2009 International Conference on Communications and Mobile Computing

Train-Ground Communication in CBTC Based on 802.11b：Design and

Performance Research
Li Zhu, Yan Zhang, Bin Ning and Hailin Jiang
Rail Traffic Control and Safety State Key Laboratory, Beijing
Jiao tong University, Beijing, China
02212090@bjtu.edu.cn

Abstract of 802.11b without violating the standard [1]; D.

Valerio studied the Wireless mesh networks based on
The 802.11b physical protocols applied in 802.11, and he also proposed some methods to
Communication Based Train Control (CBTC) Train- optimize the IEEE 802.11 parameters for wide area
Ground communication system are simulated in this coverage [3].But they didn’t study it when it is used in
paper, and the simulation result is used to determine train control system, and neither did they study the
the AP distance. The train speed impact on physical packet loss rate or the delay performance which are
FER is also studied by the simulation result. As a very significant to CBTC.
protocol used in a fully mobile outdoor environment, The paper is organized as follows: after a brief view
the packet delay and packets loss rate performance of of CBTC train-ground communication architecture, the
802.11 DCF is analyzed at last. AP deployment distance is determined by 802.11b
radio link budget in section 2; the protocol parameters
1. Introduction1 are also analyzed and optimized to meet the high real-
time in this section; in section 3, the train speed impact
Communication Based Train Control (CBTC) [6] is on physical FER, the packet loss rate and delay
a continuous and automatic train control system, performance of the designed train-ground
utilizing high-resolution, train location determination communication system are studied; A conclusion is
independent of track circuits; continuous, high- drawn in section 4.
capacity, bidirectional train-to-wayside data
communications; and train borne and wayside 2. Train-Ground Communication System
processors capable of implementing vital functions. Design
The train-ground communication, which is the basis
for train control system, is one of the key technologies 2.1. System architecture
for CBTC. At present, IEEE 802.11 series standard
WLAN has become one of directions, to which train-
ground communication can further develop.
Backbone Network
As one of IEEE 802.11 series standard, 802.11b has ZC
strong capacity of Anti-interference and Anti- AP1 AP2 AP3
Multipath[4][5] because of its spread spectrum and
rake receiver technology. But as a civil standard used
in CBTC train-ground system, the AP deployment,
protocol parameter need detailed design to meet the Train1 Train2
high reliability and high real-time of CBTC.
In the last few years, many foreign scholars have Figure 1. Train-Ground Communication System in
studied outdoor IEEE 802.11 WLAN performance. CBTC
Martin V. Clark et.al analyzed the maximum coverage

1
Figure 1 depicts a simplified view of the Train-
This paper is sponsored by National Natural Science Foundation of Ground communication system architecture, which
P. R. China under grant No.60634010, with the title "The Theory and
Key Technology Research of Train Control System", and is also
consists of the Backbone Network, zone centers (ZC),
supported by Urban Rail Transit Automation and Control Beijing Access points (AP1... AP3), wireless channel and the
Municipal Government Key Laboratory.

