IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS 1

Transcription

1 IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS 1 Finite-State Markov Modeing for Wireess Channes in Tunne Communication-Based Train Contro Systems Hongwei Wang, Fei Richard Yu, Senior Member, IEEE, Li Zhu, Tao Tang, and Bin Ning, Senior Member, IEEE Abstract Communication-based train contro (CBTC) is being rapidy adopted in urban rai transit systems, as it can significanty enhance raiway network efficiency, safety, and capacity. Since CBTC systems are mosty depoyed in underground tunnes and trains move at high speeds, buiding a train ground wireess communication system for CBTC is a chaenging task. Modeing the tunne channes is very important in designing the wireess networks and evauating the performance of CBTC systems. Most existing works on channe modeing do not consider the unique characteristics of CBTC systems, such as high mobiity speed, deterministic moving direction, and accurate train-ocation information. In this paper, we deveop a finite-state Markov channe (FSMC) mode for tunne channes in CBTC systems. The proposed FSMC mode is based on rea fied CBTC channe measurements obtained from a business-operating subway ine. Unike most existing channe modes, which are not reated to specific ocations, the proposed FSMC channe mode takes train ocations into account to have a more accurate channe mode. The distance between the transmitter and the receiver is divided into intervas and an FSMC mode is appied in each interva. The accuracy of the proposed FSMC mode is iustrated by the simuation resuts generated from the mode and the rea fied measurement resuts. Index Terms Communication-based train contro (CBTC), finite-state Markov chain (FSMC), wireess oca area network (WLAN). I. INTRODUCTION URBAN rai transit systems are rapidy deveoping around the word. Due to a great dea of urban traffic pressure, improving the efficiency and capacity of urban rai transit systems is an increasing demand. Because it is a key subsystem Manuscript received May 13, 2013; revised Juy 30, 2013 and October 17, 2013; accepted November 26, This work was supported in part by the Nationa Natura Science Foundation of China under Grant ; by the Nationa High Technoogy Research and Deveopment Program of China (863 Program) under Grant 2011AA110502; by the Doctora Program of the Ministry of Education under Grant ; by the Foundation of Beijing Scientific Committee under Grant D ; by the Key Project of the Beijing Laboratory of Urban Rai Transit and the Beijing Key Laboratory of Urban Rai Transit Automation and Contro under Grants RCS2012K010, RCS2012ZQ002, RCS2011ZT010, and 2011JBZ014; by the China Education Ministry Funding Project under Grant 2013JBM124; and by the China Schoarship Counci. The Associate Editor for this paper was L. Vacic. H. Wang, L. Zhu, T. Tang, and B. Ning are with State Key Laboratory of Rai Traffic Contro and Safety, Beijing Jiaotong University, Beijing , China. F. R. Yu is with the Department of Systems and Computer Engineering, Careton University, Ottawa, ON K1S 5B6, Canada. Coor versions of one or more of the figures in this paper are avaiabe onine at Digita Object Identifier /TITS of urban rai transit systems, communication-based train contro (CBTC) is an automated train contro system that makes use of train ground wireess communications to ensure the safe operation of rai vehices [1]. It can enhance the eve of safety and service offered to customers and improve the utiization of raiway network infrastructure [2]. CBTC is a modern successor to the traditiona raiway signaing system using interockings, track circuits, and signas [3]. Buiding a train ground wireess communication system for CBTC is a chaenging task. As urban rai transit systems are mosty depoyed in underground tunnes, there are arge amounts of refections, scattering, and barriers that severey affect the propagation performance of wireess communications. Moreover, due to the avaiabe commercia off-the-shef equipment, wireess oca area networks (WLANs) are often adopted as the main method of train ground communications for CBTC systems. However, most of the current IEEE WLAN standards were not originay designed for the high-speed environment in tunnes [4]. Furthermore, the fast movement of trains wi cause frequent handoffs between WLAN access points (APs), which can severey affect CBTC performance. Modeing the channes