ieeetransactions on intelligent transportation …...since cbtc systems are mostly deployed in...

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS 1

Finite-State Markov Modeling for WirelessChannels in Tunnel Communication-Based

Train Control SystemsHongwei Wang, Fei Richard Yu, Senior Member, IEEE, Li Zhu, Tao Tang, and Bin Ning, Senior Member, IEEE

Abstract—Communication-based train control (CBTC) is beingrapidly adopted in urban rail transit systems, as it can signifi-cantly enhance railway network efficiency, safety, and capacity.Since CBTC systems are mostly deployed in underground tunnelsand trains move at high speeds, building a train–ground wirelesscommunication system for CBTC is a challenging task. Modelingthe tunnel channels is very important in designing the wirelessnetworks and evaluating the performance of CBTC systems. Mostexisting works on channel modeling do not consider the uniquecharacteristics of CBTC systems, such as high mobility speed,deterministic moving direction, and accurate train-location infor-mation. In this paper, we develop a finite-state Markov channel(FSMC) model for tunnel channels in CBTC systems. The pro-posed FSMC model is based on real field CBTC channel mea-surements obtained from a business-operating subway line. Unlikemost existing channel models, which are not related to specificlocations, the proposed FSMC channel model takes train locationsinto account to have a more accurate channel model. The distancebetween the transmitter and the receiver is divided into intervalsand an FSMC model is applied in each interval. The accuracy ofthe proposed FSMC model is illustrated by the simulation resultsgenerated from the model and the real field measurement results.

Index Terms—Communication-based train control (CBTC),finite-state Markov chain (FSMC), wireless local area network(WLAN).

I. INTRODUCTION

URBAN rail transit systems are rapidly developing aroundthe world. Due to a great deal of urban traffic pressure,

improving the efficiency and capacity of urban rail transitsystems is an increasing demand. Because it is a key subsystem

Manuscript received May 13, 2013; revised July 30, 2013 and October17, 2013; accepted November 26, 2013. This work was supported in part bythe National Natural Science Foundation of China under Grant 61132003; bythe National High Technology Research and Development Program of China(863 Program) under Grant 2011AA110502; by the Doctoral Program of theMinistry of Education under Grant 20130009120036; by the Foundation ofBeijing Scientific Committee under Grant D131100004113002; by the KeyProject of the Beijing Laboratory of Urban Rail Transit and the BeijingKey Laboratory of Urban Rail Transit Automation and Control under GrantsRCS2012K010, RCS2012ZQ002, RCS2011ZT010, and 2011JBZ014; by theChina Education Ministry Funding Project under Grant 2013JBM124; andby the China Scholarship Council. The Associate Editor for this paper wasL. Vlacic.

H. Wang, L. Zhu, T. Tang, and B. Ning are with State Key Laboratory of RailTraffic Control and Safety, Beijing Jiaotong University, Beijing 100044, China.

F. R. Yu is with the Department of Systems and Computer Engineering,Carleton University, Ottawa, ON K1S 5B6, Canada.

Color versions of one or more of the figures in this paper are available onlineat http://ieeexplore.ieee.org.

Digital Object Identifier 10.1109/TITS.2014.2298038

of urban rail transit systems, communication-based train control(CBTC) is an automated train control system that makes useof train–ground wireless communications to ensure the safeoperation of rail vehicles [1]. It can enhance the level of safetyand service offered to customers and improve the utilization ofrailway network infrastructure [2]. CBTC is a modern successorto the traditional railway signaling system using interlockings,track circuits, and signals [3].

Building a train–ground wireless communication system forCBTC is a challenging task. As urban rail transit systemsare mostly deployed in underground tunnels, there are largeamounts of reflections, scattering, and barriers that severely af-fect the propagation performance of wireless communications.Moreover, due to the available commercial off-the-shelf equip-ment, wireless local area networks (WLANs) are often adoptedas the main method of train–ground communications for CBTCsystems. However, most of the current IEEE 802.11 WLANstandards were not originally designed for the high-speed en-vironment in tunnels [4]. Furthermore, the fast movement oftrains will cause frequent handoffs between WLAN accesspoints (APs), which can severely affect CBTC performance.

Modeling the channels of urban rail transit systems is veryimportant in designing the wireless networks and evaluatingthe performance of CBTC systems. There are some previousworks on radio wave propagation in urban rail transit systems.A path loss model of tunnel channels is given in [5], whichdescribes the characteristics of large-scale fading. Guan et al. in[6] presented the propagation characteristics based on real en-vironment measurements in the Madrid, Spain, subway. A two-layer multistate Markov model is presented in [7] for modelinga 1.8-GHz channel in urban Taipei City, Taiwan. Based on theWinner II physical-layer channel model parameters, Lin et al.in [8] proposed a channel model for a high-speed railway.

