You are on page 1of 6

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts

for publication in the WCNC 2007 proceedings.

Novel MAC Layer Handoff Schemes for IEEE 802.11 Wireless LANs
Venkata M. Chintala and Qing-An Zeng
Wireless and Mobile Networking Laboratory
Department of Computer Science, University of Cincinnati
Cincinnati, OH 45221-0030
Email: {chintavm, qzeng}@ececs.uc.edu

Abstract — Handoff in the IEEE 802.11 wireless LANs Handoff occurs whenever a STA needs to change its as-
(WLANs) occurs whenever the mobile station (STA) changes sociation from one AP to another. In order to make a hand-
its association from one access point (AP) to another owing to
poor link quality. In a WLAN, small coverage areas of APs off, the STA must first decide when to handoff. However,
create frequent handoffs. Previous studies have shown that the the IEEE 802.11 does not specify any distinct technique to
typical handoff latency is high enough to hamper the service determine when to handoff. The common mechanism is to
quality of multimedia streams like VoIP (Voice over IP). In initiate handoff whenever the current AP RSS (Received
this paper, a novel scheme to reduce the handoff latency is Signal Strength) drops below a pre-specified threshold
proposed by using inter-AP communication to receive the
probe response(s). Another adaptive scheme to enhance the (termed as handoff threshold in the literature) [2][5]. Using
quality of service (QoS) for multimedia streams during handoff current AP’s RSS alone to initiate handoff might force the
is presented which builds upon the first scheme and adaptively STA to hold on to the AP with low signal strength while
distributes the total handoff latency. Through extensive simu- there are better APs in its vicinity. Increasing the handoff
lations, we prove that our adaptive scheme decreases handoff threshold does not solve the problem as a larger value drives
latency significantly and achieves both fast and smooth hand-
off that multimedia applications entail. the STA into performing frequent handoffs. The adaptive
scheme presented in this paper provides an elegant solution
to keep track of all the neighboring AP’s RSSs with little or
I. INTRODUCTION no performance degradation.
Recent years have seen tremendous growth in the de- Once the STA decides to handoff, the next logical step is
ployment of the IEEE 802.11 [1] WLANs in various envi- to discover the best neighboring AP. It has been shown in
ronments like campuses, companies, shopping centers and [3] that the discovery phase accounts for more than 90% of
hotels. This widespread acceptance can be attributed to the the total handoff delay. The first scheme proposed in this
flexibility and the potential bandwidth that can be offered to paper decreases the delay involved in the discovery phase
users. Many believe that the IEEE 802.11 networks will significantly by receiving the probe responses from prospec-
become a part of the integrated fourth generation (4G) net- tive APs via the previous AP, thereby avoiding the probe
works. However, the small coverage of an AP results in wait delay. Another scheme is presented later which builds
frequent handoffs and may disrupt the ongoing communica- upon the first scheme and adaptively distributes the total
tion. Empirical analysis of handoffs shows that the potential handoff latency. Based on average change in RSS, the STA
latency is around 300 ~ 500 milliseconds [2]. This high predicts the amount of time it takes before the handoff
handoff latency results in play-out gaps and poor quality of threshold is reached. During this time period, the normal
service (QoS) in multimedia applications. To meet the lofty data transfer is adaptively interleaved with scanning of all
goal of becoming the next generation networks, the QoS for prospective channels. Apart from decreasing the handoff
multimedia applications during handoff should be enhanced. latency and complying with the requirements of delay sensi-
The IEEE 802.11 standard specifies two operating modes, tive applications like VoIP, this adaptive scheme makes
infrastructure-based and infrastructure-less modes. In infra- intelligent handoff decisions possible.
structure-based mode, if a mobile station (STA) wishes to The rest of the paper is organized as follows. Section II
send or receive data, it needs to associate with a special en- presents the required background and an overview of the
tity called AP. The data transfer then is facilitated via the related work. A detailed explanation of the proposed
AP. In infrastructure-less (ad hoc) mode, the source STA schemes is provided in Section III. Simulation results are
establishes communication with the destination STA with- discussed in Section IV, followed by the conclusions in Sec-
out the help of any fixed infrastructure. In both the modes, tion V.
each STA including the AP must adhere to the MAC/PHY
specification of the IEEE 802.11. There are two coordina- II. BACKGROUND AND RELATED WORK
tion functions specified in the IEEE 802.11 MAC specifica- In this section, the complete handoff process followed by
tion – the distributed coordination function (DCF) and the most vendors is presented. Various schemes proposed in the
point coordination function (PCF). In this paper, infrastruc- literature to decrease the handoff latency are discussed later.
ture-based networks using DCF have been considered as in The complete handoff process can be divided into two logic-
[3][4][5].