of urban rai transit systems is very important in designing the wireess networks and evauating the performance of CBTC systems. There are some previous works on radio wave propagation in urban rai transit systems. A path oss mode of tunne channes is given in [5], which describes the characteristics of arge-scae fading. Guan et a. in [6] presented the propagation characteristics based on rea environment measurements in the Madrid, Spain, subway. A twoayer mutistate Markov mode is presented in [7] for modeing a 1.8-GHz channe in urban Taipei City, Taiwan. Based on the Winner II physica-ayer channe mode parameters, Lin et a. in [8] proposed a channe mode for a high-speed raiway. Athough some exceent works have been done on modeing channes, most do not consider the unique characteristics in CBTC systems, such as high mobiity speed, deterministic moving direction, and accurate train-ocation information. In this paper, we deveop a finite-state Markov channe (FSMC) mode for tunne channes in CBTC systems. FSMC modes have been widey accepted in the iterature as an effective approach to characterizing the correation structure of the fading process, incuding 1.8-GHz narrow-band channes [7], high-speed raiway channes [8], sateite channes [9], indoor channes [10], Rayeigh fading channes [11], Ricean fading channes [12], and Nakagami fading channes [13]. Using IEEE. Persona use is permitted, but repubication/redistribution requires IEEE permission. See for more information. Downoaded from

2 2 IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS Fig. 1. CBTC system. FSMC modes, a variety of anaytica resuts of system performance can be derived, incuding channe capacity [14], throughput [15], and packet error distribution [16]. To the best of our knowedge, FSMC modes for tunne channes in CBTC systems have not been studied in previous works. Therefore, there is a strong motivation to deveop an FSMC mode for tunne channes in CBTC systems. Some distinct features of the proposed channe mode are as foows. The proposed FSMC mode is based on rea fied CBTC channe measurements obtained from the businessoperating Beijing Subway Changping Line. Unike most existing channe modes, which do not use train-ocation information, the proposed FSMC channe mode takes train ocations into account to create a more accurate channe mode. The distance between the transmitter and the receiver is divided into intervas, and the FSMC mode is appied in each interva. The Loyd Max technique [17] is used to determine the signa-to-noise ratio (SNR) eve boundaries in the proposed FSMC mode. The accuracy of the proposed FSMC mode is iustrated by the simuation resuts generated from the mode and the rea fied measurement resuts. The effects of different parameters are aso discussed. The rest of this paper is organized as foows. Section II describes an overview of CBTC systems. In Section III, the rea fied measurement configuration and scenario are described. In Section IV, the FSMC mode is introduced. Then, Section V presents the rea fied measurement resuts and discussions. Finay, this paper is concuded in Section VI with future work. II. OVERVIEW OF CBTC Fig. 1 describes a CBTC system. In this system, continuous bidirectiona wireess communications between each mobie station (MS) on the train and the wayside APs are adopted instead of the traditiona fixed-bock track circuit. The raiway ine is usuay divided into areas or regions. Each area is under the contro of a zone controer (ZC) and has its own radio transmission system. Each train transmits its identity, ocation, direction, and speed to the ZC. The radio ink between each train and the ZC shoud be continuous so that the ZC knows the ocations of a the trains in its area a the time. The ZC transmits to each train the ocation of the train in front of it and gives it a braking curve to enabe it to stop before it reaches that train. Theoreticay, as ong as each train is traveing at the same speed and they a have the same braking capabiity, they can trave together as cosey as a few meters between them. When a train moves away from the coverage of an AP and enters the coverage of another AP aong the raiway, the handoff procedure may resut in communication interruption and ong atency. In CBTC systems, it is important to maintain communication ink avaiabiity in order to guarantee train operation safety and efficiency. Wireess channes in CBTC systems are different