Although some excellent works have been done on modelingchannels, most do not consider the unique characteristics inCBTC systems, such as high mobility speed, deterministicmoving direction, and accurate train-location information. Inthis paper, we develop a finite-state Markov channel (FSMC)model for tunnel channels in CBTC systems. FSMC modelshave been widely accepted in the literature as an effectiveapproach to characterizing the correlation structure of thefading process, including 1.8-GHz narrow-band channels [7],high-speed railway channels [8], satellite channels [9], indoorchannels [10], Rayleigh fading channels [11], Ricean fadingchannels [12], and Nakagami fading channels [13]. Using

1524-9050 © 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

Downloaded from http://www.elearnica.ir

http://www.ieee.org/publications_standards/publications/rights/index.html


2 IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS

Fig. 1. CBTC system.

FSMC models, a variety of analytical results of system per-formance can be derived, including channel capacity [14],throughput [15], and packet error distribution [16].

To the best of our knowledge, FSMC models for tunnelchannels in CBTC systems have not been studied in previousworks. Therefore, there is a strong motivation to develop anFSMC model for tunnel channels in CBTC systems. Somedistinct features of the proposed channel model are as follows.

• The proposed FSMC model is based on real fieldCBTC channel measurements obtained from the business-operating Beijing Subway Changping Line.

• Unlike most existing channel models, which do not usetrain-location information, the proposed FSMC channelmodel takes train locations into account to create a moreaccurate channel model.

• The distance between the transmitter and the receiver isdivided into intervals, and the FSMC model is applied ineach interval.

• The Lloyd–Max technique [17] is used to determine thesignal-to-noise ratio (SNR) level boundaries in the pro-posed FSMC model.

• The accuracy of the proposed FSMC model is illustratedby the simulation results generated from the model andthe real field measurement results. The effects of differentparameters are also discussed.

The rest of this paper is organized as follows. Section IIdescribes an overview of CBTC systems. In Section III, the realfield measurement configuration and scenario are described. InSection IV, the FSMC model is introduced. Then, Section Vpresents the real field measurement results and discussions.Finally, this paper is concluded in Section VI with future work.

II. OVERVIEW OF CBTC

Fig. 1 describes a CBTC system. In this system, continuousbidirectional wireless communications between each mobilestation (MS) on the train and the wayside APs are adoptedinstead of the traditional fixed-block track circuit. The railwayline is usually divided into areas or regions. Each area is underthe control of a zone controller (ZC) and has its own radiotransmission system. Each train transmits its identity, location,direction, and speed to the ZC. The radio link between eachtrain and the ZC should be continuous so that the ZC knows

the locations of all the trains in its area all the time. The ZCtransmits to each train the location of the train in front of it andgives it a braking curve to enable it to stop before it reachesthat train. Theoretically, as long as each train is traveling atthe same speed and they all have the same braking capability,they can travel together as closely as a few meters betweenthem. When a train moves away from the coverage of an APand enters the coverage of another AP along the railway, thehandoff procedure may result in communication interruptionand long latency. In CBTC systems, it is important to maintaincommunication link availability in order to guarantee trainoperation safety and efficiency.

Wireless channels in CBTC systems are different from thosein other wireless systems, since most CBTC systems are de-ployed in underground tunnels, where there are large amountsof reflections, scattering, and barriers that severely affect thepropagation performance of wireless communications. In orderto design the wireless networks and evaluate the performance ofCBTC systems, modeling the tunnel channels in CBTC systemsshould be carefully studied.

III. REAL FIELD CBTC CHANNEL MEASUREMENTS

The objective of the real field CBTC channel measurementsis to get the real field data of WLAN propagation in tunnelsunder real conditions of the subway line, which will be usedto build an FSMC model. Here, we present the measurementequipment and the measurement scenario in our real fieldCBTC channel measurements.

With this objective, the preparation of the measurementsconsists of two parts as follows.

1) We need to make sure that the configuration of the mea-surements is the same as business-operating subway lines,including the choice of antennas, the location of antennas,and the settings of the transmitter and the receiver.

2) We need to develop a measurement method to map chan-nel data, including the signal strength and SNR, to thelocation of the receiver, which will be used in our researchthat takes train locations into account to have a moreaccurate channel model.