978-0-7695-3501-2/09 $25.00 © 2009 IEEE 368

DOI 10.1109/CMC.2009.93
transceiver in the trains. The service area along the receiver has a SNR = -76-(-100) = 24dB. Compared to
railway is divided into cells, which constitutes the the simulation result 7dB, it has a great surplus.
Extended Service Sets (ESS) and realizes the coverage In CBTC system, due to the restrictions of lines and
of larger area. The moving train receives moving vehicles, the antenna cannot be installed too high. The
authority (MA) sent by ZC from wayside AP, then it low antenna height means that the Fresnel zone
calculates its restrictive speed and sends its location typically limits the propagation of 2.4 GHz WLAN to
information to ZC. fairly short ranges. David B. Green [2] proposes an
equation for RF propagation close to the ground at
2.2 AP Distance design 2.4G frequency,
Ploss =7.6 +40log10 d - 20Log10 h t h r (2)
802.11b standard is compatible with 802.11. There
are 4 data rates in the standard, 1Mbits/s and 2Mbits/s Where Ploss is path loss, ht , hr are the height of the
adopts DSSS (Direct sequence spread spectrum) transmitting and receiving antennas; d is the overall
technology, 5.5Mbits/s ， 11Mbits/s adopts CCK distance, and all the units in (2) are dB.
modulation .The standards specified the receiver Combined with large-scale path loss and small-scale
minimum input level sensitivity as follows: The frame fading (Raleigh distribution is used to describe the
error ratio (FER) should be less than 8% at a PSDU fading envelope), the received power is:
length of 1024 octets for an input level of –76 dBm Pr = Pt − Ploss + ξ + 10 log γ + Gt + Gr (3)
measured at the antenna connector. This FER should
Where Pt is transmitted power, Pr is received power;
be specified for 11MBits/s CCK modulation. And for
ξ is a Gaussian random variable which has a variance
2Mbits/s DQPSK modulation, the frame error ratio
of δ and a mean value of 0; γ is a Rayleigh random
(FER) should be less than 8% at a PSDU length of
1024 octets for an input level of –80 dBm measured at variable, with its mean value of 1; Gt and Gr are the
the antenna connector. antenna Gain for the transmitter and receiver(Cable
This paper combines it++ library (a c++ library for loss is included).
communication simulation [8]) with C++, to simulate In China, the ISM band limits the EIRP to be less
the FER of 4 data rates under Gaussian channel (the than 27dbm when the antenna gain is more than 10 dBi.
frame size is 1024 Bytes) according to 802.11b and In order to let receivers get the best signal power, we
802.11 standards. The result is shown in Figure 2. assume the parameters in (2) (3) as follows:

Table 1. Parameter for Simulation

Figure 2. FER-SNR under AWGN channel for We randomly generate 10000 random numbers for ξ
1024 Bytes PSDU and γ , submit all the parameters into (2), then get the
probability that the received power is less than -76dBm
From Figure 2, we can see: in order to get an 8%
under different train-ground distances, as shown in
FER, The receiver SNR needs to be more than 7dB. At
table 2. We can see From Table 2 that when train-
a room temperature, thermal noise power density N0 =
ground distance is less than 200m, the probability that
-174dbm/HZ; Signal bandwidth BW= 22MHz, so we
the received power less than -76dbm is acceptable.
can get the thermal noise power:
As shown in Figure 1, the antenna on the train head
Pn =N 0 +10lg(BW) = -100dBm (1) and the antenna for the ground AP are installed face to
When the received signal power reaches the face. The wireless equipment on the train usually
minimum receive power specified by 802.11b, the associates with the closest AP whose antenna is face to

369
itself. When the train passes one associated AP, it The Ack-Timer-out period, which is not specified in
switches to associates with a new AP because its the IEEE standards, is usually chosen to be a value
antenna and the antenna for the old AP are already much larger than 10us by equipment manufacturers
back-to-back. From the above analysis, we can see that [1], but when DCF is used in CBTC, it should be
the AP deployment distance is roughly equal to the redesigned to get a small delay.
maximum train-ground distances. Figure 3 shows that as long as the ACK arrives at
In engineering design, we usually set the AP the sending station within the Ack-Timer-out period,
distance to be 200m around, the reasons are: the DCF can work normally. The Ack-Timer-out
When train-ground distance is less than 200m, the period should be greater than the sum of data
probability that the received power less than -76dbm is propagation time, SIFS, and ACK propagation time.
very small. If we set the AP distance to be much less When the train-ground distance is less than 300m, the
than 200m, more APs need to be installed on the same data and ACK propagation time are less than 1us. The
line, which cost too much. On the contrary, AP SIFS in 802.11b is set to be 10us.So Ack-Timer-out
distance much more than 200m cannot fully meets the period should be greater than 12us.We set Ack-Timer-
minimum received power. out to be 20us to give an adequate margin.
The maximum packet retransmission is also
Table 2. The probability that the received power is changeable in the standard. More packet
less than -76dBm under different train-ground retransmissions will reduce the packet loss rate. The
distances equipment manufacturers usually set the packet
retransmission at a large number, because when the
wireless link is established, FER<<1, the probability of
n times transmission (FER)n is almost zero;