1525-3511/07/$25.00 ©2007 IEEE 4438


This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the WCNC 2007 proceedings.

delay is the time spent on the channel to gather all the probe
STA APs within range on all channels responses from AP(s) operating on the channel. The stan-
dard defines two parameters to be set while scanning, Min-
ChannelTime and MaxChannelTime. The MinChannelTime
CS&T
Probe Request on is the minimum time the STA needs to spend on a channel.
channel 1 If the STA finds the channel to be busy before the Min-
ChannelTime elapses, it concludes that there exists at least
Probe Wait one AP operating on that channel. Therefore, the STA waits
till the MaxChannelTime transpires to give the AP(s) operat-
Probe Response ing on the channel enough time to send back the probe re-
sponse. Probe wait delay is therefore MinChannelTime if no
Probe Delay 2...n-1 channels are AP operates on the channel being scanned and MaxChan-
scanned likewise
nelTime otherwise.
It has been proved in [7] that the probe response delay
increases as the number of STAs and traffic load increases.
CS&T Therefore, the optimal MaxChannelTime value changes with
Probe Request on traffic conditions and no fixed value can cater all the possi-
channel n
ble traffic loads. However, the MaxChannelTime is fixed for
Probe Wait Probe Response all the handoff schemes proposed in the literature [2][6][7]
and hence results in either loss of probe responses or excess
Authentication Request New AP wait on each channel. A way of completely avoiding the use
of MaxChannelTime is presented in our schemes.
Authentication
Delay Authentication Response The reauthentication phase amounts to the delay incurred
in authenticating and transferring the state information of
Reassociation Request the STA from old AP to a new AP. The transfer of state
Association related information can be done using the IAPP (Inter Ac-
Delay Reassociation Response cess Point Protocol) [8]. In [6], the authors proposed a
scheme to reduce the handoff latency by making the APs
communicate in a distributed fashion and deduce the topol-
ogy of surroundings. Each AP maintains a list of surround-
Fig. 1 Handoff Process using Active Scanning [3] ing APs and the channels on which they operate. Therefore,
the number of channels to scan as well as the time to wait
al phases, discovery phase and reauthentication phase [3]. for the probe response on a particular channel is reduced.
Fig. 1 depicts the typical handoff process followed by most However, the scheme incurs considerable overhead to keep
network interface cards. track of the topology.
Discovery phase is the process of finding the neighboring In the SyncScan approach [9], the STA keeps track of the
AP that offers the best RSS. To accomplish this, STA scans best AP in its neighborhood by scanning channels intermit-
all the prospective channels that the neighboring APs can tently. By synchronizing the APs and forcing the APs to
operate on. Scanning can be either passive or active. In pas- transmit the beacon signals based on the channel they oper-
sive scanning, the STA switches to each channel and waits ate on, each STA is equipped with the knowledge of when
for the beacon signal. The current APs have a default bea- the beacon signals are transmitted on every channel. There-
con interval of 100 ms [6]. The STA needs to wait passively fore, the STA need not wait for the full beacon interval.
for full beacon interval and cannot stop scanning passively However, there is cost involved in the process as there will
if it receives one beacon frame as there might be more than be channel switching delay incurred for each channel scan.
one AP operation on any channel. The passive scanning has Since the STA has to come back to its current AP channel
the advantage of saving power and bandwidth. However, each time it scans, the total switching delay involved is ac-
this comes at the cost of high scanning delay. tually doubled. Though this scheme is quite effective for
interframe delay restraint applications, there will be over-
In active scanning, the STA switches to each channel,
head involved even for STAs that never make a handoff.
transmits the probe request frame and waits for probe re-
sponse(s). The total delay involved in scanning a channel Our proposed schemes significantly decrease the average
can be divided into two delays - CS&T (Channel Switch and handoff latency by using inter-AP communication in probe
Transmission) and probe wait delay [7]. CS&T delay is the phase. The scanning delay involved is decreased by forcing
time to switch and transmit on a channel while probe wait the APs to send the probe response to the previous AP (AP

