Heterogeneous LTE/802.11a Mobile Relays for Data ...

5 downloads 0 Views 248KB Size Report
International Conference on Communications (ICC'09), Dresden, Ger- many, June 2009. [10] D. Borota, G. Ivkovic, R. Vuyyuru, O. Altintas, I. Seskar, and P. Spa-.
Heterogeneous LTE/802.11a Mobile Relays for Data Rate Enhancement and Energy-Efficiency in High Speed Trains Rachad Atat1 , Elias Yaacoub2 , Mohamed-Slim Alouini1 , and Adnan Abu-Dayya2 1 Electrical

Engineering Program, King Abdullah University of Science and Technology (KAUST), Thuwal, Makkah Province, Kingdom of Saudi Arabia. Email: {rachad.atat, slim.alouini}@kaust.edu.sa 2 Qatar Mobility Innovations Center (QMIC), Qatar Science and Technology Park, Doha, Qatar. Email: {eliasy, adnan}@qmic.com

Abstract—Performance enhancements of cellular networks for passengers in high speed railway systems are investigated. Relays placed on top of each train car are proposed. These relays communicate with the cellular base station (BS) over Long Term Evolution (LTE) long range links and with the mobile terminals (MTs) inside the train cars using IEEE 802.11a short range links. Scenarios with unicasting and multicasting from the BS are studied, both in the presence and absence of the relays. In addition, LTE resource allocation is taken into account. The presence of the relays is shown to lead to significant enhancements in the effective data rates of the MTs, in addition to leading to huge savings in the energy consumption from the batteries of the MTs. Index Terms—Mobile relay, effective rate, energy efficiency, LTE, heterogeneous networks.

I. I NTRODUCTION Passengers in high speed train systems may want to have access for Internet to be able to browse a website, read/send emails, download multimedia services, etc. [1], the fact that initiated the investigation of new network architectures that can provide the passengers with high speed mobile data services. While the Global System for Mobile communications - Railway (GSM-R) is used to exchange train control information (location, schedule, speed, etc) with a maximum data rate of 200 kbps, Long Term Evolution Network (LTE), constitutes the next generation wireless communication system for high speed railways since it was shown to provide good performance with advanced channel estimation and disperse deployed antennas on train [2]. LTE is best suitable for high-speed broadband data when compared with various technologies such as WiFi, satellite, 2G/EDGE, and 3G [3]. However, many problems and challenges arise despite the wireless broadband technology in use. When the mobile terminal (MT) communicates directly with the base station (BS), it will experience a severe degradation in the Quality of Service (QoS) since the wireless signal has to travel through the train, and penetrate through the metalized This work was made possible in part by NPRP grant # 09-180-2-078 from the Qatar National Research Fund (A member of The Qatar Foundation). The statements made herein are solely the responsibility of the authors.

windows, the fact that dramatically reduces and weakens the wireless link quality [4]. In addition, as the train moves at a high speed, the high bandwidth that is required to support multimedia applications (videos, pictures, audio, etc) decreases, and users may not be able to establish a direct link with the outside world unless the passenger carries equipment that has direct access to a satellite node [1], [5]. Relays have been widely investigated in the literature because they can minimize the total power consumption of the network nodes, maximize network lifetime, extend the coverage and expand the capacity in wireless systems [6]. Relays for the purpose of reducing the energy consumption were investigated in [7], [8], while cooperative wireless communication systems employed in vehicular networks have been investigated in [9], [10]. In this paper, we suggest a new relay-based heterogeneous network architecture, where a relay is installed on top of each car of the train (on the ceiling), and the closest wireless BS in the vicinity of the train communicates with the relays using LTE technology. Then, each relay communicates with the MTs that are located inside the train car using Wireless Local Area Network (WLAN) IEEE 802.11a, that is perceived as the best candidate for high speed trains to provide in-train coverage as it can achieve a peak data rate of 54 Mbps [4]. This network approach aims at enhancing the perceived data rates and at minimizing the total energy consumption of the MTs compared to the case where the BS communicates directly with the MTs in the train. Thus, the BS does not need to communicate with the hundreds of passengers in the train which reduces radio resource management control significantly. This approach will help in avoiding the radio signal propagation losses and low QoS, and maintaining a stable high speed wireless link between the relay and the MTs inside the train car. The paper is organized as follows. The system model is presented in Section II. The rate and energy derivations for the proposed relay-based approach are derived in Section III. Scheduling of LTE resources is described in Section IV. Simulation results are studied and analyzed in Section V. Finally, conclusions are drawn in Section VI.

