Novel Handoff Scheme with MIMO ◾ 159
8.3.1.2 Synchronization in the Downlink
Each AP receiving the probe request packet will return a probe response to the
MS in sequence in the existing scheme, which spends a big part of the time in the
hando procedures. And the MS may receive a lot of probe response signals from
the rail-side APs and other interference sources in this stage. However, in CBTC
environments only the packets from the serving AP and the hando candidate AP
are the expected ones. To avoid the unnecessary transmission of the probe response
packet, the MS transmits the location information in the “Vendor Specic” domain
of probe request packet. us, the APs learn the location of the MS and will send
the probe response packets only when having received the probe request packet
from the specic hando locations, which ensures that only the two involved APs
return the probe response packets. And the other APs outside the CBTC WLAN
will not respond to the probe request packet because the service set identier (SSID)
of the CBTC WLAN is unknown to them.
To reduce the latency in this stage, we propose that the serving AP and the hand-
o candidate AP send the probe response packets synchronously in the downlink.
However, in WLAN system the packets are transmitted with carrier sense multiple
access with collision avoidance (CSMA/CA) protocol, and the random backo will
happen before the transmission of probe response. If the packets sent by the dier-
ent APs have dierent backo delay, the packets cannot arrive at the receiver at the
0 10 20 30 40 50
E
s
/N
0
(dB)
SU MX (DoF = 2)
MU STBC (DoF = 4)
SU STBC (DoF = 8)
10
−6
10
−5
10
−4
10
−3
10
−2
10
−1
10
0
FER
Figure8.3 FER of different transmission schemes.
160 ◾ Advances in Communications-Based Train Control Systems
same time even though they are transmitted simultaneously with the transmission delay
considered. To achieve the synchronous reception of the packets, revisions of the packet
transmission scheme in 802.11 standards are implemented. erefore, we set that the
two APs receiving the probe request transmit the probe response packets after the SIFS
interval, instead of the DIFS plus the backo duration dened in the specication.
In this way, the arrival time at the MS is synchronized and the interference cancel-
ation algorithm can be implemented. However, the application of SIFS also guarantees
that the probe response packets from unwished sources cannot be sent earlier than the
expected packets, and the timer of the MaxChannelTime, which is the amount of time
to wait to collect potential additional probe responses from other access points, can be
set to small enough to reduce the latency in the scanning stage. e SNRs of the signals
from the serving AP and the hando candidate AP are much stronger than the interfer-
ence signals in general; therefore, the subsequent transmissions of the probe response
packets have little impacts on the communication performance.
In the proposed scheme, the hando triggering location is predened and strong
line-of-sight (LOS) components exist between the AP and the MS, so the propaga-
tion delay between the MS and the two APs can be got in advance. For the APs
involved in the hando procedures, there are four hando locations as shown in
Figure8.4, where two are on either side of AP
2
. en the propagation delay
tp
ij
,
for
the AP
2
is calculated as
tp dc
ij
ij i,j,
,,,4
== =
/1
,1
,3
…
(8.5)
where:
c is the speed of light
d
i,j
is the distance from the MS to the AP
j
at the ith hando location
e delay dierences of the four hando locations around the AP
2
are calculated as
∆ttptpi ji j
iiij,2 ,2
==1,3, =1;=2,4,
=3
−
{} {}
,
,
(8.6)
AP
1
d
11
d
12
d
22
d
23
d
31
d
32
d
42
d
43
AP
2
AP
3
1
3
4
Handoff locations
2
Figure8.4 Delay difference between adjacent APs.
Novel Handoff Scheme with MIMO ◾ 161
e AP
2
stores the delay dierence of the hando locations, and when the MS sends
the probe request to the AP, the AP learns the position and the moving direction of
the MS from the received packets, and then knows that it is at the hando location
point i. en the AP can transmit the probe response packet to the MS, with the
delay dierence Δt
i
counted in, to guarantee the synchronous reception at the MS.
Suppose the hando from AP
1
to AP
2
initiates at hando location 1, andthe time
instants that the AP
1
and AP
2
receiving the probe request are T
rx1
and T
rx2
, respec-
tively, and the processing delay and transmission delay are
t
proc
essing
and
t
tx
, which
are supposed the same for all the APs in this chapter. Without loss of generality we
suppose the transmission delay at the rst hando location has Δ
t
1,2
<0
, then the
transmit instants for the AP
1
and AP
2
to send the probe response packet are calcu-
lated as follows:
TTtt
tx1 rx1processing tx
=+ +
(8.7)
TT
ttt
tx2 rx2processing tx
2
=+ +−×∆
1,2
(8.8)
From the equations, we can see that the APs only have to store the delay dierences
with negative values. One example is with the format of (Δt
1,j
, Δt
4,j
).
In the CBTC communication system, OFDM is adopted due to its anti-
multipath characteristics, with the cyclic prex (CP) length of 0.8μs. With the
CP, if the timing dierences between uplink clients are in the limit of half of the
CP length, 0.4μs, the signal can be still be decoded successfully [28]. e accuracy
of the location in CBTC system is about 5–10m; therefore, the delay uncertainty
between the two APs due to locating uncertainty is less than 0.067μs, which is far
less than the 0.4μs and CP length of OFDM packets. erefore, little uncertainty
of the locating results has no eects on synchronization performance.
