Quasi-Full-Duplex Wireless Communication Scheme for High-Speed Railway

Transcription

1 Quasi-Full-Duplex Wireless Communication Scheme for High-Speed Railway Li Yan, Xuming Fang, Shuang Zhong Key Lab of Information Coding & Transmission, Southwest Jiaotong University, Chengdu 6003, China Abstract Full-duplex, massive antenna and higher frequency bands are core technologies of the fifth generation (5G) wireless communications to boost the throughput to satisfy the future explosively increasing demands of traffic volume. Unfortunately, the implementation of full-duplex is badly limited by the severe selfinterference. In this paper, we propose a novel quasi-full-duplex scheme on the basis of separate transceiver arrangement and massive antenna beamforming to achieve future high reliable and efficient professional railway wireless communication systems. To obtain accurate direction of arrival (DOA) of the desired user for beamforming, real-time feedback link and novel reference signal (RS) distribution are designed to adapt to high-speed scenario. The theoretical analysis and numerical simulations demonstrate that the proposed scheme can dramatically improve the spectrum efficiency and transmission reliability, except for the scope nearby the enodeb where the receiver of enodeb is overwhelmed by its local signal. The complementary method that employs power control and higher frequency bands has settled this problem entirely. Keywords 5G; quasi-full-duplex; massive antenna beamforming; high-speed railway I. INTRODUCTION With the confirmed decision by UIC (International Union of Railways) that narrowband GSM-R (Global System for Mobile Communication for Railway) will evolve to broadband LTE (long term evolution) for railway, reliability and capacity are becoming two challenges faced by future professional railway wireless communication systems. Fortunately, lots of technologies of the fifth generation (5G) wireless communications such as full-duplex, massive antenna and higher frequency bands are emerging to address the explosively increasing traffic volume of public mobile systems in near future, which can also be applied to high-speed railway scenario [-4]. In public wireless communication systems, no matter for UE (user equipment) or enodeb, the transmitter and receiver are closely placed. Probably because of that, the implementation of full-duplex is badly limited by severe self-interference which is the local transmit signal that may overwhelm its own receiver [5]. Nevertheless, a great deal of research attempting to solve the above problem is still in the infant state. Diversely, for railway scenario where the network topology is linear and the whole train is regarded as one user (passengers access railway wireless communication system via the onboard relay which is deployed on the top of the carriage), the transceiver arrangement can be changed to prevent that problem. Based on this, we propose a novel quasi-full-duplex wireless communication scheme, in which the transmitter and receiver of the enodeb are separately placed along the railway track and those of the onboard relays are deployed on the front and rear of the train respectively. In this way, it is possible to separate the uplink and downlink transmissions in the real physical space domain, then further in the wireless communication space domain via massive antenna beamforming. For bulk train, it is much easier to deploy massive antenna system. Generally, with an L-element antenna array for beamforming, SNR (signal to noise ratio) is supposed to be improved about L times [6]. Hence, massive antenna beamforming that has larger L (e.g. 00) thereby providing larger power gain can reduce SER (Symbol Error Rate) and enhance the transmission reliability, which is very significant for high-speed railway wireless communication systems. In the proposed scheme, transmit beamforming is implemented both on the train and enodeb. For uplink, the transmit antenna array of the train adaptively steers its main lobe towards the receiver of the enodeb as well as its null towards its own receiver that has been space-isolated from the transmit array. Thus selfinterference of uplink is suppressed by the null. While for downlink, most of self-interference situations are eliminated by the same way, of which the exception and its solution will be discussed later. Because the speed of the train is at least 350km/h, the estimation error of the DOA caused by the feedback delay makes it impossible to achieve