A. Measurement Equipment

Two Cisco 3200 routers are used in our measurements. Oneis set as the AP and the other one is set as the MS. Both are setto work at the frequency of 2.412 GHz, which is also calledChannel 1. The output power of the AP is set as 30 dBm.The AP is located on the wall of the tunnel, whereas the MS islocated on the measurement vehicle. The transmitting antennais a Yagi antenna connected with the AP, which is directionaland vertically polarized. The half-power beamwidth (HPBW) is30◦ and the gain of the Yagi antenna is 13.5 dBi. In addition, theshark-fin antenna is applied as the receiving antenna connectedwith the MS, which is also directional and vertically polarized.The HPBW is 40◦ and the gain of the shark-fin antenna is10 dBi.

The location of the receiver is obtained through a velocitysensor installed on the wheel of the measurement vehicle,


WANG et al.: FINITE-STATE MARKOV MODELING FOR WIRELESS CHANNELS IN TUNNEL CBTC SYSTEMS 3

Fig. 2. Measurement equipment used in the real field CBTC channelmeasurements.

Fig. 3. Tunnel section and antenna deployment in the measurement.

which can detect the real-time velocity, and the resolution ofthe position is in millimeters per second. Fig. 2 shows themeasurement equipment used in our real field measurements.

B. Measurement Scenario

The measurements were performed in the Beijing SubwayChangping Line, China, where the cross section of the tunnelis rectangular. The cross section of the tunnel and locations ofantenna are shown in Fig. 3. The height of the tunnel is 4.91 mand the width is 4.4 m. The transmitting antenna is located0.15 m below the tunnel roof. The receiving antenna is seton the top of a measurement vehicle. As the threshold of thereceiver is −90 dBm, the coverage of one AP is about 0–500 m,which is also the experimental zone in our measurements. Thetunnel where we performed the measurement is a section of thestraight tunnel. Fig. 4 shows the cross section of the tunnel nearthe Nanshao station of the Beijing Subway Changping Line,

Fig. 4. (a) Tunnel where we performed the measurements in the BeijingSubway Changping Line. (b) Shark-fin antenna located on the measurementvehicle. (c) Yagi antenna. (d) AP set on the wall.

TABLE INOTIONS OF SYMBOLS

the shark-fin antenna, the Yagi antennas, and the AP set on thewall. We performed measurements in the tunnel of the BeijingSubway Changping Line 20 times so that enough data can becaptured.

IV. FSMC CHANNEL MODEL

To capture the characteristics of tunnel channels in CBTCsystems, we define channel states according to the differentreceived SNR levels and use an FSMC to track the statevariation. Here, we first describe the FSMC model, followedby the determination of key model parameters, including SNRlevels and SNR distribution. Table I illustrates the notions ofsymbols used in this paper.



A. FSMC Model

The SNR range of the received signal can be partitionedinto N nonoverlapping levels with thresholds {Γn, n = 0, 1, 2,3, . . . , N}, where Γ0 and ΓN can be measured. The time axis isdivided into slots of equal duration. Let γk denote the channelstate in time slot k. γk = sn when the SNR of the receivedsignal belongs to the range [Γn−1,Γn). Then, the received SNRcan be modeled as a random variable γ evolving according toan FSMC, and the transition probability pn,j can be shown as

pn,j = Pr{γk+1 = sn|γk = sj} (1)

where k = 1, 2, 3, . . ., and n, j ∈ {1, 2, . . . , N}.Based on the measurement data, we observe that most tran-

sitions within a Markov chain are between adjacent states.Therefore, we assume that each state can only transit to theadjacent states, which means pn,j = 0, if |n− j| > 1. With thedefinition, we can define a state transition probability matrix Pwith elements pn,j .

Due to the effect of large-scale fading, the amplitude ofthe SNR depends on the distance between the transmitterand the receiver. It is obvious that the SNR is usually highwhen the receiver is close to the transmitter, whereas it is lowwhen the receiver is far away from the transmitter. As a result,the transition probability from the high channel state to thelow channel state is different when the receiver is near or faraway from the transmitter, which means that the Markov statetransition probability is related to the location of the receiver.Therefore, only one state transition probability matrix, which isindependent of the location of the receiver, may not accuratelymodel the tunnel channels. Thus, we divide the tunnel into Lintervals and one state transition probability matrix is generatedfor each interval. Specifically, Pl, l ∈ {1, 2, . . . , L}, is the statetransition probability matrix corresponding to the lth intervaland the relationship between the transition probability and thelocation of the receiver can be built. Then, pln,j is the statetransition probability from state sn to state sj in the lth interval.Moreover, states n and j in the lth interval are denoted by slnand slj , respectively. Consequently, the state probabilities andthe state transition probabilities can be defined as