from those in other wireess systems, since most CBTC systems are depoyed in underground tunnes, where there are arge amounts of refections, scattering, and barriers that severey affect the propagation performance of wireess communications. In order to design the wireess networks and evauate the performance of CBTC systems, modeing the tunne channes in CBTC systems shoud be carefuy studied. III. REAL FIELD CBTC CHANNEL MEASUREMENTS The objective of the rea fied CBTC channe measurements is to get the rea fied data of WLAN propagation in tunnes under rea conditions of the subway ine, which wi be used to buid an FSMC mode. Here, we present the measurement equipment and the measurement scenario in our rea fied CBTC channe measurements. With this objective, the preparation of the measurements consists of two parts as foows. 1) We need to make sure that the configuration of the measurements is the same as business-operating subway ines, incuding the choice of antennas, the ocation of antennas, and the settings of the transmitter and the receiver. 2) We need to deveop a measurement method to map channe data, incuding the signa strength and SNR, to the ocation of the receiver, which wi be used in our research that takes train ocations into account to have a more accurate channe mode. A. Measurement Equipment Two Cisco 3200 routers are used in our measurements. One is set as the AP and the other one is set as the MS. Both are set to work at the frequency of GHz, which is aso caed Channe 1. The output power of the AP is set as 30 dbm. The AP is ocated on the wa of the tunne, whereas the MS is ocated on the measurement vehice. The transmitting antenna is a Yagi antenna connected with the AP, which is directiona and verticay poarized. The haf-power beamwidth (HPBW) is 30 and the gain of the Yagi antenna is 13.5 dbi. In addition, the shark-fin antenna is appied as the receiving antenna connected with the MS, which is aso directiona and verticay poarized. The HPBW is 40 and the gain of the shark-fin antenna is 10 dbi. The ocation of the receiver is obtained through a veocity sensor instaed on the whee of the measurement vehice,

3 WANG et a.: FINITE-STATE MARKOV MODELING FOR WIRELESS CHANNELS IN TUNNEL CBTC SYSTEMS 3 Fig. 2. Measurement equipment used in the rea fied CBTC channe measurements. Fig. 4. (a) Tunne where we performed the measurements in the Beijing Subway Changping Line. (b) Shark-fin antenna ocated on the measurement vehice. (c) Yagi antenna. (d) AP set on the wa. TABLE I NOTIONS OF SYMBOLS Fig. 3. Tunne section and antenna depoyment in the measurement. which can detect the rea-time veocity, and the resoution of the position is in miimeters per second. Fig. 2 shows the measurement equipment used in our rea fied measurements. B. Measurement Scenario The measurements were performed in the Beijing Subway Changping Line, China, where the cross section of the tunne is rectanguar. The cross section of the tunne and ocations of antenna are shown in Fig. 3. The height of the tunne is 4.91 m and the width is 4.4 m. The transmitting antenna is ocated 0.15 m beow the tunne roof. The receiving antenna is set on the top of a measurement vehice. As the threshod of the receiver is 90 dbm, the coverage of one AP is about m, which is aso the experimenta zone in our measurements. The tunne where we performed the measurement is a section of the straight tunne. Fig. 4 shows the cross section of the tunne near the Nanshao station of the Beijing Subway Changping Line, the shark-fin antenna, the Yagi antennas, and the AP set on the wa. We performed measurements in the tunne of the Beijing Subway Changping Line 20 times so that enough data can be captured. IV. FSMC CHANNEL MODEL To capture the characteristics of tunne channes in CBTC systems, we define channe states according to the different received SNR eves and use an FSMC to track the state variation. Here, we first describe the FSMC mode, foowed by the determination of key mode parameters, incuding SNR eves and SNR distribution. Tabe I iustrates the notions of symbos used in this paper.