4439
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the WCNC 2007 proceedings.

to which the STA is associated before handoff). The STA


STA APs within the range of STA old STA
therefore can avoid waiting for MinChannelTime or Max-
ChannelTime as discussed above. Further to facilitate CS&T Probe Request on
smooth handoff, a refined AP discovery process is used in channel 1
the second scheme presented. Using change in RSS at regu-
lar intervals to envisage the time the STA takes to reach Probe Response
handoff threshold, the enhanced probe phase presented in Probe Delay 2...n-1 channels are
scanned likewise
the first scheme is distributed and interleaved with normal
data transmission. With this approach, the handoff latency is
reduced substantially and the interframe delay requirements
of multimedia applications like VoIP are met. DISCOVERY
PHASE
CS&T Probe Request on
III. PROPOSED SCHEMES channel n

Two schemes are proposed in this section to improve the Probe Response
CS&T
handoff latency.
Probe Responses request
A. Fast Handoff by Avoiding Probe Wait (FHAP)

In FHAP scheme, the probe wait is avoided by forcing all


the neighboring AP(s) operating on the channel to send the
probe responses to the previous AP using IAPP or by a pro- Probe Responses of all
prietary protocol specific to a vendor. The STA just APs within the range of
switches to all the channels and sends the probe request. STA
After the probe phase, the STA switches back to the previ-
Fig. 2 Discovery Phase for FHAP (Fast Handoff by Avoiding
ous AP and receives the probe responses from previous AP Probe Wait)
after sending a request. The discovery phase ends at this
point and the STA resumes reauthentication process at the unanswered in the discussion. The SyncScan mechanism [9]
new AP. The whole procedure of the scheme is outlined in discussed in Section II presents a solution to this problem.
Fig. 2. However, the STA probes channels periodically all the time
Some important aspects are to be considered to make this and therefore reduces the throughput achieved. The APFH
scheme function as desired. Firstly, the handoff threshold scheme provides a better way for preemptive AP discovery.
has to be adjusted so that the STA is able to communicate In APFH scheme, the total discovery phase is divided into
with the previous AP after the probe phase. This implies several subphases and a single channel is scanned in each
that the handoff threshold be decreased slightly. Secondly, subphase. Based on the RSS, the total coverage area of an
the issue of nonarrival of the probe response from the pro- AP is divided into three zones – safe zone, grey zone and
spective APs to the previous AP needs to be addressed. To handoff zone as shown in Fig. 3. As the name implies, the
improve the effectiveness of the scheme, the STA prioritizes safe zone is the part of coverage area where the STA is not
the channels to be scanned in case the discovery phase fails. under threat of imminent handoff. Therefore, the STA does
The STA pushes all the channels on which it received the not trigger the discovery phase as yet and data transfer is
probe responses to the bottom of the prioritized list. The carried out normally.
scan is preceded to this list only if the STA is unable to find
best AP on other channels. This is to make sure that the time The grey zone is defined as the area where the probability
spent by the STA during the scanning process is not wasted. of handoff is high. The STA starts gathering information
about the next best AP once it enters the grey zone. As long
B. Adaptive Preemptive Fast Handoff (APFH) as the STA is in grey zone, the scanning process is done
intermittently with the normal data transfer. The maximum
The AP discovery process is vendor specific. There is a velocity of the STA is chosen based on previous observa-
mechanism discussed in [10] called preemptive AP discov- tions and the grey threshold is selected to comfortably inter-
ery, that compels the STA to predetermine the AP to hand- leave normal data with scanning and yet complete the dis-
off. Once the handoff threshold is reached, the STA skips covery phase. The maximum velocity chosen for our simu-
the discovery phase and triggers the reauthentication phase lations is 15 m/s as in [5]. The absolute RSS and the change
directly. This process reduces the total handoff latency and in the RSS during fixed intervals are used to estimate the
consequently the application impact as well. However, the amount of time the STA stays in grey zone. The channel
question of how the STA predetermines the AP to handoff is scan can be scheduled according to this estimated stay time.

