MIMO V2V Communications Via Multiple Relays: Relay Selection Over Space-Time Correlated Channels

Transcription

1 EuCNC216-Posters MIMO V2V Counications Via Multiple elays: elay election Over pace-tie Correlated Channels Eanouel T. Michailidis, Konstantinos Maliatsos, and Athanasios G. Kanatas Departent of Digital ystes, chool of Inforation and Counication Technologies, University of Piraeus 8 Karaoli & Diitriou t., 18534, Piraeus, Greece {eichail, kaliat, kanatas}@unipi.gr Abstract This paper presents an opportunistic relay selection policy related to the propagation characteristics of a ultiple-input ultiple-output (MIMO) vehicle-to-vehicle (V2V) ulti-relay syste, where the vehicle nodes are equipped with unifor linear ulti eleent antenna arrays. As Channel tate Inforation (CI) overhead is introduced by the relay selection process, reactive relay selection techniques over space-tie correlated MIMO channels are presented based on the CI of the links between the relays and the destination. ince the correlation of MIMO subchannels strongly depends on the distribution of the effective scatterers, a three-diensional (3-D) geoetrical odel for ΜΙΜΟ V2V relay channels is utilized. The results depict the throughput and the packet error rate (PE) perforance for different scenarios. Keywords Correlation; decode-and-forward relaying; ultiple-input ultiple-output (MIMO) channels; relay selection; vehicular counications; 3-D scattering. I. INTODUCTION Vehicle-to-vehicle (V2V) counications, which are an integral part of Intelligent Transportation ystes (IT), are expected to iprove convenience and safety of transportation, efficiently control road traffic, and provide obile infotainent applications. Nevertheless, the transission links of V2V systes are usually vulnerable to harsh ultipath fading effects. Cobining the features of cooperative diversity [1] with the benefits of ultiple-input ultiple-output (MIMO) technology [2] can potentially lead to robust and reliable signal transission in difficult terrains and/or long distances with enhanced channel capacity. Leveraging MIMO technology in V2V scenarios sees practically attractive, since ultiple antennas can be easily ounted on large vehicle surfaces. This paper investigates the perforance of a MIMO V2V syste in decode-and-forward (DF) ulti-relay wireless networks, where one vehicle relay node is assigned to assist a source in forwarding its inforation to a destination [3]. A full-duplex (FD) operation ode is applied, which facilitates frequency reuse and enables data reception and transission at the sae tie in a single frequency band. An opportunistic relay selection policy is perfored, where only one relay is activated [4], [5] in order to iprove the resource utilization and reduce the hardware coplexity. In the DF relaying networks, there are two typical relay selection policies: reactive relay selection and proactive relay selection [6]. ince instantaneous channel state inforation (CI) knowledge is required and ultiple relays ust be exained each tie, the ipleentation coplexity is increased in proactive relay selection scenarios leading to CI overhead, especially in propagation scenarios with obility properties, i.e., high vehicle velocity and rapid changing road topologies. Hence, a reactive relay selection policy is exained, where only the CI of the links between the relays and the destination is required to perfor distributed selection. The proposed relay selection policy is directly related to the underlying radio channel [7], and the space-tie correlation properties. ince the degree of correlation is a coplicated function of the degree of scattering and the antenna inter-eleent spacing at the source, relay, and destination [8], the establishent of a particular geoetry describing the location of the scattering objects is highly critical. It is assued that the radio waves travel in both the horizontal and vertical plane. This assuption is realistic, especially in densely built-up urban areas, where the antenna arrays are usually located lower than the surrounding scatterers and the scattered waves propagate by diffraction fro vertical structures, i.e., buildings, to the street level. In this paper, a odified version of the three-diensional (3-D) odel for MIMO V2V fading channels