4 4 IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS A. FSMC Mode The SNR range of the received signa can be partitioned into N nonoverapping eves with threshods {Γ n,n= 0, 1, 2, 3,...,N}, where Γ 0 and Γ N can be measured. The time axis is divided into sots of equa duration. Let γ k denote the channe state in time sot k. γ k = s n when the SNR of the received signa beongs to the range [Γ n 1, Γ n ). Then, the received SNR can be modeed as a random variabe γ evoving according to an FSMC, and the transition probabiity p n,j can be shown as p n,j = P r {γ k+1 = s n γ k = s j } (1) where k = 1, 2, 3,..., and n, j {1, 2,...,N}. Based on the measurement data, we observe that most transitions within a Markov chain are between adjacent states. Therefore, we assume that each state can ony transit to the adjacent states, which means p n,j = 0, if n j > 1. With the definition, we can define a state transition probabiity matrix P with eements p n,j. Due to the effect of arge-scae fading, the ampitude of the SNR depends on the distance between the transmitter and the receiver. It is obvious that the SNR is usuay high when the receiver is cose to the transmitter, whereas it is ow when the receiver is far away from the transmitter. As a resut, the transition probabiity from the high channe state to the ow channe state is different when the receiver is near or far away from the transmitter, which means that the Markov state transition probabiity is reated to the ocation of the receiver. Therefore, ony one state transition probabiity matrix, which is independent of the ocation of the receiver, may not accuratey mode the tunne channes. Thus, we divide the tunne into L intervas and one state transition probabiity matrix is generated for each interva. Specificay, P, {1, 2,...,L}, is the state transition probabiity matrix corresponding to the th interva and the reationship between the transition probabiity and the ocation of the receiver can be buit. Then, p n,j is the state transition probabiity from state s n to state s j in the th interva. Moreover, states n and j in the th interva are denoted by s n and s j, respectivey. Consequenty, the state probabiities and the state transition probabiities can be defined as p n = Pr { γ k = s } n p n,j = Pr { γ k+1 = s n γk = s } j p n,j = 0, if n j > 1 N p n,j = 1, n {1, 2, 3,...,N} (2) j=1 where p n is the probabiity of being in state n in the th interva and γk is the SNR eve in time sot k in the th interva. Based on the measurement resuts, we can determine the vaue of state probabiity p n and state transition probabiity p n,j as { } p a n γ n = k = s n N { (3) n=1 a n γ k = sn} where a n{γk = s n} is the number of times state s n appears in the th interva, i.e., { p a n,j γ k+1 = s n γk n,j = = } s j N { j=1 a n,j γ k+1 = s n γk = (4) j} s where a n,j {γ k+1 = s n γk = s j } is the number of times state j transits to state n in the th interva. B. Determine the SNR Leve Threshods of the FSMC Mode Determining the threshods of SNR eves is the key factor that affects the accuracy of the FSMC mode. There are many methods to seect the SNR eve boundaries, among which the equiprobabe partition method is frequenty used in previous works [11] [13]. As nonuniform ampitude partitioning can be usefu to obtain more accurate estimates of system performance measures [18], we choose the Loyd Max technique [17] instead of the equiprobabe method to partition the ampitude of the SNR in this paper. Loyd Max is an optimized quantizer, which can decrease the distortion of scaar quantization. First, distortion function D is defined as D = N n=1 Γ n Γ n 1 f( Γ n γ)p(γ) dγ (5) where Γ n is the quantized vaue of the SNR whose ampitude is in the range [Γ n 1 Γ n ), f(x) is the error criterion function, and p(γ) is the probabiity distribution function of the SNR. The distortion function can be minimized through optimay seecting Γ n and Γ n. Then, the necessary conditions for minimum distortion are obtained by differentiating D with respect to Γ n and Γ n.the resut of this minimization is a pair of equations [19] f( Γ n Γ n )=f( Γ n+1 Γ n ) (6) Γ n Γ n 1 f ( Γ n γ)p(γ) dγ = 0. (7) The error criterion function f(x) is often taken as x 2 [19]. As a resut, (6) and (7) become Γ n = Γ n + Γ n+1 2 Γ n (8) Γ n 1 ( Γ n γ)p(γ) dγ = 0. (9) As mentioned above, we partition the ampitude of the SNR into N eves and there are N + 1 corresponding threshods {Γ n,n= 0, 1, 2, 3,...,N}. Generay, the first and ast threshods are known, which are denoted by the minimum and maximum measurement vaues of the SNR. Furthermore, the Loyd Max agorithm is used to divide 2 r eves, which means N = 2 r,r = 1, 2, 3,..., and N is an even number. As a resut, since Γ 0 and Γ N are known, Γ (N/2) can be obtained from (9).