pln =P lr

{γlk = sln

}pln,j =P l

r

{γlk+1 = sln|γl

k = slj}

pln,j = 0, if |n− j| > 1

N∑j=1

pln,j = 1, ∀n ∈ {1, 2, 3, . . . , N} (2)

where pln is the probability of being in state n in the lth intervaland γl

k is the SNR level in time slot k in the lth interval.Based on the measurement results, we can determine the

value of state probability pln and state transition probabilitypln,j as

pln =aln

{γlk = sln

}∑N

n=1 aln

{γlk = sln

} (3)

where aln{γlk = sln} is the number of times state sn appears in

the lth interval, i.e.,

pln,j =aln,j

{γlk+1 = sln|γl

k = slj}

∑Nj=1 a

ln,j

{γlk+1 = sln|γl

k = slj} (4)

where aln,j{γlk+1 = sln|γl

k = slj} is the number of times state jtransits to state n in the lth interval.

B. Determine the SNR Level Thresholds of the FSMC Model

Determining the thresholds of SNR levels is the key factorthat affects the accuracy of the FSMC model. There are manymethods to select the SNR level boundaries, among which theequiprobable partition method is frequently used in previousworks [11]–[13]. As nonuniform amplitude partitioning can beuseful to obtain more accurate estimates of system performancemeasures [18], we choose the Lloyd–Max technique [17] in-stead of the equiprobable method to partition the amplitude ofthe SNR in this paper. Lloyd–Max is an optimized quantizer,which can decrease the distortion of scalar quantization.

First, distortion function D is defined as

D =

N∑n=1

Γn∫

Γn−1

f(Γ̃n − γ)p(γ) dγ (5)

where Γ̃n is the quantized value of the SNR whose amplitudeis in the range [Γn−1Γn), f(x) is the error criterion function,and p(γ) is the probability distribution function of the SNR.The distortion function can be minimized through optimallyselecting Γ̃n and Γn.

Then, the necessary conditions for minimum distortion areobtained by differentiating D with respect to Γn and Γ̃n. Theresult of this minimization is a pair of equations [19]

f(Γ̃n − Γn) = f(Γ̃n+1 − Γn) (6)

Γn∫

Γn−1

f ′(Γ̃n − γ)p(γ) dγ = 0. (7)

The error criterion function f(x) is often taken as x2 [19].As a result, (6) and (7) become

Γn =Γ̃n + Γ̃n+1

2(8)

Γn∫

Γn−1

(Γ̃n − γ)p(γ) dγ = 0. (9)

As mentioned above, we partition the amplitude of the SNRinto N levels and there are N + 1 corresponding thresholds{Γn, n = 0, 1, 2, 3, . . . , N}. Generally, the first and last thresh-olds are known, which are denoted by the minimum andmaximum measurement values of the SNR. Furthermore, theLloyd–Max algorithm is used to divide 2r levels, which meansN = 2r, r = 1, 2, 3, . . ., and N is an even number. As a result,since Γ0 and ΓN are known, Γ(N/2) can be obtained from (9).



Then, Γ(N/4) and Γ(3N/4) can also be calculated according tothe new variable Γ(N/2), when r is larger than 2. With theprocess being repeated, all elements of {Γn} can be obtained as

Γb∫

Γa

(Γ̃ b+a

2− γ

)p(γ) dγ = 0

b > a and b ∈ {2, 3, . . . , N}, a ∈ {1, 2, 3, . . . , N − 1} (10)

where p(γ) is the probability distribution function of the SNR.According to the calculated {Γn}, combined with (8) and

(9), we can update the value of {Γn} until the value of D isthe minimum, and the optimal thresholds of the SNR levels canbe obtained. As p(γ) is still unknown, we should discuss thedistribution of the SNR in the following subsection accordingto the real field measurement data, which is the last step inobtaining the thresholds of SNR levels.

C. Determine the Distribution of SNR

Deriving the distribution of the SNR is the crucial step inpartitioning the levels of the SNR. In fact, there are some classicmodels to describe the distribution of signal strength, such asRice, Rayleigh, and Nakagami, and then the correspondingmodels of the SNR can also be obtained [20]. We first derivethe distribution of the signal strength in order to determinethe distribution model of the SNR. According to [20] and[21], voltage/meter (v/m) is usually used when studying thedistribution of small-scale fading. Therefore, we convert dBmto the linear unit v/m following [22].