4440
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the WCNC 2007 proceedings.

scanned after the transmission. If the estimated stay time is


AP boundary
more than the actual stay time, the STA ends up scanning all
Grey zone the channels before handoff threshold is reached. On the
other hand if the estimated stay time is less, all the channels
Safe zone
Handoff threshold
might not be scanned. In such a scenario, the remaining
channels are scanned using the normal scanning process.
. AP The probe responses are sent to the previous AP by the pro-
Handoff zone spective APs similar to the scan process of the FHAP
Grey threshold scheme. The probe responses are either piggybacked with
the normal packets in case there are packets scheduled to the
STA at that instant or sent just like the normal packets. The
line dividing the grey zone and handoff zone is the handoff
threshold. The STA triggers the reauthentication process in
the handoff zone after all the prospective channels are
Fig. 3 AP Coverage Area Divided into Zones based on Re-
ceived Signal Strength (RSS) scanned.
There are certain other benefits we can reap from the
The other alternative is to fix the channel scan schedule and APFH scheme. Since the probe response passes through the
vary the grey threshold. The former mechanism is chosen previous AP, it is equipped with the required knowledge to
over the later as varying the grey threshold might result in make intelligent decisions like load balancing. Also, optimal
premature scan of channels and consequently the STA handoff decisions can be made by taking the neighboring
might get an unclear picture of surrounding APs. APs signal strengths into account along with the previous
AP RSS unlike the conventional method described in Sec-
Each STA keeps track of the moving average of the
tion I.
change in RSS from time to time as follows. Let currsi be
the current RSS, prevsi be the RSS calculated at the previ-
ous interval, α be the smoothing factor and ∆Ψavg be the IV. SIMULATION RESULTS AND DISCUSSIONS
moving average of the change in the RSS. ∆Ψavg is calcu-
In this section, the performance of the proposed schemes
lated as follows:
is evaluated and compared with the basic handoff scheme
If ∆Ψ = currsi – prevsi, (handoff scheme currently followed by most network inter-
Then, ∆Ψavg = ∆Ψavg(1 – α) + ∆Ψα face cards). The handoff latencies of all the three schemes
for different traffic loads are presented. This is followed by
The average change in RSS is then used to estimate the discussion on the total amount of time spent for handoff for
remaining amount of time the STA stays in the grey zone. all the schemes. The effect of the proposed schemes on real
The correlation between the average change in RSS, ∆Ψavg time traffic is explored and weighed against basic handoff
and estimated stay time in grey zone is found theoretically scheme. We used C++ for simulation and a total of 100
prior to the deployment by considering a number of STAs APs, arranged as a 10 by 10 grid are considered. The IEEE
with velocities in small increments and taking a graph be- 802.11b [12] networks are considered for testing the
tween the stay time in the grey zone and the ∆Ψavg of the schemes. The other parameters taken for our simulations are
STA. Considering incremental velocities allows us to find outlined in Table I.
correlation between all possible ∆Ψavg and the associated
stay time in grey zone. The indoor path loss model given in Table I: Simulation Parameters
[11] is used to estimate RSS in our simulations. Let the es-
timated stay time at the instant the channel i is scanned be si. Parameter Value
Speed of STA 0.1 – 15 m/s
Let the number of channels to be scanned be n and the cur- Mobility Model random way point
rent time be t. Then, the channels are scheduled to be MaxChannelTime 11 ms
scanned at MinChannelTime 7 ms
{t +s0/n, t + s1/ (n-1), t + s2/ (n-2), t + s3/ (n-3)…} Switch Delay 5 ms
Number of STAs 500
The channel scan is scheduled by dividing the estimated Grey Threshold -5 dB
stay time into equal intervals and scanning the channel at Handoff Threshold for FHAP -2 dB
the end of the first interval. Normal data transmission is Handoff Threshold for -1.2 dB
carried out during the interval. The estimated stay time is Basic Handoff Scheme and APFH
reassessed after the channel is scanned and the process is α 0.1
repeated until all the channels are scanned. In case the STA Transmitting Power 13.97 dB
is in transmission at the scheduled scan time, the channel is

4441
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the WCNC 2007 proceedings.