the expected effect of beamforming. The most of high-speed railway scenario is LOS and fast time varying, thus it has wide coherence bandwidth and short coherence time, which is quite opposite to the public mobile scenario [7]. Based on this, we propose a novel RS distribution scheme, which is time-continuation but frequencyrarefaction and entirely different from the RS distribution of traditional LTE, to adapt to the special scenario in the proposed scheme. Furthermore, a real-time feedback link separated from the common data transmission is setup between the transmit array and receiver for both of the uplink and downlink. Once the receiver obtains CSI (Channel State Information) through RSs which are continuously assigned in time, CSI will be fed back to the transmit array via the real-time feedback link immediately instead of waiting for the next sub-frame. Since the time interval between two time-continuation RSs is about 70s in LTE, much smaller than the coherence time of highspeed railway, it is able to obtain the accurate and real-time DOA for beamforming in the proposed scheme. The theoretical analysis and simulation show that the proposed scheme improves 50% of the spectrum efficiency, as well as keep the same reliability compared with conventional SDMA (space division multiple access) technology, except for the small scope nearby the enodeb. That is because, in that area, three points inclusive of the transmit antenna array and /4/$ IEEE

2 receiver of the enodeb and receiver of the train stay on a straight line, that is, the uplink and downlink simultaneously using the same frequency band are of space-collision. And the leakage power of downlink results in interference to the receiver of enodeb. To mitigate this situation, power control and higher frequency bands with larger path loss are taken into account to supplement the proposed scheme. With power control technique, the leakage power, i.e. interference to others, is under control. On the other hand, although, the larger path loss of higher frequency bands is always regarded as a weakness considering the coverage, it can be utilized to reduce the strength of interference. The rest of this paper is organized as follows. Section II details the proposed scheme including the novel RS distribution and real-time feedback link design. Section III performs the theoretical analysis and numerical simulations and presents the complementary method to mitigate the only self-interference situation. Finally, Section IV concludes the paper. II. PROPOSED SCHEME A. Quasi-full-duplex scheme For high-speed railway scenario, the network topology is linear and the whole train is considered as a single user. In addition, the train just runs on the determined rail track, unlike the public mobile scenario where lots of users move to dispersive directions. Hence, the transceiver placement can be changed for the specific application scenario. And thanks to another emerging technology massive antenna, much narrower beamforming is achievable to realize better space multiplexing and much higher power gain is attainable to enhance the SNR. Therefore, the quasi-full-duplex wireless communication scheme is proposed to extend the capacity and enhance the reliability of future professional wireless communication systems for high-speed railway as depicted in Fig.. Full-duplex means that the uplink and downlink data can be transmitted at the same time in the same frequency band. Hence, to isolate the uplink and downlink in the real physical space domain then further in the wireless communication space doamin, the transmitter and receiver of the enodeb are separately installed at each side of the railway track. And for the onboard relay, the transmitter and receiver, which are connected via broadband fiber and both managed by the control center, are deployed on the front and rear of the train respectively as depicted in Fig.. Since the train is not volume limited, the massive antenna arrays are deployed both on the train and enodeb to implement narrower transmit beamforming. The data of the uplink and downlink are transmitted through the narrower main lobes formed by massive antenna arrays of the enodeb and train respectively. Based on the combination, the uplink and downlink transmission are entirely isolated in the wireless space domain as shown in Fig.. Meanwhile, to avoid self-interference the null of the transmit antenna array is steered towards its local receiver. Hence, the uplink and downlink can use identical