in relay-based networks in [9] is utilized, where the relay is equipped with both transit and receive antennas. The results deonstrate the throughput and the Packet Error ate (PE) perforance for various fading conditions. The rest of the paper is organized as follows. ection II presents the syste odel, while ection III outlines the applicability of the proposed syste on IT. ection IV details the syste geoetry and generates the MIMO channel atrix. esults are provided in ection V. Finally, conclusions are drawn in ection VI. II. YTEM MODEL This paper considers an inter-vehicular counication syste and frequency-flat MIMO channels. The source counicates with a destination through a set of M FD DF 1

2 relays, while the direct link between source and destination is obstructed due to high attenuation. To aid our analysis, the subscripts, D, and are affiliated with the source, the destination, and the -th selected relay, respectively. As shown in Fig. 1, the source and destination are equipped with one unifor linear array (ULA), whereas the relay is equipped with two ULAs; one for reception of the source s signal and one for transission towards the destination. The antennas are oni-directional with low height and are nubered as 1 p p L, 1 q q L, 1 qt qt L T, and 1 u u L D, respectively, where L, L, L T, and L D are the antenna eleents at the source, relay (reception), relay (transission), and destination, respectively. It is considered that the ultiple-antenna receivers use a Maxiu atio Cobining (MC) diversity schee. All the links exhibit block fading, which is considered constant during one tieslot and changes independently in the next tie-slot. Additive white Gaussian noise (AWGN) at the receiver is assued, where the noise N has zero ean and variance n, i.e., N CN O, n. Due to DF relaying, the relays decode the received signal and then re-encode it for transission to the destination. Thus, siultaneous reception and transission take place resulting in loop-interference (LI) fro the relay s output antenna to the relay s input antenna. However, depending on the vehicle size, the antennas at the relays can be isolated and perfect LI cancellation can be achieved. Additional LI cancellation and suppression techniques can be also applied to further itigate LI effect. The source is considered saturated and has always data to transit. Besides, the data rate is given in Bits-Per- Channel-Use (BPCU). In each tie-slot, one relay is selected to establish counication between the source and the destination through FD transissions. Aiing at reducing the aount of CI overhead, a reactive relay selection policy is exained. In this case, the source power level is fixed, the signal is broadcasted towards all the relays, and the relay s transission power levels are dynaically changed, i.e., power adaptation is perfored. This procedure intends to achieve a specific quality of service (Qo) level apped to a signal-to-noise ratio (N) value based on the CI of the -D link, which is fed back to the source in order to perfor the relay selection procedure. Then, power expenditure can be significantly reduced and network service provision is assured for extended tie periods. ource H M FD DF elays Loop-Interference -th elay H Destination Fig. 1. iple representation of a MIMO V2V ulti-relay counication syste, where one FD DF relay is selected. To denote whether or not the destination has successfully received the source s packet, Acknowledgeent/Negative- Acknowledgeent (ACK/NACK) essages are sent by the destination, which in turn triggers a contention procedure, towards the source through the selected relay. Providing that a node fulfills the selection requireents, it can always deliver the packet to the destination. Otherwise, no candidates are available for relaying, and hence the source retransits the data itself. At the start of each tie-slot the relays should exchange signaling with the destination by sending pilot sequences. The latter estiates the CI and notifies the relays on this CI. The selection takes place aong the relays which decoded the source s signal. The set of these relays is denoted as U : N N, where N is the N in the -D link and N is an instantaneous N value at the receiver. The best relay is denoted by b and is chosen as follows [7] b nn arg