The Akaike information criterion (AIC) is often used to getthe approximate distribution model of the signal strength [23].The AIC is a measure of the relative goodness of fit of astatistical model. It was developed by Hirotsugu Akaike, underthe name of an information criterion, and was first published in1974. The general case of AIC is

AIC = −2 lnLm + 2η (11)

where η is the number of parameters of the statistical modeland Lm is the maximized value of the likelihood function forthe estimated model. In fact, according to the relationship of ηand the number of samples ns, AIC needs to be changed to AICwith a correction (AICc) when ns/η < 40 [23], i.e.,

AICc = AIC +2η(η + 1)ns − η − 1

. (12)

In this paper, AICc is adopted to estimate the model of the sig-nal strength distribution instead of the classic AIC. In practice,one can compute AICc for each of the candidate models andselect the model with the smallest value of AICc. The candidatemodels in this paper include Rice, Rayleigh, and Nakagami.

Since our channel model is related to the distance betweenthe transmitter and the receiver, the tunnel should be dividedinto intervals. Assume that there are L intervals and we applyAICc for each candidate model in every interval. As a result, wecan select the most appropriate model based on the frequencyof the minimum AICc value of different candidate models. Inorder to obtain enough data for each interval and ensure theaccuracy of the model, we set the length of each interval as

Fig. 5. Frequencies of AICc selecting a candidate distribution.

40 wavelengths of WLANs [21], and then there are 100 inter-vals. Based on the frequencies of AICc of different distributionsin the real field measurements, we observe that the Nakagamidistribution provides the best fit compared with Rayleigh andRicean distributions, as shown in Fig. 5. As a result, we candefine p(γ) as the Nakagami distribution.

After the distribution of the signal strength is obtained,according to [20], we can derive the distribution of the SNR.Thus

p(γl) =μμll γμl−1

l

γ̄μll Γ(μl)

e(−μlγlγ̄l ) (13)

where γl is the SNR of the received signal in the lth interval,γ̄l is the mean of the SNR in the lth interval, μl is the fadingfactor of Nakagami distribution in the lth interval, and Γ(·) isthe gamma function. In fact, μl can be calculated when apply-ing AICc through the maximum-likelihood estimator for eachinterval.

V. REAL FIELD MEASUREMENT RESULTS

AND DISCUSSIONS

Here, we compare our FSMC model with real field testresults to illustrate the accuracy of the model. The effects ofdifferent parameters in the proposed model are discussed. Thenumber of states in our model is first set to four. We also useeight states to study the effects of the number of states on theaccuracy of the proposed model. In order to obtain the effectsof distance intervals on the model, we choose the intervals as5, 10, 20, 50, and 100 m. We perform measurements in thetunnel of the Beijing Subway Changping Line 20 times so thatenough data can be captured. The accuracy of the FSMC modelis verified through another set of measurement data.

Based on the measurement data, (8)–(10), and (13), we derivethe thresholds {Γn, n = 0, 1, 2, . . . , N} of the SNR in eachdistance interval. Tables II and III demonstrate the thresholdsof the SNR levels at the location of 100 m for differentintervals, where we divide the SNR into four and eight levels.As the distance intervals are different, the range of the SNR isdifferent, and it brings different thresholds, which can providea more accurate model.



TABLE IITHRESHOLDS OF SNR LEVELS (FOUR LEVELS) AT THE

LOCATION OF 100 m FOR DIFFERENT INTERVALS

TABLE IIITHRESHOLDS OF SNR LEVELS (EIGHT LEVELS) AT THE

LOCATION OF 100 m FOR DIFFERENT INTERVALS

TABLE IVSTATE TRANSITION PROBABILITIES OF THE FSMC MODEL

AND THE MEASUREMENT DATA WITH FOUR STATES AND

5-m INTERVAL AT THE LOCATION (35–40 m)

After we get the thresholds, we can get the state probabilitiesand the state transition probabilities from the real field data.Table IV illustrates the state transition probabilities of theFSMC model and the measurement data at the same location(35–40 m), when there are four states and the distance intervalis 5 m. We can observe that the sum of the transition probabilityof each channel state is not equal to 1. This is because, inthe measurement data, there are some state transitions that donot happen in adjacent states, such as transitions from state1 to state 3. However, in our FSMC models, for the sake ofsimplicity, we assume that states can only transit to the adjacentstates. Therefore, in Table IV, we only consider state transitionsbetween adjacent states. Consequently, the sum of the transitionprobability of each state of the measurement data is not equalto 1. Nevertheless, it is very close to 1, which means that ourassumption is realistic for practical tunnel channels in CBTCsystems.