frequency band f at the same time achieving double spectrum efficiency compared with that of conventional wireless communication systems. Except for the scope where the train goes through the enodeb, three points including the transmit antenna array and receiver of the enodeb (i.e. enodeb-t and enodeb-r as shown in Fig.) and the receiver of the train (i.e. Train-R) stay on a straight line, that is, the uplink and downlink are of space-collision. The leakage power from the enodeb-t turns into self-interference to the enodeb-r. However, the beamwidth formed by massive antenna array is much narrower, the scope will be very small even just several meters. The detailed influence on performance is presented in the following section. Fig.. f enodeb-r f f Train-T Control center Train-R Omni-directional antenna Transmit Antenna array Quasi-full-duplex scheme. f enodeb-t Transmit beam Feedback link B. Real-time feedback link and novel RS distribution Since the velocity of the high-speed train is very high, the feedback delay of CSI badly impacts the estimation accuracy of DOA, which will be proved by the following example. Suppose the speed of the train is 360km/h and the carrier frequency is GHz, then the maximum Doppler shift is 667Hz and corresponding coherence time is 0.63ms. In traditional LTE, CSI feedback is coupled with common data, that is, the CSI feedback period is at least one sub-frame which is ms for LTE. Because the feedback period is larger than the coherence time, the estimated DOA loses effectiveness [7]. Hence, for high-speed scenario, a real-time feedback link separated from the common data transmission is highly needed. Consequently, novel feedback link to meet the above demands is designed as shown in Fig.. Because the public mobile scenario is almost NLOS and slow time varying, the RSs of conventional LTE are densely distributed both in time and frequency domains as shown in Fig. [8]. While the high-speed railway is almost regarded as LOS, thereby experiencing much smaller multipath delay spread (e.g.50ns) [9]. Consequently, the coherence bandwidth over which the channel is highly correlated in frequency domain is much wider. For example, the coherence bandwidth corresponding to delay spread of 50ns is 4MHz. Hence, it is not necessary to densely assign RSs in frequency dimension for high-speed railway scenario. Instead, denser distribution in time domain is beneficial to overcome the fast time varying fading. Based on this, novel RSs distribution as shown in Fig. is designed to assist the proposed scheme to adapt to highspeed scenario. Along with the real-time feedback link, once the receiver detects the time-continuation RSs, it will feed the CSI back to the transit antenna array immediately within the

3 effective time. As a result, real-time estimated DOA is obtained to achieve expected effect of beamforming. Furthermore, this method is also suitable for other link adaptive technologies applied to high-speed scenario to get real-time CSI feedback. f(hz ) Reference signal f(hz ) s j ae h h receiver fc t(ms) fc t(ms) j ae RS distribution of existing LTE Novel RS distribution Fig. 3. The schematic of transmit beamforming with -element antenna array. Fig.. RS distribution. III. PERFORMANCE EVALUATION A. Theoretical analysis For spectrum efficiency and SER are highly relevant with SNR, the following mathematical model aims at acquiring the average received SNR at the receiver. The small-scale and large-scale fading inclusive of path loss and shadow effect is taken into account. Conditional on the channel gain h of small-scale fading, the received SNR is SNR Pr,h = P r σ h () where σ represents the power of Gaussian white noise; P r which is the received power of the desired signal including the large-scale fading can be expressed as [0] P r (x)[db] =P t PL(x 0 ) 0nlg x x 0 + X ε () where x is the propagation distance of the desired signal; P t is the transmit power of the antenna array; PL(x 0 ) is the path loss at reference distance x 0 which is evaluated by Hata Model []; n is the fading factor and its value ranges from to 4; and X ε stands for the shadow fading which obeys Gaussian distribution with mean zero and standard deviation ε. It can be easily certified that P r is also a random variable following Gaussian distribution with mean P t PL(x 0 ) 0nlg x x 0 and standard deviation ε. Considering -element antenna array, the schematic of beamforming is depicted in Fig.3, where α n is the fraction of power allocated to each