in P arg in, U U g D where P is the -th relay s transission power and g D is the channel power in the -D link. III. APPLICATION OF THE V2V ELAYING CHEME ON IT In this section, the application of the proposed syste odel and the relaying schee on IT G5 standard [1] is investigated. The physical and edia access control (MAC) layer of IT G5 is based on the ost popular relevant standard, the IEEE 82.11p [11], with 1 MHz signal bandwidth. It is assued that a safety critical IT application is used, e.g., Cooperative Adaptive Cruise Control, which eans that the IT G5 Control Channel (CCH) is used. In IT applications, the vehicles will transit data using rando access, i.e., the IEEE 82.11p. For safety critical applications, all vehicles have the obligation to state their presence and infor other vehicles for status changes periodically. For typical applications, the iniu transission frequency of IT essages is 1 to 25 Hz. In the investigated syste, 25 Hz iniu transission frequency is assued, which eans that all vehicles participating in the network should broadcast at least one data packet every 4 sec. For the considered syste, this eans that every 4 sec a given vehicle will receive data fro all the vehicles in the network. The size of each IT essage is typically less than 1kByte for reduced latency and they are transitted using orthogonal frequency division ultiplexing (OFDM). These signals can be used to extract CI for all the links with the adjacent nodes. However, in ost cases, the nodes will exchange signals ore frequently with uch higher rate and therefore CI is updated uch ore often. elays are assued to have FD capabilities. This eans that each relay is able to directly retransit the received essage fro the source without the need to content for access in the ediu. Moreover, possible throughput perforance degradation due to the retransission procedure is significantly reduced. (1) 2

3 IV. A 3-D MODEL FO MIMO V2V ELAY CHANNEL This section proposes a odified version the 3-D geoetrical odel for MIMO V2V relay channels in [9] and considers that the -th relay is equipped with one transit and one receive ULA. Fig. 2 deonstrates this odel for a 2x2x2 channel. One observes that the source, the -th relay, and the destination are oving with speeds v, v, and v D, respectively, in the directions deterined by the angles γ, γ, and γ D, respectively. The antenna inter-eleent spacing at the source, relay (reception), relay (transission), and destination is denoted by δ, δ, δ T, and δ D, respectively, whereas the angles θ, θ, θ T, and θ D represent the orientation of the corresponding antenna arrays, relative to the x-axis. Moreover, the angles ψ, ψ, ψ T, and ψ D describe the elevation angle of the eleents p, q, q T, and u, respectively, relative to the x-y plane. As shown in Fig. 2, the scattering objects in the vicinity of the vehicle nodes lie on the surface of three cylinders, which reflect the influence of three heterogeneous scattering environents. It is assued that M scatterers denoted by 1, 2,..., M are situated around the source, on the surface of a cylinder of radius. iilarly, k l k 1, 2,..., K and l 1, 2,..., L scatterers in the k l vicinity of the relay with K L and for k l lie on a surface of a cylinder of radius, while n N scatterers denoted by D n 1,2,..., N are situated at the destination, on the surface of a cylinder of radius. D Only double-bounce non-line-of-sight (NLo) propagation conditions are considered, which are doinant in urban acrocells propagation environents [12]. The angles a and denote the aziuth angle of departure (AAoD) and the elevation angle of departure (EAoD), respectively, of the wave transitted fro the source and ipinged on, k k while a and are the aziuth angle of arrival (AAoA) and the elevation angle of arrival (EAoA), respectively, of the k wave scattered fro and received at the relay. Besides, l l a and denote the AAoD and the EAoD, respectively, l of the wave transitted fro the relay and ipinged on, n n while ad and D are the AAoA and the EAoA, n respectively, of the wave scattered fro D and received at the destination. Due to the heterogeneity of the propagation environents and the double-bounce scattering, the angles of departure are independent fro the angles of arrival, while the aziuth and