5 WANG et a.: FINITE-STATE MARKOV MODELING FOR WIRELESS CHANNELS IN TUNNEL CBTC SYSTEMS 5 Then, Γ (N/4) and Γ (3N/4) can aso be cacuated according to the new variabe Γ (N/2), when r is arger than 2. With the process being repeated, a eements of {Γ n } can be obtained as Γ b ( Γ b+a 2 ) γ p(γ) dγ = 0 Γ a b>a and b {2, 3,...,N},a {1, 2, 3,...,N 1} (10) where p(γ) is the probabiity distribution function of the SNR. According to the cacuated {Γ n }, combined with (8) and (9), we can update the vaue of {Γ n } unti the vaue of D is the minimum, and the optima threshods of the SNR eves can be obtained. As p(γ) is sti unknown, we shoud discuss the distribution of the SNR in the foowing subsection according to the rea fied measurement data, which is the ast step in obtaining the threshods of SNR eves. C. Determine the Distribution of SNR Deriving the distribution of the SNR is the crucia step in partitioning the eves of the SNR. In fact, there are some cassic modes to describe the distribution of signa strength, such as Rice, Rayeigh, and Nakagami, and then the corresponding modes of the SNR can aso be obtained [20]. We first derive the distribution of the signa strength in order to determine the distribution mode of the SNR. According to [20] and [21], votage/meter (v/m) is usuay used when studying the distribution of sma-scae fading. Therefore, we convert dbm to the inear unit v/m foowing [22]. The Akaike information criterion (AIC) is often used to get the approximate distribution mode of the signa strength [23]. The AIC is a measure of the reative goodness of fit of a statistica mode. It was deveoped by Hirotsugu Akaike, under the name of an information criterion, and was first pubished in The genera case of AIC is AIC = 2 n L m + 2η (11) where η is the number of parameters of the statistica mode and L m is the maximized vaue of the ikeihood function for the estimated mode. In fact, according to the reationship of η and the number of sampes n s, AIC needs to be changed to AIC with a correction (AICc) when n s /η < 40 [23], i.e., 2η(η + 1) AICc = AIC + n s η 1. (12) In this paper, AICc is adopted to estimate the mode of the signa strength distribution instead of the cassic AIC. In practice, one can compute AICc for each of the candidate modes and seect the mode with the smaest vaue of AICc. The candidate modes in this paper incude Rice, Rayeigh, and Nakagami. Since our channe mode is reated to the distance between the transmitter and the receiver, the tunne shoud be divided into intervas. Assume that there are L intervas and we appy AICc for each candidate mode in every interva. As a resut, we can seect the most appropriate mode based on the frequency of the minimum AICc vaue of different candidate modes. In order to obtain enough data for each interva and ensure the accuracy of the mode, we set the ength of each interva as Fig. 5. Frequencies of AICc seecting a candidate distribution. 40 waveengths of WLANs [21], and then there are 100 intervas. Based on the frequencies of AICc of different distributions in the rea fied measurements, we observe that the Nakagami distribution provides the best fit compared with Rayeigh and Ricean distributions, as shown in Fig. 5. As a resut, we can define p(γ) as the Nakagami distribution. After the distribution of the signa strength is obtained, according to [20], we can derive the distribution of the SNR. Thus p(γ) = μμ γ μ 1 γ μ Γ(μ ) e( μ γ γ ) (13) where γ is the SNR of the received signa in the th interva, γ is the mean of the SNR in the th interva, μ is the fading factor of Nakagami distribution in the th interva, and Γ( ) is the gamma function. In fact, μ can be cacuated when appying AICc through the maximum-ikeihood estimator for each interva. V. R EAL FIELD MEASUREMENT RESULTS AND DISCUSSIONS Here, we compare our FSMC mode with rea fied test resuts to iustrate the accuracy of the mode. The effects of different parameters in the proposed mode are discussed. The number of states in our mode is first set to four. We aso use eight states to study the effects of the number of states on the accuracy of the proposed mode. In order to obtain the effects of distance intervas on the mode, we choose the intervas as 5, 10, 20, 50, and 100 m. We perform measurements in the tunne of the Beijing Subway Changping Line 20 times so that enough data can be captured. The accuracy of the FSMC mode is verified through another set of measurement data. Based on the measurement data, (8) (10), and (13), we derive the threshods {Γ n,n= 0, 1, 2,...,N} of the SNR in each distance interva. Tabes II and III demonstrate the threshods of the SNR eves at the ocation of 100 m for different intervas, where we divide the SNR into four and eight eves. As the distance intervas are different, the range of the SNR is different, and it brings different threshods, which can provide a more accurate mode.