of the transmit antennas and θ n is the phase shift of each antenna applied to the signal [6]. Suppose the number of antennas of the transmit array is N, then the same symbol s is transmitted from all of the antennas, each of which is multiplied by the complex value α n e jθn, for n =,,N, such that N n= α n =, preserving the total transmit power. As mentioned before, the high-speed railway scenario is almost LOS, thus, the small-scale fading follows Ricean distribution. Then, the fading channel between the nth transmit antenna and the receiver can be modeled as h n = k k + ejφn + +k b n (3) where k is the Ricean factor which denotes the ratio of the energy in the direct component to that in the diffused component; b n follows complex Gaussian distribution with mean zero and variance ; and φ n is [] φ n = πnδdf c cos Θ (4) c where the space Δd between two close antennas is not larger than half a wavelength; f c is the carrier frequency; c represents the velocity of the light; and Θ denotes the DOA of the direct component. Since each of the transmit antennas is multiplied by the complex component α n e jθn, the entire channel gain h of small-scale fading is [] h = = = N αn e jθn h n n= ( ) N k αn e jθn k + ejφn + +k b n n= k N αn e j(θn+φn) + B (5) k + n= N where B = +k n= αn e jθn b n and it is easy to prove that B is a complex-valued Gaussian random variable with mean zero and variance +k. With the specially designed real-time feed link and novel RS distribution, it is reasonable to believe that the transmit antenna array can obtain accurate DOA to let θ n = φ n. Then, for the former term k N A = k+ n= αn e j(θn+φn), the largest value Nk +k is achieved with setting the power allocation α n = N for all transmit antennas [6]. In (), because the random variables of large-scale fading and small-scale fading are independent mutually, the average received SNR can be further expressed as SNR = E Pr,h [SNR Pr,h] = σ E [P r] E [ h ] (6)

4 f f where E [P r ]=P t PL(x 0 ) 0nlg x x 0. On the basis of (5), h is consist of two parts A and B, then E [ h ] = Nk + +k Suppose the distance between the enodeb transmit antenna array and railway track is d min (d min denotes the distance between the enodeb receiver and railway track), and the vertical projection of enodeb is at the original point as shown in Fig.7 of Section IV, then the signal propagation distance x can be given by x = d min + d where d is the distance between the train and enodeb. Hence, the average received SNR can be expressed as ) SNR = p σ 0 t PL(x 0) 0nlg (d min +d /x 0 Nk + 0 +k (7) (8) Obviously, from Fig.4, it can be found that the spectrum efficiency of the proposed scheme is.5 times as high as that of conventional SDMA. Compared to the case without beamforming (i.e. without any space multiplexing), the spectrum efficiency of the conventional SDMA is more than /3 times higher on account of the power gain of beamforming. For the proposed scheme, transmit beamforming is applied both to the uplink and downlink. Hence, the power gain of beamforming proportional to the number of antennas can be achieved both in the uplink and downlink which goes the same for the conventional SDMA. The difference is that in the enodeb of conventional SDMA system, there exists transmit and receiver beamforming for downlink and uplink respectively. Thanks to this, the proposed scheme shown in Fig. has the same SER (i.e. transmission reliability) as that of the conventional SDMA. With the expression of average received SNR and above theoretical analysis, the numerical simulation of the Shannon limit of spectrum efficiency and SER is conducted. The simulation parameters are set as TABLE I, and the results are shown in Fig.5 and Fig.6. While for the case without beamforming, the average channel gain of small-scale fading E [ h ] = can be easily derived. And the corresponding average SNR is ) SNR = p σ 0 t PL(x 0) 0nlg (d min +d /x 0 0 (9) B. Numerical simulations For conventional SDMA system based on beamforming, the space multiplexing is only achieved in downlink as shown in Fig.4, because it is hard to equip the volume-limited user terminal with multiple antennas. Suppose there are M users accessing the same enodeb, then at least M + frequency bands are needed to realize reliable communications, among which at least one is used for the downlinks of all M users through the SDMA technology and residual M frequency bands are assigned to each of the M users for uplink transmission respectively. While in the proposed