     

   

     

 

Fig. 1. General system model.

where the train passengers are interested in a similar content. For example, this could be the case of entertainment services provided by the railroad operator in cooperation with the mobile operator: live news websites, broadcast of a live sports match, etc. In fact, cooperative relaying of content of common interest is receiving significant research attention, e.g. as in [15], where a low mobility scenario is considered though. Scenarios where passengers are downloading different content correspond to SR unicasting, and will be considered in future research. A. Data Rates (x)

II. S YSTEM M ODEL A high speed moving train is considered. Cellular coverage inside the train is ensured by LTE BSs deployed parallel to the train path. These BSs could be co-located with GSMR BSs. Typical deployments consist of having a separation distance dBS between BSs on the order of 7-15 km along the railroad. However, distinction should be made between GSMR used for railway control, which is out of the scope of this paper, and LTE, which is used to ensure cellular connectivity to passengers inside the train. The system model is depicted in Fig. 1. The train consists of a number K of train cars. On top of each car, a mobile relay is fixed in the ceiling. The relay performs heterogeneous communications using two antennas (or two sets of antennas in case MIMO communication is used, which could be an interesting future extension of this work): one antenna located outside the train car, used for communication with the BS on the long range (LR) cellular links, and another antenna inside the train car, used to communicate with the MTs inside the train using WLAN. In a given car, relay k serves a number Mk of MTs belonging to the train passengers. LTE communication is considered between the BS and the relays. In LTE, the available spectrum is divided into resource blocks (RB) consisting of 12 adjacent subcarriers, allocated in a 0.5 ms time slot. The shortest assignment unit consists of two consecutive slots, i.e., for a duration of 1 ms, which is the duration of one transmission time interval (TTI) [11], [12]. A total bandwidth of Wtot = 5 MHz, subdivided into NRB = 25 RBs of 12 subcarriers each is assumed. Each subcarrier has a bandwidth of Wsub = 15 kHz, such that the bandwidth of an RB is WRB = 180 kHz [11]. We will consider that one RB is allocated to each relay. The BS power is assumed to be subdivided equally among the RBs. The short range (SR) links between the relays and the MTs inside the car use the IEEE 802.11a protocol which uses Orthogonal Frequency Division Multiplex (OFDM) technique including 8 different data rates and 64 carriers. It supports a bandwidth of 16.6 MHz, with a carrier spacing of 312.5 KHz and a data rate ranging between 6 and 54 Mbps [13], [14]. In this paper, the downlink direction is studied. Unicasting and multicasting on the LR LTE links are investigated. On the SR, it is considered that multicasting is performed by the relays inside each train car. This corresponds to a scenario

Given the transmit power Ps,d that source s is using in order to transmit to destination d over subcarrier x, the (x) channel gain Hs,d of the channel between s and d over subcarrier x, the antenna gain of the source Gs , the antenna gain of the destination Gd , and the thermal noise power (x) σ 2 , the received signal-to-noise ratio (SNR) γs,d on the link between s and d over subcarrier x can be calculated as: (x)

(x)

γs,d =

(x)

Ps,d Gs Gd Hs,d

· (1) σ2 In the presence of relays, the source is either the BS transmitting to the relays (destination), or it could be the relay transmitting to the MTs inside the train car. In the traditional case without relays, the source is the BS and the destinations are the MTs. In this paper, discrete sets of Modulation and Coding Schemes (MCS), as is the case in practical standards, are considered. Thus, the discrete rates used in LTE and 802.11a are adopted. The 14 MCS used in LTE can be found in [16] and the eight MCS used in 802.11a in addition to their corresponding data rates can be found in [13], [14]. Assuming L possible discrete rates such that r1 < r2 < · · · < rL , then rate rl is used between s and d over subcarrier x if the SNR (x) is above a certain threshold ηl , i.e. ηl ≤ γs,d < ηl+1 . B. Channel Model The channels are modeled by pathloss, shadowing and fading. The channel gain on the link between source s and destination d over subcarrier x is given by: (x)

(x)