Fig. 6 shows the simulation results generated from our FSMCmodel and the experimental results from real field measure-ments. We can observe the great agreement between them.Next, we derive the mean square error (MSE) to measure thedegrees of approximation, as shown in Fig. 7, where the y-axisis the MSE between the FSMC results and the measurementresults, and the x-axis is the interval distance (5, 10, 20, 50,and 100 m). As shown in Fig. 7, when the distance intervalincreases, the MSE also increases, which means that the accu-racy of the model decreases. Moreover, we can also observe

Fig. 6. Simulation results generated from the FSMC model and experimentalresults from real field measurements.

Fig. 7. MSE between the FSMC model and the experimental data with fourstates and eight states.

that the MSE of the FSMC model with four states is larger thanthat with eight states. The number of states in the FSMC modelplays a key role in the accuracy. Nevertheless, when the distanceinterval is 5 m, the MSE difference is small for the four-stateFSMC model (0.032) and the eight-state FSMC model (0.028).In this figure, we can see that the FSMC model with four statesand the 5-m distance interval can provide an accurate enoughchannel model for tunnel channels in CBTC systems.

VI. CONCLUSION AND FUTURE WORK

Modeling the tunnel wireless channels of urban rail transitsystems is important in designing the wireless networks andevaluating the performance of CBTC systems. In this paper, wehave proposed an FSMC model for tunnel channels in CBTCsystems. Since the train location is known in CBTC systems,the proposed FSMC channel model takes train locations into ac-count to have a more accurate channel model. The distance be-tween the transmitter and the receiver is divided into intervals,and an FSMC model is designed in each interval. The accuracyof the proposed model has been illustrated by the simulationresults generated from the proposed model and the real field



measurements. In addition, we have shown that the number ofstates and the distance interval have impacts on the accuracyof the proposed FSMC model. Future work is in progress tostudy the effects of wireless channels on the control perfor-mance of CBTC systems based on the proposed channel model.

REFERENCES

[1] R. Pascoe and T. Eichorn, “What is communication-based train control?”IEEE Veh. Tech. Mag., vol. 4, no. 4, pp. 16–21, Dec. 2009.

[2] S. Su, X. Li, T. Tang, and Z. Gao, “A subway train timetable optimiza-tion approach based on energy-efficient operation strategy,” IEEE Trans.Intell. Transp. Syst., vol. 14, no. 2, pp. 883–893, Jun. 2013.

[3] Standard for Communications-based Train Control (CBTC) Performanceand Functional requirements, IEEE Std 1474.1-2004 (Revision of IEEEStd 1474.1-1999), 2004, 0_1–45.

[4] L. Zhu, F. R. Yu, B. Ning, and T. Tang, “Cross-layer handoff design inMIMO-enabled WLANs for communication-based train control (CBTC)systems,” IEEE J. Sel. Areas Commun., vol. 30, no. 4, pp. 719–728,May 2012.

[5] Y. Zhang, “Novel model for propagation loss prediction in tunnels,” IEEETrans. Veh. Tech., vol. 52, no. 5, pp. 1308–1314, Sep. 2003.

[6] K. Guan, Z. Zhong, J. Alonso, and C. Briso-Rodriguez, “Measurementof distributed antenna systems at 2.4 GHz in a realistic subway tun-nel environment,” IEEE Trans. Veh. Tech., vol. 61, no. 2, pp. 834–837,Feb. 2012.

[7] H.-P. Lin and M.-J. Tseng, “Two-layer multistate Markov model for mod-eling a 1.8 GHz narrow-band wireless propagation channel in urban Taipeicity,” IEEE Trans. Veh. Tech., vol. 54, no. 2, pp. 435–446, Mar. 2005.

[8] S. Lin, Z. Zhong, L. Cai, and Y. Luo, “Finite state Markov modellingfor high speed railway wireless communication channel,” in Proc. IEEEGlobecom, Anaheim, CA, USA, 2012, pp. 5421–5426.

[9] F. Babich, G. Lombardi, and E. Valentinuzzi, “Variable order Markovmodeling for LEO mobile satellite channels,” Electron. Lett., vol. 35,no. 8, pp. 621–623, Apr. 1999.

[10] F. Babich and G. Lombardi, “A measurement based Markov model forthe indoor propagation channel,” in Proc. IEEE VTC, Phoenix, AZ, USA,May 1997, vol. 1, pp. 77–81.