scheme, one frequency band is used for both of the uplink and downlink of one user. To avoid any possible co-channel interference, each of the M users is allocated its exclusive frequency band which has no overlap with any of other users. Consequently, only M frequency bands are enough to achieve identically reliable communications as shown in Fig.4. Since the application scenario is high-speed railway, only two trains are taken into account in Fig.4. Fig. 5. spectrum efficiency (bit/s/hz) TABLE I. SIMULATION PARAMETERS Parameters Value Carrier frequency GHz Bandwidth 0MHz Transmit power 43dBm Number of antennas 00 Coverage radius R 3km Width of track L.5m d min d min Thermal noise density -74dBm/Hz Beamwidth γ 0 o modulation scheme 6QAM Spectrum efficiency comparison. without beamforming conventional SDMA proposed scheme UE f f enodeb f f 3 Downlink uplink UE Conventional SDMA Fig. 4. Space multiplexing comparison. UE f f enodeb Proposed scheme UE C. Self-interference analysis From Fig.5 and Fig.6, we can find that the proposed scheme improves the spectrum efficiency, as well as keeps the same reliability compared with the conventional SDMA, except for the scope where the train goes through the enodeb (i.e. the area nearby the original point in Fig.5 and Fig.6). That is because the three points including the enodeb-t, Train-R and enodeb-r stay on a straight line, then the local transmitted signal from enodeb-t overwhelms its own receiver enodeb- R. What is still worse, the power gain resulting from transmit

5 SER self interference (dbm/hz) without beamforming conventional SDMA proposed scheme Fig. 6. SER i.e. transmission reliability comparison. Fig. 8. Strength of the self-interference in the scope nearby the enodeb. beamforming of the downlink enlarges the strength of selfinterference. As a result, the uplink transmission is badly impacted by the downlink signal causing the decline of the spectrum efficiency and reliability in this scope. Still thanks to the narrower beam formed by the massive antenna, the scope is very small. Suppose the beamwidth is γ, then the mapping area D of the angle 3γ is the self-interference scope as shown in Fig.8. And the length of the scope is D = tan ( ) 3γ (dmin + d min + L). The strength of the self-interference of this scope can be calculated through (), and the numerical simulation is described in Fig.9 of which the parameters are set as TABLE I. Fig. 7. db Self-interference Train-R enodeb-r d min L d min D 0 d(km) Train-T Control center Geometric sketch for self-interference analysis. R 3 enodeb-t D. Method based on power control and higher frequency bands Power control aims at adaptively adjusting the transmit power, so that the SNR achieved at the receiver is maintained at a relatively stable level (e.g. the power contour of ΓdB shown in Fig.7). Hence, this technique can save energy to improve the power efficiency in the robust communication environment, as well as enhance the transmission reliability under the adverse communication circumstance which will contribute a lot to the reliable transmission of crucial train control information. Furthermore, with this technique, the leakage power, i.e. interference to others, is under control. To mitigate the self-interference situation, power control is employed to supplement the proposed scheme. On the other hand, extension of higher frequency bands (which are higher than 3GHz and even higher than 0GHz []) to exploit more available spectrum is being considered by 5G. Despite the extended capacity, operation at these bands also improves the feasibility for new technical solutions such as the use of massive antenna in the volume-limited UE. Still, there are drawbacks curbing the application of higher frequency bands. Considering the coverage, the larger path loss of these bands is always regarded as a weakness. Reversely, in the proposed scheme, we take advantage of this weakness to further reduce the strength of self-interference combined with power control. In fact, as the above analysis results reveals, with an L-element antenna array for beamforming the SNR is supposed to be improved about L times. Hence, massive antenna beamforming which has larger L thereby providing larger power gain is also viewed as a potential technology to support practical coverage of the small cells using higher frequency bands.the simulation parameters are set as TABLE II. The strength of self-interference under this method is depicted in Fig.9. TABLE II. SIMULATION PARAMETERS Parameters Value Carrier frequency 5GHz Pathloss model M.35 Bandwidth 00MHz SNR threshold Γ 0dB Number of antennas 00 Coverage radius R km Width of track L.5m d min d min Beamwidth γ 0 o With the consideration of reliability, the SNR threshold of power control is set to 0dB, of which the corresponding leakage power of the downlink is restricted to a much lower value for the scope nearby the enodeb. Additionally, owing to the larger path loss of the higher frequency band, the downlink signal is extremely faded when it arrives at its own receiver, which is even 60dB lower than that of the thermal noise as