Hs,d,dB = (−κ − υ log10 ds,d ) − ξs,d + 10 log10 Fs,d · (2) In (2), the first factor captures propagation loss, with ds,d the distance between s and d, κ the pathloss constant, and υ the path loss exponent. The second factor, ξs,d , captures lognormal shadowing with a standard deviation σξ , whereas the (x) last factor, Fs,d , corresponds to Rayleigh fading (generally considered with a Rayleigh parameter a such that E[a2 ] = 1). In the channel model, spatial shadowing correlation is taken into account as the train moves along the railroad. The correlated shadowing model of [17] and [18] is applied in this paper. In addition, for fast Rayleigh fading, a block fading model is considered, where the fast fading remains constant for a fixed time Tdec which is the channel de-correlation time. Then the channel conditions change and remain constant for another Tdec , and so on.

B. Scenario of LR Multicasting

III. R ATE AND E NERGY C ALCULATIONS WITH THE R ELAY- BASED A PPROACH We denote by R the set of relays, and by Mk the set of MTs served in car k, with |Mk | = Mk , where | · | represents set cardinality. In addition, we denote by RL,d the transmission rate on the LR links from the BS to destination d (could be a relay in the proposed approach or an MT in the traditional approach), and by RS,kj the achievable transmission rate on the SR links from relay k to MT j.

In the case of LR multicasting, the BS transmits the content of common interest at the lowest rate achievable by the relays, in order to ensure that all relays receive the data correctly. Following the same analysis as in Section III-A, the results for LR multicasting with relays can be obtained by replacing RL,k with min RL,k in (3)-(4). The expressions of (5) and (6) k∈R remain the same. In addition, (7) becomes: Ew/o−relay,jk (n) = PL,Rx ·

Bjk (n) , min RL,i (n) 

i∈

A. Scenario of LR Unicasting The time needed to transmit one data bit to any MT jk ∈ Mk with LR unicasting and SR multicasting is given by: 1bit = Drelay,j k

1 1 + , RL,k min RS,ki

where the first term corresponds to the time needed by the relay to receive the data bit from the BS and the second term corresponds to the multicast transmission from relay k to the MTs of Mk . Hence, the effective data rate of MT jk ∈ Mk when relays are used can be expressed as the inverse of (3), or: Req,jk =

RL,k · min RS,ki i∈Mk

RL,k + min RS,ki

Mk

i.e. in the case of multicasting without relays, the minimization takes place over all MTs in all the cars of the train since the relays do not exist. IV. LTE R ESOURCE A LLOCATION

(3)

i∈Mk

In this section, we present the resource allocation techniques adopted on the LTE LR communications for each of the investigated scenarios. We denote by rd,y the achievable rate of a destination d over RB y. The destination could be a relay in the proposed approach or an MT in the traditional approach. A. LTE Resource Allocation with LR Unicasting

·

(4)

i∈Mk

The rates vary with the channel variations, and hence depend on the channel realization. Consequently, the number of bits that can be transmitted at the nth channel realization to all MTs jk ∈ Mk can be expressed as: Bjk (n) = Req,jk (n) · Tdec ·

In the case of LR unicasting, we consider that one RB is allocated to each destination according to Algorithm 1. This approach allocates RBs to destinations in a way to Algorithm 1 LTE resource allocation with unicasting 1: while All destinations have not been assigned an RB do 2: Find the pair (Destination d∗ ,RB y ∗ ) such that:

(5)

(d∗ (n), y ∗ (n)) = {arg max rd,y (n)} d,y

With PS,Rx denoting the power drained from the MTs’ batteries when their SR IEEE 802.11a interface is active, then the energy drained from the battery of each MT jk ∈ Mk in order to receive Bjk (n) bits with SR multicasting is given by: Bjk (n) · (6) Erelay,jk (n) = PS,Rx · min RS,ki (n) i∈Mk

In other words, the energy in (6) corresponds to the power drained multiplied by the reception time. In the absence of relays, MTs communicate directly with the BS and in this case Req,jk = RL,jk . In addition, the energy consumption becomes: Ew/o−relay,jk (n) = PL,Rx ·

(8)

Bjk (n) , RL,jk (n)

(7)

with PL,Rx denoting the power drained from the MTs’ batteries when their LR LTE interface is active to receive data from the BS.