2.3 Mac protocol Parameter design

802.11 series standards share the same Mac protocol. Figure 3.Data Transmission Process in DCF Basic
It supports two operation modes: the point access method
coordination function (PCF) and the distributed
function (DCF). The PCF has not been implemented, 3. Train-Ground Communication System
but only the DCF is available in commercial products. Performance Analysis
This paper studies the DCF performance applied in
train-ground communication system. 3.1 The Impact of Train Speed on FER
The DCF is based on CSMA/CA and it consists of
both a basic and an optional channel access method When 802.11b is used in CBTC system, we need to
using RTS/CTS exchanges. The basic access method is think about the impact of the train speed on physical
studied in this paper. Firstly a station (including the transmission. After AP deployment in section 2.2, we
AP) with a new packet ready for transmission senses have a SNR more than 20dB along the railway. This
whether the channel is busy. After finding the channel paper simulate the FER of Train-Ground
idle for a DIFS interval, the station starts packet communication system which works in 11Mbits/s
transmission. Otherwise the station continues to sense operation mode, under Rayleigh channel with different
the channel. After finding the channel idle for a DIFS Doppler shifts, according to 802.11b and 802.11
interval, the station waits for a random bakeoff interval standard. The frame size is 400Bytes (including high-
to send the packet. When the packet is transmitted, the level overhead) which is also the typical frame size in
ACK is expected to be received within a Ack-Timer- CBTC. The simulation result is shown in Figure 4:
out period. If the ACK is not received in the period, the
station will continue to sense the channel. After finding
the channel idle for a DIFS interval, the station waits
for a random bakeoff interval to send the packet, and
the backoff window is twice of the last backoff widow.
The data transmission process is shown in Figure 3.

370
 The 128 synchronization bits in the packet preamble
are used to estimate the channel response in 802.11b.
Table 2 shows that the coherence time is closer to the
frame duration 483us when the train speed increases,
which means the channel response fluctuate widely in
the duration of the frame and the channel response
estimated by the preamble is not correct. This leads to
an increased FER.
From Figure 4, we can see that the frame with a
PSDU of 400 Bytes has a FER less than 8% when the
trains speed is less than 120km/h.
Figure 4：FER under Rayleigh Channel
As shown in Figure 4, the FER increases when the 3.2 Packet loss rate and Packet Delay
train speed increases. The reasons are: the 802.11b Performance
standard is designed for indoor wireless LAN, Figure 5
illustrates the frame structure of IEEE 802.11b: Packet loss rate and Packet delay are two important
parameters for train-ground system. The control
information cannot be sent to train due to packet loss
which has a great impact on the efficiency of the traffic.
The packet delay must be strictly less than CBTC train
Figure 5. Frame structure of IEEE 802.11b communication cycle, otherwise the information
received by the train is not real-time information,
From the figure it can be seen that each data frame which has a bad impact on control accuracy.
consists of a 144 bit preamble, 48 bit header and a We simulate the packet loss rate for different
MAC frame. The first 192 bits, i.e. the preamble and retransmissions, given an 8% FER. The result is shown
the header, are always transmitted using a bit rate of 1 in Table 3:
Mbit/s, while the MAC frame is transmitted with the The result in Table 3 shows that the packet loss rate
selected data rates. So we can get the frame duration (when the physical FER is 8%) decreases to 10-8 when
for 400 Bytes PSDU at 11Mbits/s data rates is we have 6 retransmissions. Retransmissions more than
192+400*8/11 = 483us. 6 are negligible because it has a very small probability.
The Doppler shift which is caused by train
movement will lead to a smaller coherence time. When Table 3. Packet loss rate with different
the normalized coherence function value is above 0.5, retransmissions
the coherence time is approximately [7]:
9
Tc = (4)
16π f m
Where Tc is coherence time, fm is the maximum
Doppler shift in frequency from carrier frequency:
v
fm = * fc (5)
c According to the DCF basic access method, as
Where fc is carrier frequency (for 802.11b it is 2.4G), shown in Figure 3, when the packet is transmitted for n
v is train speed, we calculate the coherence time times, we calculated the maximum packet delay as
coherence time for different train speed, and the result follow:
is shown in table 2:
Tdelay =Tdata +Tack _timerout + DIFS +Tbackoff (1)
Table 2. Coherence time Vs Train speed
+Tdata +Tack_timerout + DIFS +Tbackoff (2) +
...+Tdata +Tack_timerout + DIFS +Tbackoff (n−1) (6)