6 6 IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS TABLE II THRESHOLDS OF SNR LEVELS (FOUR LEVELS) AT THE LOCATION OF 100 m FOR DIFFERENT INTERVALS TABLE III THRESHOLDS OF SNR LEVELS (EIGHT LEVELS) AT THE LOCATION OF 100 m FOR DIFFERENT INTERVALS Fig. 6. Simuation resuts generated from the FSMC mode and experimenta resuts from rea fied measurements. TABLE IV STATE 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 threshods, we can get the state probabiities and the state transition probabiities from the rea fied data. Tabe IV iustrates the state transition probabiities of the FSMC mode and the measurement data at the same ocation (35 40 m), when there are four states and the distance interva is 5 m. We can observe that the sum of the transition probabiity of each channe state is not equa to 1. This is because, in the measurement data, there are some state transitions that do not happen in adjacent states, such as transitions from state 1 to state 3. However, in our FSMC modes, for the sake of simpicity, we assume that states can ony transit to the adjacent states. Therefore, in Tabe IV, we ony consider state transitions between adjacent states. Consequenty, the sum of the transition probabiity of each state of the measurement data is not equa to 1. Nevertheess, it is very cose to 1, which means that our assumption is reaistic for practica tunne channes in CBTC systems. Fig. 6 shows the simuation resuts generated from our FSMC mode and the experimenta resuts from rea fied measurements. We can observe the great agreement between them. Next, we derive the mean square error (MSE) to measure the degrees of approximation, as shown in Fig. 7, where the y-axis is the MSE between the FSMC resuts and the measurement resuts, and the x-axis is the interva distance (5, 10, 20, 50, and 100 m). As shown in Fig. 7, when the distance interva increases, the MSE aso increases, which means that the accuracy of the mode decreases. Moreover, we can aso observe Fig. 7. MSE between the FSMC mode and the experimenta data with four states and eight states. that the MSE of the FSMC mode with four states is arger than that with eight states. The number of states in the FSMC mode pays a key roe in the accuracy. Nevertheess, when the distance interva is 5 m, the MSE difference is sma for the four-state FSMC mode (0.032) and the eight-state FSMC mode (0.028). In this figure, we can see that the FSMC mode with four states and the 5-m distance interva can provide an accurate enough channe mode for tunne channes in CBTC systems. VI. CONCLUSION AND FUTURE WORK Modeing the tunne wireess channes of urban rai transit systems is important in designing the wireess networks and evauating the performance of CBTC systems. In this paper, we have proposed an FSMC mode for tunne channes in CBTC systems. Since the train ocation is known in CBTC systems, the proposed FSMC channe mode takes train ocations into account to have a more accurate channe mode. The distance between the transmitter and the receiver is divided into intervas, and an FSMC mode is designed in each interva. The accuracy of the proposed mode has been iustrated by the simuation resuts generated from the proposed mode and the rea fied

7 WANG et a.: FINITE-STATE MARKOV MODELING FOR WIRELESS CHANNELS IN TUNNEL CBTC SYSTEMS 7 measurements. In addition, we have shown that the number of states and the distance interva have impacts on the accuracy of the proposed FSMC mode. Future work is in progress to study the effects of wireess channes on the contro performance of CBTC systems based on the proposed channe mode. REFERENCES [1] R. Pascoe and T. Eichorn, What is communication-based train contro? IEEE Veh. Tech. Mag., vo. 4, no. 4, pp , Dec [2] S. Su, X. Li, T. Tang, and Z. Gao, A subway train timetabe optimization approach based on energy-efficient operation strategy, IEEE Trans. Inte. Transp. Syst., vo. 14, no. 2, pp , Jun [3] Standard for Communications-based Train Contro (CBTC) Performance and Functiona requirements, IEEE Std (Revision of IEEE Std ), 2004, 0_1 45. [4] L. Zhu, F. R. Yu, B. Ning, and T. Tang, Cross-ayer handoff design in MIMO-enabed WLANs for communication-based train contro (CBTC) systems, IEEE J. Se. Areas Commun., vo. 30, no. 4, pp , May [5] Y. Zhang, Nove mode for propagation oss prediction in tunnes, IEEE Trans. Veh. Tech., vo. 52, no. 5, pp , Sep [6] K. Guan, Z. Zhong, J. Aonso, and C. Briso-Rodriguez, Measurement of distributed antenna systems at 2.4 GHz in a reaistic subway tunne environment, IEEE Trans. Veh. Tech., vo. 61, no. 2, pp , Feb [7] H.