(9)

Mark RB y ∗ (n) as occupied and Mark destination d∗ (n) as served Set RL,d∗ (n) = rd∗ ,y∗ (n) Repeat (9) for the remaining RBs and destinations 7: until all destinations are served or all RBs are allocated 8: end while 3: 4: 5: 6:

maximize LR performance by selecting the best RB for each destination. B. LTE Resource Allocation with LR Multicasting The transmission on a given RB is limited by the rate achieved by the destination having the worst channel conditions on that RB. Thus, the BS performs multicasting on the RB having the highest minimum rate, i.e., according to the following:    ∗ · (10) y (n) = arg max min rd,y (n) y

d

V. R ESULTS AND A NALYSIS

A. Effective Rate Results As the train moves along the railroad, the distribution of a content of common interest by the BS might occur at any position of the train with respect to nearest BS. Fig. 2 shows the effective rate results assuming that the nearest BS is located at dLR = 1 Km away from the first car in the train when the content distribution starts. The positions of the cars, relays, and MTs, in addition to the shadowing and fading values, are updated as the train moves during content distribution. It can be seen from Fig. 2 that the presence of relays leads to significant enhancements, both in the case of LR unicasting and multicasting. The best results are achieved with relays in the case of LR unicasting. In fact, in this case, each relay is granted its best LTE RB to receive the BS transmission on the LR according to (9), whereas high multicasting rates are achieved on the SR over IEEE 802.11a. With LR multicasting, the LR rate is limited by that of the relay having the worst LR performance according to (10). In the absence of relays, MTs receive their data directly on the LR. In the case of LR multicasting, the performance is limited by the MT having the worst performance. Hence, as the number of MTs increases, the probability of finding an MT with bad LR channel conditions increases. Consequently, the implementation of (10) with MTs as destinations leads to performance degradation as the number of MTs increases. On the other hand, in the case of LR multicasting with relays, the number of relays is constant, since one relay is available per car (K = 10 in the model considered). This explains the almost constant performance of the multicasting plot with relays, especially that high rates can be achieved on the SR relay-MT links as explained above, which makes the LR rate the limiting factor in determining the effective rate. In the case of LR unicasting without relays, one RB is dedicated to each MT. However, the number of MTs that can be simultaneously served cannot exceed NRBs = 25. When the number of MTs increases beyond this limit, performance

0.9

Effective Rate (Mbps)

0.8 0.7 0.6

Relay: unicast Relay: multicast w/o Relay: unicast w/o Relay, 200 RBs: unicast w/o Relay: multicast

0.5 0.4 0.3 0.2 0.1 0 0

50

100 Number of MTs

150

200

Fig. 2. Effective rate versus the total number of MTs.

1 Relay: d

LR

0.9

= 2km

w/o Relay: d

LR

Relay: d

LR

0.8

= 2km

= 3km

w/o Relay: d

LR

Effective Rate (Mbps)

In this section, we compare the performance of the scenarios with and without relays. In the simulations, we assume that the train moves along a track with a speed of 250 km/hour. The train consists of 10 cars, each of 20 meters in length and 5 meters in width. A relay is placed in the middle of the ceiling of each train car, in which up to 20 MTs can be uniformly distributed inside. The time where fading is considered constant is taken to be Tdec = 10 ms. In addition, we consider an IEEE 802.11a bandwidth of WSR = 16.6 MHz on the SR and an LTE bandwidth on the LR of WLR = 5 MHz. With WLR = 5 MHz, the bandwidth is subdivided into NRBs = 25 RBs of 12 subcarriers each [11], [12]. The BS transmit power is set to 5 Watts, equally divided among the LTE RBs. Moreover, we consider that the BS antenna gain is 10 dBi and that of the relay is 6 dBi. Channel parameters are obtained from [19]: κ = −128.1 dB, υ = 3.76, and σξ = 8 dB. Furthermore, we set PL,Rx = 1.8 Joules/s and PS,Rx = 0.925 Joules/s according to the measurements made in [20].