Inter Frame Delay Basic Handoff Scheme


VoIP restriction FHAP Scheme
Handoff occurence APFH Scheme
60
180
Inter Frame Delay (ms)

Grey/Handoff Latency (ms)


50 160
140
40 120
30 100
80
20 60
10 40
20
0 0
0 100 200 300 400 500 1 2 3 4 5 6 7 8 9 10 11 12
Frame Number Traffic Load

Fig. 4 Inter Frame Delay for APFH Scheme Fig. 6 Grey/Probe Latency with Varying Traffic Load

Basic Handoff Scheme APFH Scheme


FHAPScheme Basic Handoff Scheme
APFH Scheme FHAP Scheme
VoIPConstraint 1.00E+00
200
Handoff Latency (ms)

Droping Probability 1.00E-01


150

1.00E-02
100
1.00E-03
50
1.00E-04
0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.00E-05
Traffic Load 0 0.2 0.4 0.6 0.8 1 1.2 1.4
Traffic Load

Fig. 5 Handoff Latency with Varying Traffic Load


Fig. 7 Dropping Probability with Varying Traffic Load
Fig. 4 evaluates the performance of APFH scheme for different traffic loads for all the three schemes. The FHAP
VoIP traffic. The interarrival time for VoIP applications is scheme achieves 57% and APFH attains 94.7% improve-
considered to be 20 ms as in [13]. It is recommended that ment over the basic handoff scheme. The handoff latency
the interframe delay be less than 50 ms for VoIP applica- for the basic handoff scheme is consistent with the simula-
tions [14][15]. This restriction is depicted as a horizontal tion results in [6] for similar parameters. It can also be ob-
line at 50 ms in the Fig. 4. A node with the VoIP interarrival served from Fig. 5 that the average handoff latency for basic
time is taken and the corresponding inter frame delays are handoff scheme and FHAP scheme is above 50 ms. The
shown. The vertical dotted lines represent the occurrence of average handoff latency for APFH scheme however is well
handoff. within VoIP constraints.
As shown in the Fig. 4, for APFH scheme, the interarrival We define Grey latency as the total time spent for AP dis-
time is within VoIP restrictions even when handoff occurs. covery process in the grey zone. To get an estimate of the
As discussed in section III, the handoff latency for APFH total time spent for handoff in all the schemes, the grey la-
scheme is just the reauthentication delay if all the prospec- tency incurred in APFH scheme is compared with the probe
tive channels are scanned in grey zone. On the contrary, if latency of basic handoff and FHAP schemes. It can be ob-
all the channels are not scanned, the handoff latency is the served from the figure that the FHAP scheme spends the
delay involved in scanning the remaining channels in addi- least amount of time to hand it over to a new AP. In APFH
tion to the reauthentication delay. The reauthentication scheme, there is twofold switching delay involved since the
phase is considered to be a null authentication for simula- STA has to come back to the current AP channel after scan-
tions as in [3]. Fig. 5 shows the average handoff latency for ning. However, the delay suffered from AP discovery in

4442
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the WCNC 2007 proceedings.