8.3.2 Communication Latency in WLANs
In the WLAN system in CBTC, CSMA/CA is the media access control (MAC)
protocol for all the stations. Two reasons resulting in the loss of packets are as
follows: (1) the loss because of the collisions when the dierent nodes send the
packets simultaneously, which is caused by the contention of dierent transmitters,
and (2)the transmission error, where a packet is received without packet collisions
and is corrupted due to low SNR. In this chapter, we only consider the latter case
because in typical CBTC environments there is only one train in the same direc-
tion in the AP’s coverage range, which is decided by the requirements of the safety
distance between consecutive trains.
With packet loss caused only by transmission error, the packet delay
T
ho sig_
of
one single hando signaling packet in CSMA/CA systems is calculated as follows:
TT
TT TT
TT
NT
ho_sig PtxDIFS CCA RxTx
Preamble PLCP backof s
=++++
++
+×
ll ot
(8.9)
162 ◾ Advances in Communications-Based Train Control Systems
where:
T
P
is the propagation latency
T
tx
is the transmission latency
T
DIFS
is the DIFS
T
CCA
is the time to sense the channel
T
RxTx
is the latency when the station transits from receiving to transmitting status
T
Preamble
and
T
PL
CP
are the times to transmit preamble and physical layer conver-
gence protocol (PLCP) packet, respectively
N
backof
is the number of backo slots before transmitting data
T
slot
is the slot length in 802.11
T
tx
can get as
TL
R
tx fr
=
/ , where R is the data rate and
L
fr
is the frame length in bits.
In normal hando procedure, the hando is divided into scanning, authentica-
tion, and reassociation stages, where in the scanning and reassociation stages, there
are two signaling packets’ transmission, and in the authentication stage, there are
four signaling packets [2] in the fast transition protocol. Suppose that the sizes
of all the hando signaling are the same, and during the hando procedure, the
transmission error happens at most once due to the fact that the probability that
two successive signaling packets both fail is very small. If any one of the authenti-
cation and reassociation packets is lost, the hando procedure will start from the
beginning of the hando procedure. It will take some time before the SA starts over
again. We refer to that time as
T
wait
. e hando latency is calculated as
TP
TmChTime
PP T
HO_latency fr ho_sig
fr fr ho_si
)
=− ×+
+−×
(1 )( )
(1 (
8
7
ggbackofslo
tw
ait
+××+++
++
NT mChTimeT)(01 7)
(8.10)
where:
mChTime
is the maximum time to spend on each channel when nding that the
channel is not idle during the scanning stage
P
fr
is the FER
8.4 Analysis of Handoff Performance
8.4.1 Wireless Channel Model
e transmitted signal will experience dierent fading of three factors: path loss,
shadow fading, and fast fading. Fast fading such as Rician or Rayleigh fading has
shorter correlation distance and faster variation compared with shadow fading, and
the eects are usually averaged out when applying the trigger conditions of hando.
e path loss model for 2.4GHz signal near the ground is proposed in [29] as follows:
Pd
dh
h
loss
tr
lo
gl
og() 7.6 40 20
10 10
=+ −
(8.11)
Novel Handoff Scheme with MIMO ◾ 163
where:
d is the distance between the transmitter and the receiver
h
t
and h
r
are the height of the transmit antenna and receive antenna, respectively
erefore, the received signal power from the serving AP
1
and candidate AP
2
,
expressed in dBm, when the distance between the MS and AP
1
is d, is formulated
as follows:
ad PGGLLPdud() ()
()=+ +−−− +
ttrtrloss
(8.12)
bd PG GLLPDd vd()
()=+ +−−− −+
ttrtloss
r
()
(8.13)
where:
P
t
is the transmit power
G
t
and G
r
are the transmitter and receiver antenna gains, respectively
L
t
and L
r
are the insertion losses in the transmitter and the receiver
D is the distance between the two adjacent APs
ud()
and
vd()
are zero-mean Gaussian random variables, which model the shadow
fading
e shadow fading is supposed to be auto-correlated with the auto-correlation
function as follows:
Fudu
dddd
{( )( )} (| |/ )
12 120
=−−σ
s
2
exp
(8.14)
where d
0
is the correlation distance and determines the fading rate of the correla-
tion with the distance.
8.4.2 Optimal Handoff Location
As introduced in Section 8.3, when the train reaches the predetermined hand-
o location, the MIMO transmission mode will switch to multiuser multi-
plexing from the single-user STBC transmission mode, which will degrade
the transmission performance. To determine the optimal hando location
forthe multiple antenna conguration, we suppose that the hando happens
at theplace where the FER of data transmission with the serving AP is same as
the FER of transmission of the hando signaling, which is transmitted between
the train and the candidate AP. Because the FER is identical, the BER of the
communication links is the same as well according to Equation 8.4. erefore,
we have
αγ αγ
ββ
A
A
B
B
Dd d
21
() ()−
[]
=
[]
−−
(8.15)
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.