= 3km

Relay: dLR = 4km

0.7

w/o Relay: dLR = 4km

0.6 0.5 0.4 0.3 0.2 0.1 0 0

20

40

60

80 100 120 Number of MTs

140

160

180

200

Fig. 3. Effective rate for LR multicasting with different distances to the BS.

degrades significantly, as many MTs do not have an RB allocated to them. Therefore, in Fig. 2, we compare the results to a “fictitious” scenario where we assume that an RB is available for each MT (200 RBs in total). This could correspond in practice to an LTE-Advanced (LTE-A) deployment with carrier aggregation, where two 20 MHz bandwidth slots can be aggregated to lead to a total of 40 MHz, subdivided into 200 RBs. However, even in this extreme scenario, the case of 5 MHz bandwidth and 25 RBs performs better in the presence of relays. Fig. 3 shows the effect on the effective rate when dLR is increased, considering a BS antenna gain of 15 dBi. Results for LR multicasting are shown. As expected, the effective rates decrease as the distance to the BS increases, due to lower achievable LR rates since the SNR on the LR decreases. However, the superiority of the proposed approach is maintained. The same conclusions can be reached for LR unicasting, although the results are not shown here due to space limitations. B. Energy Results In this section, we investigate the total energy consumed by the MTs in order to receive a file of total size BT = 1 Mbits. Fig. 4 shows the energy consumption in the same scenario investigated in Fig. 2, i.e. with a BS located at dLR = 1 km from the first train car with a BS antenna gain of 10 dBi.

R EFERENCES

Energy consumption (Joules)

600

500

Relay w/o Relay: unicast w/o Relay: multicast

400

300

200

100

0 0

50

100 Number of MTs

150

200

Fig. 4. Energy consumption in Joules versus the total number of MTs.

In Fig. 4, the scenario of unicasting without relays in the case of 25 RBs assumes the LR interface is put to sleep when an MT is not allocated an RB on the LR. If this is not the case, the plot for unicasting without relays in Fig. 4 would correspond only to the case of LTE-A with 200 RBs, and the energy consumed in the case of 25 RBs would be orders of magnitude higher when the number of MTs increases, since most of the MTs would be spending energy most of the time without actually receiving any data. The case of LR multicasting leads to higher energy consumption since the transmission occurs at the lowest LR rate as expressed in (8), whereas with unicasting each MT receives at the rate it can achieve as expressed in (7), which leads to lower overall energy consumption. The superiority of the proposed approach is obvious, where the scenarios with LR unicasting and multicasting lead to the same results, since the interest is in the energy consumed by the MTs (not relays) according to (6). Hence, the energy is drained from the MTs’ batteries only during SR multicasting regardless of the LR transmission method. Consequently, only one plot for the scenarios with relays is presented in Fig. 4. From Fig. 4, it can be noted that cooperative schemes achieve significant energy savings, around 3.45 Joules with 200 MTs, whereas the energy for multicasting without relays reaches 600 Joules. The results of varying the distance as in the scenarios of Fig. 3 were investigated but are not shown due to space limitations. These results show that increasing the distance leads to increased energy consumption for the cases without relays, since a larger distance to the BS leads to lower achievable rates and higher energy consumption according to (7) and (8). However, the energy consumption for the scenarios with relays remains the same as in Fig. 4, regardless of the LR distance, since MTs receive the content via SR multicasting, and the distances between the MTs and relays inside each train car remain the same regardless of the LR distance to the BS. VI. C ONCLUSIONS The use of heterogeneous LTE/802.11a relays placed on top of each train car in high speed trains was proposed and investigated. LTE resource allocation on the LR BSrelay links was considered, both in the case of unicasting and multicasting. Relays were shown to lead to significant enhancements in the effective data rates of the MTs, in addition to huge savings in their energy consumption.