-P. Lin and M.-J. Tseng, Two-ayer mutistate Markov mode for modeing a 1.8 GHz narrow-band wireess propagation channe in urban Taipei city, IEEE Trans. Veh. Tech., vo. 54, no. 2, pp , Mar [8] S. Lin, Z. Zhong, L. Cai, and Y. Luo, Finite state Markov modeing for high speed raiway wireess communication channe, in Proc. IEEE Gobecom, Anaheim, CA, USA, 2012, pp [9] F. Babich, G. Lombardi, and E. Vaentinuzzi, Variabe order Markov modeing for LEO mobie sateite channes, Eectron. Lett., vo. 35, no. 8, pp , Apr [10] F. Babich and G. Lombardi, A measurement based Markov mode for the indoor propagation channe, in Proc. IEEE VTC, Phoenix, AZ, USA, May 1997, vo. 1, pp [11] H. S. Wang and N. Moayeri, Finite-state Markov channe A usefu mode for radio communication channes, IEEE Trans. Veh. Tech., vo. 44, no. 1, pp , Feb [12] C. Pimente, T. Fak, and L. Lisboa, Finite-state Markov modeing of correated Rician-fading channes, IEEE Trans. Veh. Tech.,vo.53,no.5, pp , Sep [13] C. Iskander and P. Mathiopouos, Fast simuation of diversity Nakagami fading channes using finite-state Markov modes, IEEE Trans. Broadcasting, vo. 49, no. 3, pp , Sep [14] A. Godsmith and P. Varaiya, Capacity, mutua information, and coding for finite-state Markov channes, IEEE Trans. Inf. Theory, vo. 42, no. 3, pp , May [15] A. Chockaingam, M. Zorzi, L. Mistein, and P. Venkataram, Performance of a wireess access protoco on correated Rayeigh-fading channes with capture, IEEE Trans. Commun., vo. 46, no. 5, pp , May [16] F. Babich, O. Key, and G. Lombardi, Generaized Markov modeing for fat fading, IEEE Trans. Commun., vo. 48, no. 4, pp , Apr [17] S. Loyd, Least squares quantization in PCM, IEEE Trans. Inf. Theory, vo. 28, no. 2, pp , Mar [18] P. Sadeghi, R. Kennedy, P. Rapajic, and R. Shams, Finite-state Markov modeing of fading channes A survey of principes and appications, IEEE Signa Process. Mag., vo. 25, no. 5, pp , Sep [19] J. G. Proakis, Digita Communications. New York, NY, USA: McGraw- Hi, [20] M. K. Simon and M.-S. Aouini, Digita Communication over Fading Channes.. Hoboken, NJ, USA: Wiey, [21] S. Wyne, A. Singh, F. Tufvesson, and A. Moisch, A statistica mode for indoor office wireess sensor channes, IEEE Trans. Wireess Commun., vo. 8, no. 8, pp , Aug [22] T. S. Rappaport, Wireess Communications Principes and Practice, Second ed. Beijing, China: Pubishing House of Eectronics Industry, [23] K. P. Burnham and D. R. Anderson, Mode Seection and Muti-mode Inference: A Practica Information-Theoretic Approach. New York, NY, USA: Springer-Verag, Hongwei Wang received the B.S. degree in eectronics and information engineering from Beijing Jiaotong University, Beijing, China, in 2008, where he is currenty working toward the Ph.D. degree with the State Key Laboratory of Rai Traffic Contro and Safety. His research interests incude train ground communication technoogy in communication base train ground communication systems and cooperative scheduing approaches in subway systems. Fei Richard Yu (S 00 M 04 SM 08) received the Ph.D. degree in eectrica engineering from the University of British Coumbia, Vancouver, BC, Canada, in From 2002 to 2004, he was with Ericsson (in Lund, Sweden), where he worked on the research and deveopment of wireess mobie systems. From 2005 to 2006, he was with a startup in Caifornia, USA, where he worked on the research and deveopment in the areas of advanced wireess communication technoogies and new standards. In 2007, he joined the Careton Schoo of Information Technoogy and the Department of Systems and Computer Engineering, Careton University, Ottawa, ON, Canada, where he is currenty an Associate Professor. His research interests incude crossayer design, security, green IT, and quaity-of-service provisioning in wireess networks. Dr. Yu was a recipient of the IEEE Outstanding Leadership Award in 2013; the Careton Research Achievement Award in 2012; the Ontario Eary Researcher Award (formery Premier s Research Exceence Award) in 2011; the Exceent Contribution Award at the 2010 IEEE/IFIP Internationa Symposium on trusted Computing and Communications (IEEE/IFIP TrustCom 2010); the Leadership Opportunity Fund Award from Canada Foundation of Innovation in 2009; and the Best Paper Awards at the 2012 IEEE Goba Communications Conference (GLOBECOM 2012), IEEE/IFIP TrustCom 2009, and Internationa Conference on Networking He serves on the editoria boards