+Tdata +Ttransfer

371
 Where Tdata is the time needed to transmit a data 4. Conclusion
frame. Tack_timerout is the Ack-Timer-out period, as
analyzed in section 2.3, we assume Tack_timerout = This paper introduces the CBTC train-ground
20us.Tbackoff(i) is the backoff time of the communication architecture. By simulation, we set the
retransmission at i times. Ttransfer is the propagation AP distance to be 200m around in order to meet the
time of the data,in this paper, the train-ground distance minimum received power. The FER increases as the
is less than 300m,so we have Ttransfer=1us. train speed increases due to the channel estimation
We calculate the maximum packet delay under error. The packet delay increases rapidly when more
different retransmissions and at different data rates retransmissions are needed .Given an 8% FER, typical
when transmitting the typical 400Bytes frame in CBTC CBTC frame with six retransmissions lead to a packet
(we set Tbackoff(i) to be the backoff window size to get delay of 90ms,when the data rates is 1Mbits/s.We need
the maximum packet delay). The result is shown in to improve the received signal power to reduce the
Figure 6: retransmissions as far as possible.

5. Reference
[1] Kin K.Leung ,Zoran Kostic Outdoor IEEE 802.11
Cellular Networks: Radio and Mac Design and Their
Performance. IEEE Trans.Veh.Technol, vol.56,
pp.2673-2684, September 2007:2 - 4
[2] David B. Green ,M. S. Obaidat An Accurate Line of
Sight Propagation Performance Model for Ad-Hoc
802.11 Wireless LAN (WLAN) Devices.
Communications,2002.ICC 2002.IEEE International
Conference on, vol.5, pp. 3424 - 3428, May 2002:2 – 4
[3] D. Valerio, L. De Cicco et. al. Optimization of IEEE
802.11 parameters for wide area coverage. Proceedings
Figure 6. Maximum Packet delay under different of MEDHOCNET 2006
retransmissions and at different data rates
[4] IEEE 802.11 Wireless LAN medium access control
(MAC) and physical layer (PHY) specifications.1999.
When the packet is transmitted for 6 times, and the
[5] IEEE 802.11b Part 11: Wireless LAN Medium Access
data rates are 1Mbits/s, the maximum packet delay
almost reaches 90ms, which means that the CBTC Control (MAC) and Physical Layer (PHY)
communication cycle must be strictly more than specifications. 1999.
90ms.In order to decrease the CBTC communication [6] IEEE Standard for Communications-Based Train
cycle to improve control accuracy, we need to improve Control (CBTC) Performance and Functional
the received signal power to reduce the retransmissions Requirements.1999
as far as possible. And a stronger received signal [7] Theodore S. Rappaport Wireless Communications[M]
power can make the system work in high data rates Beijing: Publishing House of Electron
which also reduce the packet delay. [8] IT++ help http://itpp.sourceforge.net/ics industry 2004

372