uniformly distributed in [0, BI],
(k)
the average MIT is 0.5 BI. Thus, TI = (0.5 + m)BI where
(k)
(k) (k)
T
1,2 (1 z0 )
1,2 e 1,2 I
(k)
m
is a positive integer [2]. We fix BI=0.1 s [1]. Additionally,
PI = {1 +
+
(k) (k)
(k) (k)
(k)
(k)
2 z0
(1 e1,2 TI )(1 e1,2 TD ) for the k-th STA, we introduce the notation R(k) (k)
1 /2 .
(k)
(k)
(k)
(k)
(k)
(k)
the STA is heavier than
1 z0 + 2 z0 TD + z0 e2 TD e(1,2 1 z0 )TD 1 Thus, download (or upload) traffic of(k)
(k)
}
upload
(or
download)
traffic
when
R
(k)
1 (or R 1).
2 z0
In the IEEE 802.11 standard, a two-octet field in the AR
(21)
frame is used for LI [1], which indicates that LI ranges from 0
(k)
Moreover,
to 216 1 BI. Thus, TD takes a value over [0, 6553.5] s when
(k) (k)
BI =0.1 s. As in [4], we set the power consumptions of the
(k) (k) (k)
PI 1,2 TD e1,2 TI
(k)
PD =
(22) STA in active, idle, and doze states to EA =1W, EI =0.83W,
(k) (k)
(k) (k)
1,2 TI
1,2 TD
ED =0.13W, respectively. Obviously, power consumption of
[1 e
][1 e
]
(k)
480
IEEE TRANSACTIONS ON NETWORK AND SERVICE MANAGEMENT, VOL. 9, NO. 4, DECEMBER 2012
TABLE II
SIMULATION PARAMETERS
Parameter
Value
144 b
48 b
BI (Beacon interval)
0.1 s
Data rate
11 Mbps
1300 B
34 B
20 us
10 us
50 us
31
1023
N (Number of STAs)
20
100 s
Fig. 5.
maximum data rate to more than 500 Mbps and increases the
frame transmission rate accordingly. Therefore, in the numeric
evaluations, we select =2000 fps [2] as the nominal value.
Intuitively, increases in transmitting/receiving rate should
reduce the ANBF-AP. However, this turns out to be incorrect.
The experiments, in which is set to 1000, 1500, 2000, ...,
(k)
(k)
and 5000, respectively, indicate that, for given 1 and 2 ,
increases in only exert a small influence on the ANBFAP (we omit the figure of this experiment due to space
limitation). The main reason is that, increasing only makes
the STA finish transmitting/receiving frames and enter doze
state sooner, but the ANBF-AP depends on the arrival rate of
the incoming frames. Hence, increasing does not help much
in reducing the ANBF-AP except that it increases the power
consumption. Hence, the observations from the experiments
with =2000 can be applied to other cases in which a different
is chosen.
(k)
Next, we investigate the impact of parameters R and
on the ANBF-AP of the k-th STA, given in (26). Fixing m=1,
(k)
(k)
(k)
i.e., TI =1.5 BI=0.15 s, setting TD =30 s, 2 =10, 20, ...,
(k)
50 fps and R =1, 2, ..., 16, we obtain Fig. 6. In addition,
(k)
setting R to 1, 1/2, ..., and 1/16 leads to Fig. 7.
From Fig. 6, we observe that in the case where the download
(k)
traffic is heavier than the upload traffic, for a given 2 (e.g.,
(k)
(k)
2 =10 fps), ANBF-AP first goes up and then down as R
(k)
(k)
increases (or 1 increases due to 2 fixed). The reason
is that, for a given doze period, more frames arrive at the
(k)
AP when 1 is increased, which causes ANBF-AP to grow.
(k)
But, when 1 continues to increase, the idle time of the
STA decreases, which causes the STA to have less probability
of entering doze state, i.e., the doze period of the STA is
shortened, making ANBF-AP decrease. The same observation
can be obtained from Fig. 7 when upload traffic is heavier
than download traffic. The above observations imply that, the
AP does not need to enlarge its buffer size if the number of
ARs of STAs increases (i.e., if more STAs intend to connect to
(k)
(k)
the WLAN), which contributes to make 1 as well as R
increase.