of severa journas, incuding IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, IEEE Communications Surveys and Tutorias, ACM/Springer Wireess Networks, EURASIP Journa on Wireess Communications Networking, Ad Hoc and Sensor Wireess Networks, Wiey Journa on Security and Communication Networks, and Internationa Journa of Wireess Communications and Networking. He is aso a Guest Editor for IEEE Systems Journa for the specia issue on Smart Grid Communications Systems. He has served on the Technica Program Committee (TPC) of numerous conferences, as the TPC Co- Chair of the IEEE GLOBECOM 2014, the IEEE INFOCOM 2014 Workshop on Mobie Coud Computing (INFOCOM-MCC 2014), GLOBECOM 2013, the 2013 Internationa Conference on Green Computing and Communications (GreenCom 2013), the 2013 IEEE Consumer Communications and Networking Conference (CCNC 2013), the IEEE INFOCOM 2012 Workshop on Green Networking and Smart Grids (INFOCOM-CCSES 2012), the IEEE ICC 2012 Workshop on Green Communications and Networking (ICC-GCN 2012), the 2012 Vehicuar Technoogy Conference (VTC Spring 2012), GLOBECOM 2011, INFOCOM-GCN 2011, the IEEE INFOCOM 2010 Workshop on Cognitive Wireess Communications and Networking (INFOCOM-CWCN 2010), the 2009 IEEE Wireess Communications and Mobie Computing Conference (IEEE IWCMC 2009), VTC Fa 2008, and the 1st Internationa Workshop on Wireess Networking for Inteigent Transportation Systems (WiN-ITS 2007); as the Pubication Chair of the 7th Internationa ICST Conference on Heterogeneous Networking for Quaity Reiabiity and Robustness (ICST Qshine 2010); and as the Co-Chair of the IEEE ICUMT Workshop on Cognitive Wireess Communications and Networking (ICUMT-CWCN 2009). Li Zhu received the B.S. and Ph.D. degrees in eectronics and information engineering from Beijing Jiaotong University, Beijing, China, in 2006 and 2012, respectivey. He was a Visiting Ph.D. Student with Careton University, Ottawa, ON, Canada, and the University of British Coumbia, Vancouver, BC, Canada. He is currenty a Facuty Member with Beijing Jiaotong University. His research interests incude train-ground communication technoogy in communication-based train-ground communication (CBTC) systems and cross-ayer design in train-ground communication systems.

8 8 IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS Tao Tang received the Ph.D. of Engineering degree from the Chinese Academy of Sciences, Beijing, China, in He is currenty a Professor with the Schoo of Traffic and Transportation, Beijing Jiaotong University, Beijing, China, where he is the Associate Director of the State Key Laboratory of Rai Traffic Contro and Safety. His research interests incude communication-based train contro, high-speed train contro systems, and inteigent transportation systems. Dr. Tang is a member of the Experts Group of the High Technoogy Research and Deveopment Program of China (863 Program), and he undertakes as the Leader of the Fied of Modern Transportation Technoogy Experts Group. He is aso a Speciaist of the Nationa Deveopment and Reform Commission and the 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, respectivey, from the Northern Jiaotong University, which has been caed Beijing Jiaotong University since He was a Visiting Schoar with Brune University, London, UK, studying eectronics and eectrica power engineering. From October 2002 to February 2003, he joined University of Caifornia, Berkeey, CA, USA, as a Senior Visiting Schoar. He is currenty a Professor with Beijing Jiaotong University. His research mainy focus on high-speed train contro systems and raiway transportation train contro systems, incuding main ocomotive signas, communication-based train contro systems, inteigent transportation, faut-toerant design of signa systems, faut diagnosis, system reiabiity, and security design. Dr. Ning is a Feow of the Association of Internationa Raiway Signaing Engineers (IRSE), The Institute of Engineering and Technoogy, and the China Raiway Society. He is the Deputy Director of the China Traffic System Engineering Society and the Beijing Raiway Society. He is a member of the China Overseas Returned Schoars Association and the editoria board of the Journa of Raiways in China. He served as the Chair of the Technica Committee on Rairoad Systems and Appications of the IEEE Inteigent Transportation Systems Society. He aso served as an Associate Editor of the IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS, in , and of Acta Automatica Sinica, in