6 self interference (dbm/hz) Fig. 9. Strength of self-interference using power control and higher frequency bands. shown in Fig.9. Hence, with this method to supplement the proposed scheme, the self-interference of the scope nearby the enodeb can be neglected. IV. CONCLUSION In the present paper, the scheme based on novel transceiver placement and massive antenna that can form much narrower beam for space-multiplexing are proposed for the peculiar scenario of high-speed railway to achieve full-duplex. Furthermore, the larger power gain of beamforming improves the SNR thereby enhancing the transmission reliability. In order to obtain accurate DOA to realize expected effete of beamforming, novel RS distribution and real-time feedback link are designed. The theoretical analysis and numerical simulations demonstrate that the proposed scheme can improve the spectrum efficiency, as well as keep the same SER, i.e. reliability compared to conventional SDMA technology. Except for the scope where the train goes through the enodeb, three points inclusive of the transmit antenna array and the receiver of the enodeb and the receiver of the train stays on a straight line, then the local signal from enodeb overwhelms its own receiver. For this reason, the spectrum efficiency and transmission reliability are badly impacted in that area. To mitigate the situation, power control combined with higher frequency bands which experience larger path loss is applied to supplement the proposed scheme. The result shows that the strength of the self-interference is extremely reduced by this method, which is even 60dB lower than that of the thermal noise thereby being neglected. [] Y. Kishiyama, A. Benjebbour, H. Ishii, T. Nakamura, Evolution concept and candidate technologies for future steps of LTE-A, IEEE International Conference on Communication Systems (ICCS), pp , Nov. 0. [3] Ericsson. The Spectrum Crunch - busting the solutions myth. [online].available: c. [4] J. Hoydis, S. ten Brink, M. Debbah, Massive MIMO in the UL/DL of Cellular Networks: How Many Antennas Do We Need? IEEE J. Select. Areas Commun., vol. 3, no., pp. 60-7, Feb. 03. [5] J. I. Choi, M. Jain, K. Srinivasan, P. Levis, S. Katti, Achieving single channel, full duplex wireless communication, in Proc. 6th ACM MOBICOM, Chicago, Illinois, USA, Sep. 00. [6] P. Viswanath, DNC Tse, R. Laroia, Opportunistic beamforming using dumb antennas, IEEE Trans. Inf. Theory, vol. 48, no.6, pp. 77C94, Jun. 00. [7] L. liu, C. Tao, J. Qiu, H. Chen, L. Yu, W. Dong, Y. Yuan, Position- Based Modeling for Wireless Channel on High-Speed Railway under a Viaduct at.35ghz, IEEE J. Select. Areas Commun., vol. 30, no. 4, pp , May 0. [8] E. Dahlman, S. Parkvall, J. Sköld, LTE/LTE-Advanced for Mobile Broadband, Elsevier Ltd, 0. [9] K. Pekka, WINNER II Channel Models part II Radio Channel Measurement and Analysis Results, 007. [0] M. Cheng, X. Fang, W. Luo, Beamforming and positioning-assisted handover scheme for long-term evolution system in high-speed railway, IET Commun., vol. 6, no. 5, pp , Oct. 0. [] M. Cheng, X. Fang, Location information-assisted opportunistic beamforming in LTE system for high-speed railway, Eurasip Journal on Wireless Communications and Networking, Jul. 0. [] C. Xiao, Y. R. Zheng, N. C. Beaulieu, Novel Sum-of-Sinusoids Simulation Models for Rayleigh and Rician Fading Channels, IEEE Trans. Wireless Commun., vol. 5, no., pp , Dec ACKNOWLEDGMENT The work of authors was supported partially by the 973 Program under Grant 0CB3600, NSFC under Grant 60300, and the Program for Development of Science and Technology of China Railway Corporation under Grant 03X06-A. REFERENCES [] R. Baldemair, E. Dahlman, G. Fodor, G. Mildh, S. Parkvall, Y. Selen, H. Tullberg, K. Balachandran, Evolving Wireless Communications: Addressing the Challenges and Expectations of the Future, IEEE Veh. Technol. Mag., vol. 8, no., pp.4-30, Mar. 03.