ZHU et al.: ACCESS POINT BUFFER MANAGEMENT FOR POWER SAVING IN IEEE 802.11 WLANS
(k)
(k)
Fig. 6.
Impacts of R
Fig. 7.
Impacts of R
(k)
1.
(k)
1.
Fig. 8.
Fig. 9.
481
482
Fig. 10.
(2)
IEEE TRANSACTIONS ON NETWORK AND SERVICE MANAGEMENT, VOL. 9, NO. 4, DECEMBER 2012
Fig. 11. The PDF in BPM and the ratio of power consumption between
APP-TPM and BPM.
(3)
Max x(1)
(1)
(2)
(3)
(27)
(2)
(3)
TI , TI , TI
2000
Via genetic algorithm, we obtain the optimal power consumption of the above PBMS-OPT, which is shown in the
upper part in Fig. 10, where R = i represents the case
(1)
(2)
(3)
when R = R = R = i(i = 1, 2, 3). Moreover, the
(1)
(2)
(3)
corresponding best TI , TI and TI for R =3 are depicted
in the lower part in the figure.
From Fig. 10, we observe that, as the buffer size constraint
is gradually relaxed, power consumption can be gradu(1)
(2)
(3)
ally decreased by finding the optimal TI , TI and TI .
For example, when =200, for the case R =3, by setting
(1)
(2)
(3)
TI = TI =0.25 and TI =0.45 (see the lower part of the
figure), the power consumption is about 2400 mW (see the
upper part of the figure), which can be reduced to about 1700
(1)
(2)
(3)
mW by choosing TI =0.05 and TI =TI =0.25 when =600.
This supports the key idea of our APP-TPM that trades power
consumption for memory usage.
As mentioned above, APP-TPM is able to trade off more
power consumption for less frame discarding by adjusting
the pair of parameters TI and TD , whereas BPM does not
have this flexibility. To illustrate this tradeoff, we evaluate
the Percentage of Frame Discarded (PFD) by the AP under
BPM due to insufficient memory. Let D1 represent ANBFAP under BPM. Since the AP buffer size is , D1 is
the average number of the frames discarded by the AP due to
buffer overflow. Hence, PFD = max{(D1 )/D1 , 0}100%
since PFD must be non-negative. Additionally, we evaluate the
ratio of the power consumption with APP-TPM to that with
BPM.
Fig. 11 illustrates the PFD in BPM (see the upper part of the
figure) and the ratio of power consumption between APP-TPM
and BPM (see the lower part of the figure) when =200, 400,
..., 1200. From Fig. 11, we observe that: 1) PFD decreases as
grows and no frame discarding occurs when is sufficient for
buffering frames (e.g., =1000 in the case when R =1), which
agrees with our intuitions; and 2) when memory shortage
exists in the AP (e.g., <1000), APP-TPM consumes more
energy than BPM (i.e., the ratio of power consumption is
larger than 1), but APP-TPM and BMP consume about the
same power when the AP has enough memory to buffer frames
(e.g., =1000 in the case when R =1). In the case that R =2,
with tight memory constraint, e.g., =200, with BPM nearly
90 % of frames are discarded, while APP-TPM can reduce
frame discarding by consuming nearly 5.5 times more power
than BPM. However, when a larger memory, e.g., =1200 or
more, is available, power consumption of APP-TPM reduces
to less than twice the power consumption of BPM, which still
drops about 35% of the frames.