[11] H. S. Wang and N. Moayeri, “Finite-state Markov channel—A use-ful model for radio communication channels,” IEEE Trans. Veh. Tech.,vol. 44, no. 1, pp. 163–171, Feb. 1995.

[12] C. Pimentel, T. Falk, and L. Lisboa, “Finite-state Markov modeling ofcorrelated Rician-fading channels,” IEEE Trans. Veh. Tech., vol. 53, no. 5,pp. 1491–1501, Sep. 2004.

[13] C. Iskander and P. Mathiopoulos, “Fast simulation of diversity Nakagamifading channels using finite-state Markov models,” IEEE Trans. Broad-casting, vol. 49, no. 3, pp. 269–277, Sep. 2003.

[14] A. Goldsmith and P. Varaiya, “Capacity, mutual information, and codingfor finite-state Markov channels,” IEEE Trans. Inf. Theory, vol. 42, no. 3,pp. 868–886, May 1996.

[15] A. Chockalingam, M. Zorzi, L. Milstein, and P. Venkataram, “Perfor-mance of a wireless access protocol on correlated Rayleigh-fading chan-nels with capture,” IEEE Trans. Commun., vol. 46, no. 5, pp. 644–655,May 1998.

[16] F. Babich, O. Kelly, and G. Lombardi, “Generalized Markov modelingfor flat fading,” IEEE Trans. Commun., vol. 48, no. 4, pp. 547–551,Apr. 2000.

[17] S. Lloyd, “Least squares quantization in PCM,” IEEE Trans. Inf. Theory,vol. 28, no. 2, pp. 129–137, Mar. 1982.

[18] P. Sadeghi, R. Kennedy, P. Rapajic, and R. Shams, “Finite-state Markovmodeling of fading channels—A survey of principles and applications,”IEEE Signal Process. Mag., vol. 25, no. 5, pp. 57–80, Sep. 2008.

[19] J. G. Proakis, Digital Communications. New York, NY, USA: McGraw-Hill, 1995.

[20] M. K. Simon and M.-S. Alouini, Digital Communication over FadingChannels.. Hoboken, NJ, USA: Wiley, 2005.

[21] S. Wyne, A. Singh, F. Tufvesson, and A. Molisch, “A statistical model forindoor office wireless sensor channels,” IEEE Trans. Wireless Commun.,vol. 8, no. 8, pp. 4154–4164, Aug. 2009.

[22] T. S. Rappaport, Wireless Communications Principles and Practice,Second ed. Beijing, China: Publishing House of Electronics Industry,2008.

[23] K. P. Burnham and D. R. Anderson, Model Selection and Multi-modelInference: A Practical Information-Theoretic Approach. New York, NY,USA: Springer-Verlag, 2002.

Hongwei Wang received the B.S. degree in elec-tronics and information engineering from BeijingJiaotong University, Beijing, China, in 2008, wherehe is currently working toward the Ph.D. degree withthe State Key Laboratory of Rail Traffic Control andSafety.

His research interests include train–ground com-munication technology in communication base train–ground communication systems and cooperativescheduling approaches in subway systems.

Fei Richard Yu (S’00–M’04–SM’08) received thePh.D. degree in electrical engineering from the Uni-versity of British Columbia, Vancouver, BC, Canada,in 2003.

From 2002 to 2004, he was with Ericsson (inLund, Sweden), where he worked on the research anddevelopment of wireless mobile systems. From 2005to 2006, he was with a startup in California, USA,where he worked on the research and developmentin the areas of advanced wireless communicationtechnologies and new standards. In 2007, he joined

the Carleton School of Information Technology and the Department of Systemsand Computer Engineering, Carleton University, Ottawa, ON, Canada, wherehe is currently an Associate Professor. His research interests include cross-layer design, security, green IT, and quality-of-service provisioning in wirelessnetworks.

Dr. Yu was a recipient of the IEEE Outstanding Leadership Award in 2013;the Carleton Research Achievement Award in 2012; the Ontario Early Re-searcher Award (formerly Premier’s Research Excellence Award) in 2011; theExcellent Contribution Award at the 2010 IEEE/IFIP International Symposiumon trusted Computing and Communications (IEEE/IFIP TrustCom 2010); theLeadership Opportunity Fund Award from Canada Foundation of Innovationin 2009; and the Best Paper Awards at the 2012 IEEE Global Communi-cations Conference (GLOBECOM 2012), IEEE/IFIP TrustCom 2009, andInternational Conference on Networking 2005. He serves on the editorialboards of several journals, including IEEE TRANSACTIONS ON VEHICULAR