APFH scheme is still considerably less than the basic hand- Conference on Computer Communications (INFOCOM), June
off scheme. This is because of the time saved in scanning 2002.
the channels as explained in Section III. Though the FHAP [5] M. Raghavan, A. Mukherjee, H. Liu, Q-A. Zeng, and D. P.
scheme performs the best among all the three schemes, the Agarwal, “Improvement in QoS for Multimedia Traffic in
delay suffered is more than the permitted delay in multime- Wireless LANs during Handoff,” Proceedings of the 2005 In-
dia applications and hence is best suitable for applications ternational Conference on Wireless Networks (ICWN’05), Las
Vegas, Nevada, USA, pp. 251-257, June 27-30, 2005.
where the interframe delay is not bounded.
The multimedia packets can be dropped if the inter frame [6] H. Velayos and G. Karlsson, “Techniques to Reduce IEEE
802.11b MAC Layer Handover Time,” Proceedings of IEEE
delay at the transmitting STA exceeds a particular value, ICC, vol. 7, pp. 3844-3848, June 2004.
preset depending on the multimedia application. Fig. 7
shows the probability of the packets being dropped for dif- [7] M. Shin, A. Mishra, and W. A. Arbaugh, “Improving the La-
tency of 802.11 Hand-offs using Neighbor Graphs, ” Process-
ferent arrival rates for all the three schemes. For this pur-
ings of the ACM MobiSys Conference, Boston, MA, USA, pp.
pose, the traffic load for all the nodes is divided into real- 70-83, June 2004.
time and non real-time traffic in the ratio of 7/3 and the real-
time data packets are dropped if the inter frame delay is 50 [8] “IEEE Trial-use Recommended Practice for Multi-Vendor
Access Point Interoperability via An Inter-access Point Proto-
ms. The APFH scheme outperforms the other two schemes
col across Distribution Systems supporting IEEE 802.11 Op-
as long as the traffic load is restricted. At low traffic load, eration,” IEEE Std 802.11F, July 2003.
only handoff accounts to the violation in VoIP restriction.
However, at high traffic load, the channel access delay [9] I. Ramani and S. Savage, “SyncScan: Practical Fast Handoff
for 802.11 Infrastructure Networks,” Proceedings of the IEEE
breaches the multimedia limitations and hence all the three Infocom, vol. 1, pp. 675-684, March 2005.
schemes suffer equally.
[10] P. Roshan and J. Leary, 802.11 Wireless LAN Fundamentals,
CISCO Press, ISBN No.1587050773.
V. CONCLUSIONS
[11] R. Gandhi, “Empirical Path Loss Models for 802.11b Links,”
The handoff delay perceived in the existing IEEE 802.11 Master’s Thesis, Indian Institute of Technology, Kanpur,
WLANs is well above the delay constraints of multimedia 2003.
applications like VoIP. Two novel schemes to reduce hand- [12] IEEE Std. 802.11b, Supplement to Part 11: Wireless LAN
off latency by using inter-AP communication were proposed Medium Access Control (MAC) and Physical Layer (PHY)
in this paper. The first scheme proposed avoided probe wait specifications: Higher-speed Physical Layer Extension in the
delay by receiving all the probe responses from the old AP. 2.4 GHz Band, IEEE Std. 802.11b-1999, 1999.
Though the handoff latency is decreased considerably, this [13] Y. Chen, N. Smavatkul, and S. Emeott, “Power management
scheme does not meet the recommended multimedia restric- for VoIP over IEEE 802.11 WLAN,” Proceedings of the
tions. Another scheme has been presented which spreads out IEEE WCNC 2004, vol.5, pp.1648–1653, March 2004.
the total handoff latency and interleaves normal data trans- [14] International Telecommunication Union, “General Character-
mission with channel scan. This scheme can meet the VoIP istics of International Telephone Connections and Interna-
constraints. tional Telephone Circuits,” ITU-TG.114, 1988.
[15] R. Shirdokar, J. Kabara, and P. Krishnamurthy, “A QoS-
based Indoor Wireless Data Network Design for VoIP,” Pro-
REFERENCES ceedings of IEEE Vehicular Technology Conference 2001,
[1] IEEE Std. 802.11-1999, Part 11: Wireless LAN Medium Ac- vol. 4, pp. 2594–2598, October 2001.
cess Control (MAC) and Physical Layer (PHY) Specifica-
tions, IEEE Standard 802.11, 1999.
[2] A. Mishra, Min ho Shin, and W. Arbaugh, “An Empirical
Analysis of the IEEE 802.11 MAC Layer Handoff Process,”
ACM SIGCOMM Computer Communication Review (ACM
CCR), vol. 33, pp. 93-102, April 2003.
[3] G. Bianchi, L. Fratta, and M. Oliveri, “Performance Evaluation
and Enhancement of the CSMA/CA MAC Protocol for
802.11 Wireless LANS,” Proceedings of the 7th International
Symposium on Personal, Indoor and Mobile Radio Communi-
cations (PIMRC), vol. 2, pp. 392 – 396, 1996.
[4] H. Wu, Y. Peng, K. Long, S. Cheng, and J. Ma, “Performance
of Reliable Transport Protocol over IEEE 802.11 Wireless
LAN: Analysis and Enhancement,” Proceedings of the IEEE

4443

You might also like