(k)
ZHU et al.: ACCESS POINT BUFFER MANAGEMENT FOR POWER SAVING IN IEEE 802.11 WLANS
(k)
(28)
j=+1
483
484
IEEE TRANSACTIONS ON NETWORK AND SERVICE MANAGEMENT, VOL. 9, NO. 4, DECEMBER 2012
Additionally, we have
G(x) = Pr{D x} =
0, x < TD ;
1, x TD
(a.2)
pDi,j (s, x) =
(a.3)
ZHU et al.: ACCESS POINT BUFFER MANAGEMENT FOR POWER SAVING IN IEEE 802.11 WLANS
Let lim spD0,0 (s, 0). Then (a.1) and (a.3) leads to
s0
K(y, z) =
s0
(a.4)
i j
= 2 1 (TD )i+j e1,2 TD (i, j = 0, 1, ...)
i!j!
+ 1 TD e
+ 2 TD e
(a.5)
=0
(a.6)
(a.7)
1,2 e
(1 e1,2 TI )(1 e1,2 TD )
(a.10)
(a.12)
/[1 z ( + 1,2 )z + ]
Obviously, the denominator of (a.12) has a unique root z0 in
the interval (0, 1), which is shown in (18). Clearly, z-transform
Q0 (z) is an analytic function over (0, 1). Thus, the numerator
of (a.12) has to be 0 when z = z0 , which yields
1
{[( + 1,2 )z0 ]qA0,0
z0
+ e1,2 TD (e1 TD z0 1 TD z0 1)}
(a.13)
(2 TD ) 1 TD z
(e
1 TD z 1)
z
i!
(a.14)
( + 1,2 )(Qi (z) qAi,0 )
z
+ 2 (Qi1 (z) qAi1,0 ) qAi,1 = 0, i = 1, 2, ...
y i qAi,0 = qA0,0 +
zy
{yqA1,0 + 2 y 2 Q0 (zy )}
(y zy )
1,2 TI
qA0,1 =
i=0
i=0
i
1,2 TD 2 TD y
z
y qAi,1 + e
(e
1)
i=1
(e1 TD z 1 TD z 1)}
i=1
485
(a.18)
(1 x)j 1 x
e
P AP (j) =
PDi,j =
dx
j!
0
i=0
which yields (25) through using (a.10) and integration by
parts. Moreover, the preceding equation also yields (26) by
using Talor series of ex as follows:
LAP =
jP AP (j) =
i=0
1 (TD )2
2
486
IEEE TRANSACTIONS ON NETWORK AND SERVICE MANAGEMENT, VOL. 9, NO. 4, DECEMBER 2012
[3] M. S. Gast, 802.11 Wireless Networks: The Definitive Guide, 2nd edition.
OReilly, 2005.
[4] R. Zheng, J. C. Hou, and L. Sha, Performance analysis of power management policies in wireless networks, IEEE Trans. Wireless Commun.,
vol. 5, pp. 13511361, June 2006.
[5] S. G. Sitharaman, Modeling queues using Poisson approximation in
IEEE 802.11 ad-hoc networks, in Proc. 2005 IEEE Workshop on Local
and Metropolitan Area Networks, pp. 16.
[6] S. M. Ross, Introduction to Probability Models, 9th edition. Elsevier
(Singapore) Pte LTD, 2007.
[7] Y. Rong, A. Y. Teymorian, L. Ma, X. Cheng, and H.-A. Choi, A novel
adaptation scheme for 802.11 Networks, IEEE Trans. Wireless Commun.,
vol. 8, no. 2, pp. 862870, Feb. 2009.
[8] N. Aste, L. Atzori, and L.S. Randaccio, WLC24-2: selective power
management in IEEE 802.11 infrastructure WLANs, in Proc. 2006
GLOBECOM, pp. 16.
[9] C.-H. Gan and Y.-B. Lin, An effective power conservation scheme for
IEEE 802.11 wireless networks, IEEE Trans. Veh. Technol., vol. 58, no.
4, pp. 19201929, May 2009.
[10] Y. He, R. Yuan, X. Ma, J. Li, and C. Wang, Scheduled PSM for
minimizing energy in wireless LANs, in Proc. 2007 IEEE International
Conference on Network-Protocols, pp. 154163.
[11] Y. Xie, X. Luo, and R. K. C. Chang, Centralized PSM: an AP-centric
power saving mode for 802.11 infrastructure networks, in Proc. 2009
IEEE Sarnoff Symposium, pp. 15.