[1] K.-D. Lin and J.-F. Chang, “Communications and entertainment onboard a high-speed public transport system,” IEEE Wireless Communications, vol. 9, no. 1, pp. 84–89, February 2002. [2] K. Guan, Z. Zhong, and B. Ai, “Assessment of LTE-R using high speed railway channel model ,” In Proc. of the Third International Conference on Communications and Mobile Computing (CMC 2011), Qingdao, China, April 2011. [3] J. Garstenauer, “GSM-R evolution towards LTE ,” In Proc. of the Institution of Railway Signal Engineers (IRSE 2010) International Convention, New Delhi, India, October 2010. [4] Y. Zhou, Z. Pan, J. Hu, J. Shi, and X. Mo, “Broadband wireless communications on high speed trains,” In Proc. of the 20th Annual Wireless and Optical Communications Conference (WOCC 2011), New Jersey, USA, April 2011. [5] F. Greve, B. Lannoo, L. Peters, T. Leeuwen, F. Quickenborne, D. Colle, F. Turck, I. Moerman, M. Pickavet, B. Dhoedt, and P. Demeester, “FAMOUS: A network architecture for delivering multimedia services to FAst MOving USers,” Wireless Personal Communications, New Jersey, USA, April 2011. [6] M. Salem, A. Adinoyi, M. Rahman, H. Yanikomeroglu, D. Falconer, Y.-D. Kim, E. Kim, and Y.-C. Cheong, “An overview of radio resource management in relay-enhanced OFDMA-based networks,” IEEE Communications Surveys and Tutorials, vol. 12, no. 3, pp. 422–438, Third Quarter 2010. [7] M. Hajiaghayi, M. Dong, and B. Liang, “Energy-aware power allocation for lifetime maximization in single-source relay cooperation,” 25th Bienniel Symposium on Communications, Kingston, Ontario, Canada, May 2010. [8] R. Madan, N. Mehta, A. Molisch, and J. Zhang, “Energy-efficient decentralized cooperative routing in wireless networks,” IEEE Transactions on Automatic Control, vol. 54, no. 3, March 2009. [9] O. Trullols-Cruces, J. Morillo-Pozo, J. Barcelo, and J. Garcia-Vidal, “A cooperative vehicular network framework,” In Proc. of the IEEE International Conference on Communications (ICC’09), Dresden, Germany, June 2009. [10] D. Borota, G. Ivkovic, R. Vuyyuru, O. Altintas, I. Seskar, and P. Spasojevic, “On the delay to reliably detect channel availability in cooperative vehicular environments,” In Proc. of the 73rd IEEE Vehicular Technology Conference (VTC Spring 2011), Budapest, Hungary, May 2011. [11] 3rd Generation Partnership Project (3GPP), “3GPP TS 36.211 3GPP TSG RAN evolved universal terrestrial radio access (E-UTRA) physical channels and modulation, version 8.3.0, Release 8,” 2008. [12] 3rd Generation Partnership Project (3GPP), “3GPP TS 36.213 3GPP TSG RAN evolved universal terrestrial radio access (E-UTRA) physical layer procedures, version 8.3.0, Release 8,” 2008. [13] Rohde and Schwarz, “WLAN 802.11p measurements for vehicle to vehicle (V2V) DSRC,” Rohde and Schwarz Application Note, September 2009. [14] IEEE 802.11, “Wireless LAN medium access control (MAC) and physical (PHY) layer specifications, amendment 6: wireless access in vehicular environments,” 2010. [15] L. Keller, A. Le, B. Cici, H. Seferoglu, C. Fragouli, and A. Markopoulou, “Microcast: Cooperative Video Streaming on Smartphones,” In Proc. of the 10th International Conference on Mobile Systems, Applications and Services (MobiSys 2012), Low Wood Bay, Lake District, United Kingdom, June 2012. [16] 3rd Generation Partnership Project (3GPP), “3GPP TR 36.942 3GPP TSG RAN evolved universal terrestrial radio access (E-UTRA) radio frequency (RF) system scenarios, version 8.1.0, Release 8,” 2008. [17] Z. Wang, E. Tameh, and A. Nix, “Joint shadowing process in urban peer-to-peer radio channels,” IEEE Transactions on Vehicular Technology, vol. 57, no. 1, pp. 52–64, January 2008. [18] K. Yamamoto, A. Kusuda, and S. Yoshida, “Impact of shadowing correlation on coverage of multihop cellular systems,” In Proc. of the IEEE International Conference on Communications (ICC 2006), pp. 4538–4542, Istanbul, Turkey, June 2006. [19] 3rd Generation Partnership Project (3GPP), “3GPP TR 25.814 3GPP TSG RAN physical layer aspects for evolved UTRA, v7.1.0,” 2006. [20] K. Mahmud, M. Inoue, H. Murakami, M. Hasegawa, and H. Morikawa, “Energy Consumption Measurement of Wireless Interfaces in MultiService User Terminals for Heterogeneous Wireless Networks,” IEICE Transactions on Communications, vol. E88-B, no. 3, pp. 1097–1110, March 2005.