elevation angles are also independent [13]. A. Generation of the MIMO Channel Matrix The effective MIMO channel transfer atrix describing the relay channel is expressed as L D L H H H, (2) z v y x p k v a k q p q q T k a -th elay T q T v D T l D ource D l a Destination D l u D u D n D n a D n D Fig. 2. The three-cylinder odel for a 2x2x2 MIMO V2V relay channel, where the -th FD DF relay equipped with one transit and one receive ULA is selected. L where L H is the MIMO channel transfer atrix for L the - link and D L H is the MIMO channel transfer atrix for the -D link. Considering a 2x2 MIMO channel, H can be derived as follows [14] H H (3) 1/2 T /2 w, where and are the 2x2 source and 2x2 relay (- link) correlation atrices, respectively, H w is a 2x2 stochastic atrix with independent and identically distributed zero-ean coplex Gaussian entries, 1/2 denotes the atrix square root operation, and T denotes the transpose operation. ince the nuber of the scatterers in the vicinity of the source is infinite, the discrete AAoD a and the discrete EAoD can be replaced with continuous rando variables α and β, respectively, with a joint probability density function (pdf). The aforeentioned rando variables are utually independent. Hence, the joint pdf can be decoposed to a product of two separate pdfs. The von Mises pdf [15] is used to characterize the AAoD α and is defined as exp k cos a f a, - a, (4) 2 I k where I the first kind,, is the zeroth-order odified Bessel function of is the ean angle at which the scatterers in the vicinity of the source are distributed in the 3

4 x-y plane, and k controls the spread around the ean. etting, k i.e., 1/2, f a incurs isotropic scattering. As k increases, the scattering becoes increasingly non-isotropic. The Parson s pdf [16] is used to characterize the EAoD β and is defined as f cos, < 4 2 2,, otherwise where is the axiu value of. Using (4) and (5) and carrying out extensive calculations, one can derive the source space-tie correlation function (TCF) and then obtain the eleents of the source correlation atrix as follows [9], [17] pp, ' where I x y T, cos 2 4 I 2 2 cos k (5) exp j2 p p sin sin / d, (6) x j2 p p cos cos / j2 f cos k cos / cos, (7),ax y j2 p p sin cos / j2 f sin k sin /cos, (8),ax f,ax v / is the axiu Doppler shift associated with the source, and λ is the carrier wavelength. Note that the source TCF in (6)-(8) is accurate only over a tie duration that is uch saller than in / v. ince a pair of source and relay (- link) correlation atrices is required to derive H, one can siilarly define the relay (- link) TCF q,, q fro (6)-(8) by replacing the indices and then construct the relay (- link) correlation atrix. By following a siilar procedure to that used to generate H, H can be also generated. Then, using (2), the channel transfer atrix H describing the overall MIMO relay channel can be obtained. V. EULT This section presents results regarding the perforance evaluation of the proposed reactive relay selection technique. The evaluation was perfored using a developed IT G5 siulator. The siulator includes a coplete ipleentation of the IT G5 PHY and MAC layers. ix nodes were assued, where nodes 1 and 2 are unable to counicate directly with each other. The relaying echanis was introduced in the IT G5 standard in order to establish connectivity between nodes 1 and 2. Nodes 5 and 6 have FD DF relaying capability and therefore can be used as relays, while nodes 3 and 4 are conventional single-antenna transceivers. Moreover, nodes 1, 2, 5, and 6 use two antennas. In transitting ode, nodes 1 and 2 equally share the available power at both antennas. In receiving ode, an MC algorith is applied in order to fully exploit the diversity gain. It is noted that no standard odification is necessary in order to ebed the MC operation in an IT G5 syste. The relaying echanis was ipleented in the IT G5 standard with the following odifications: Nodes 1 and 2 use the Address 4 field of the MAC header in order to request relaying and indicate the preferred relaying node. Therefore, if the Address 4 is epty then conventional transission occurs. Otherwise, the node indicated in the specific address will act as a relay. Transission is perfored in the IT G5 Control Channel (1 MHz bandwidth), where 48 of the 64 IT G5 OFDM