TECHNOLOGY, IEEE Communications Surveys and Tutorials, ACM/SpringerWireless Networks, EURASIP Journal on Wireless Communications Network-ing, Ad Hoc and Sensor Wireless Networks, Wiley Journal on Security andCommunication Networks, and International Journal of Wireless Communica-tions and Networking. He is also a Guest Editor for IEEE Systems Journal forthe special issue on Smart Grid Communications Systems. He has served on theTechnical Program Committee (TPC) of numerous conferences, as the TPC Co-Chair of the IEEE GLOBECOM 2014, the IEEE INFOCOM 2014 Workshopon Mobile Cloud Computing (INFOCOM-MCC 2014), GLOBECOM 2013,the 2013 International Conference on Green Computing and Communications(GreenCom 2013), the 2013 IEEE Consumer Communications and NetworkingConference (CCNC 2013), the IEEE INFOCOM 2012 Workshop on GreenNetworking and Smart Grids (INFOCOM-CCSES 2012), the IEEE ICC 2012Workshop on Green Communications and Networking (ICC-GCN 2012), the2012 Vehicular Technology Conference (VTC Spring 2012), GLOBECOM2011, INFOCOM-GCN 2011, the IEEE INFOCOM 2010 Workshop on Cog-nitive Wireless Communications and Networking (INFOCOM-CWCN 2010),the 2009 IEEE Wireless Communications and Mobile Computing Conference(IEEE IWCMC 2009), VTC Fall 2008, and the 1st International Workshop onWireless Networking for Intelligent Transportation Systems (WiN-ITS 2007);as the Publication Chair of the 7th International ICST Conference on Heteroge-neous Networking for Quality Reliability and Robustness (ICST Qshine 2010);and as the Co-Chair of the IEEE ICUMT Workshop on Cognitive WirelessCommunications and Networking (ICUMT-CWCN 2009).

Li Zhu received the B.S. and Ph.D. degrees in elec-tronics and information engineering from BeijingJiaotong University, Beijing, China, in 2006 and2012, respectively.

He was a Visiting Ph.D. Student with CarletonUniversity, Ottawa, ON, Canada, and the Univer-sity of British Columbia, Vancouver, BC, Canada.He is currently a Faculty Member with BeijingJiaotong University. His research interests in-clude train-ground communication technology incommunication-based train-ground communication

(CBTC) systems and cross-layer design in train-ground communicationsystems.



Tao Tang received the Ph.D. of Engineering degreefrom the Chinese Academy of Sciences, Beijing,China, in 1991.

He is currently a Professor with the Schoolof Traffic and Transportation, Beijing JiaotongUniversity, Beijing, China, where he is the AssociateDirector of the State Key Laboratory of Rail TrafficControl and Safety. His research interests includecommunication-based train control, high-speedtrain control systems, and intelligent transportationsystems.

Dr. Tang is a member of the Experts Group of the High Technology Researchand Development Program of China (863 Program), and he undertakes as theLeader of the Field of Modern Transportation Technology Experts Group. Heis also a Specialist of the National Development and Reform Commission andthe Beijing Urban Traffic Construction Committee.

Bin Ning (M’94–SM’12) received the B.S., M.S.,and Ph.D. of Engineering degrees in 1982, 1987,and 2005, respectively, from the Northern JiaotongUniversity, which has been called Beijing JiaotongUniversity since 2003.

He was a Visiting Scholar with Brunel Univer-sity, London, UK, studying electronics and elec-trical power engineering. From October 2002 toFebruary 2003, he joined University of California,Berkeley, CA, USA, as a Senior Visiting Scholar.He is currently a Professor with Beijing Jiaotong

University. His research mainly focus on high-speed train control systemsand railway transportation train control systems, including main locomotivesignals, communication-based train control systems, intelligent transportation,fault-tolerant design of signal systems, fault diagnosis, system reliability, andsecurity design.

Dr. Ning is a Fellow of the Association of International Railway SignalingEngineers (IRSE), The Institute of Engineering and Technology, and the ChinaRailway Society. He is the Deputy Director of the China Traffic SystemEngineering Society and the Beijing Railway Society. He is a member ofthe China Overseas Returned Scholars Association and the editorial board ofthe Journal of Railways in China. He served as the Chair of the TechnicalCommittee on Railroad Systems and Applications of the IEEE IntelligentTransportation Systems Society. He also served as an Associate Editor of theIEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS, in2010–2012, and of Acta Automatica Sinica, in 2011-2012.

ieeetransactions on intelligent transportation …...since cbtc systems are mostly deployed in...

Documents