[12] M. Sarkar and R. L. Cruz, A MAC layer dynamic power management
scheme for multiple users in a WLAN, in Proc. 2007 IEEE Wireless
Communications and Networking Conference, pp. 21262130.
[13] X. Perez-Costa and D. Camps-Mur, APSM: bounding the downlink
delay for 802.11 power save mode, in Proc. 2005 IEEE International
Conference on Communications, vol. 5, pp. 36163622.
[14] S.-C. Lo, G. Lee, and W.-T. Chen, An efficient multipolling mechanism
for IEEE 802.11 wireless LANs, IEEE Trans. Computers, vol. 52, no.
6, pp. 764778, 2003
[15] J.-R. Hsieh, T.-H. Lee, and Y.-W. Kuo, Energy-efficient multi-polling
scheme for wireless LANs, IEEE Trans. Wireless Commun., vol. 8, no.
3, pp. 15321541, Mar. 2009.
[16] T. Li, D. Leith, and D. Malone, Buffer sizing for 802.11-based
networks, IEEE/ACM Trans. Netw., vol. 19, no. 1, pp. 156169, 2011.
[17] Y. He, R. Yuan, and W. Gong, Modeling power saving protocols
for multicast services in 802.11 wireless LANs, IEEE Trans. Mobile
Computing, vol. 9, no. 5, pp. 657671, 2010.
[18] J. H. Holland, Adaptation in Natural and Artificial Systems: An Introductory Analysis with Applications to Biology, Control, and Artificial
Intelligence. The MIT Press/Bradford Books edition, 1992.
Han-Cheng Lu received his B.S. degree in information science and technology from Nangjing University of Posts and Telecommunications in 2008. He
is currently a candidate student for his master degree in computer science.
Victor C. M. Leung (S75-M89-SM97-F03) received the B.A.Sc. (Hons.) degree in electrical engineering from the University of British Columbia
(U.B.C.) in 1977, and was awarded the APEBC
Gold Medal as the head of the graduating class
in the Faculty of Applied Science. He attended
graduate school at U.B.C. on a Natural Sciences and
Engineering Research Council Postgraduate Scholarship and completed the Ph.D. degree in electrical
engineering in 1981.
From 1981 to 1987, Dr. Leung was a Senior
Member of Technical Staff at MPR Teltech Ltd. In 1988, he was a Lecturer
in the Department of Electronics at the Chinese University of Hong Kong.
He returned to U.B.C. as a faculty member in 1989, and currently holds
the positions of Professor and TELUS Mobility Research Chair in Advanced
Telecommunications Engineering in the Department of Electrical and Computer Engineering. Dr. Leung has co-authored more than 500 technical papers
in international journals and conference proceedings, and several of these
papers had been selected for best paper awards. His research interests are in
the areas of wireless networks and mobile systems.
Dr. Leung is a registered professional engineer in the Province of British
Columbia, Canada. He is a Fellow of IEEE, a Fellow of the Engineering
Institute of Canada, and a Fellow of the Canadian Academy of Engineering.
He is a Distinguished Lecturer of the IEEE Communications Society. He
has served on the editorial boards of the IEEE J OURNAL ON S ELECTED
A REAS IN C OMMUNICATIONS Wireless Communications Series, the IEEE
T RANSACTIONS ON W IRELESS C OMMUNICATIONS and the IEEE T RANS ACTIONS ON V EHICULAR T ECHNOLOGY , and is serving on the editorial
boards of the IEEE T RANSACTIONS ON C OMPUTERS , IEEE W IRELESS
C OMMUNICATIONS L ETTERS , Computer Communications, the Journal of
Communications and Networks, as well as several other journals. He has
guest-edited many journal special issues, and served on the technical program
committee of numerous international conferences, and contributed to the
organization of many international conferences. He is a winner of the IEEE
Vancouver Section Centennial Award and a 2011 U.B.C. Killam Research
Prize.