subcarriers were used. The transitted packet was generated randoly with an average size of 1kbyte. The average N at the receiver varied fro 1 to 2 db. The carrier frequency was 5.9 GHz. Counication is perfored with the IT G5 basic rate (BPK ½ Coding). The values of the odel paraeters used to generate the MIMO channel atrix are δ = δ = δ T = δ D = λ, θ = θ = θ T = θ D = 9 ο, ψ = ψ = ψ T = ψ D = 45 ο, μ = μ D =, μ = μ = π/6, γ = γ D =, γ = π/4, and f,ax = f,ax = f D,ax = 1 Hz. The selected axiu Doppler frequency corresponds to an average speed of 24 k/h. The siulation odel evaluates the gains offered by the use of an MC diversity receiver. Moreover, the outdated CI effects are considered. ince the preferred relay station is defined in the Address 4 field of the previously transitted packet, the outdated CI ay have significant ipact on the syste perforance, when no significant teporal correlation exists. Perforance was evaluated for different distributions of the effective scattering objects, i.e., different degree of local scattering in the aziuth doain (paraeter k = k = k = k = k D ) and different axiu value of the ultipath elevation angles (paraeter β ax = β ax = β ax = β ax = β Dax ). Note that large k and sall β ax values indicate that the local scatterers are concentrated in specific angles at aziuth and elevation planes correspondingly. On the contrary, sall k and large β ax values indicate that the local scatterers are distributed in a uch larger range of aziuth and elevation angles. Then, the wireless radio channel rapidly changes and the Doppler effects becoe ore significant. Fig. 3 presents the achieved net throughput between nodes 1 and 2 with DF relaying and MC reception. To evaluate the diversity and relaying gain, the perforance of conventional single-antenna point-to-point transission between nodes 3 and 4 is also presented. In Fig. 4, the results of PE for the IT G5 network are also deonstrated. Packet losses ay occur either due to poor channel conditions or due to collisions in the rando access schee. More specifically, in high N packet loss occurs ostly due to packet collisions and thus the gain of the diversity technique is decreased. The results show that a) there are significant gains fro the use of the MC diversity receiver in the link perforance and b) the perforance degrades, when the scatterers are distributed in a large range of aziuth and/or elevation angles. These results also justify the inclusion of the third diension of the geoetrical channel odel. 4

5 Net Throughput (kbps) between nodes 1-2 and nodes 3-4 Packet Error ate between nodes 1-2 and nodes Nodes 3-4, k=2, β ax =1 o Νodes 1-2, k=2, β ax =1 ο Nodes 3-4, k=5, β ax =2 ο Νodes 1-2, k=5, β ax =2 ο Nodes 3-4, k=, β ax =8 ο Nodes 1-2, k=, β ax =8 o N (db) Fig. 3. Network throughput versus N for a) nodes 1 and 2 with relaying and MC reception and b) nodes 3 and 4 with single-antenna point-topoint links for different degree of local scattering and different axiu ultipath elevation angles Nodes 3-4, k=2, β ax =1 o Nodes 1-2, k=2, β ax =1 ο Nodes 3-4, k=5, β ax =2 ο Nodes 1-2, k=5, β ax =2 ο Nodes 3-4, k=, β ax =8 ο Nodes 1-2, k=, β ax =8 ο N (db) Fig. 4. PE versus N for the traffic between a) nodes 1 and 2 with relaying and MC reception and b) nodes 3 and 4 with single-antenna point-to-point links for different degree of local scattering and different axiu ultipath elevation angles. VI. CONCLUION In this paper, an inter-vehicular MIMO wireless network, where a source and a destination counicate via ultiple relays, has been studied. A reactive relay selection policy over space-tie correlated MIMO relay channels has been applied based on instantaneous CI knowledge, which selects the relay requiring the iniu power expenditure to forward the source s signal towards the destination. To copute the spacetie correlation, a 3-D channel odel for MIMO V2V relay channels, where the selected FD DF relay is equipped with one transit and one receive ULA, has been proposed. Finally, the network throughput and PE perforance for an IT G5 network has been evaluated for various scattering conditions. ACKNOWLEDGMENT This research has received funding fro the European Union s Horizon 22 research and innovation prograe under OADAT Grant Agreeent No EFEENCE [1] A. Nosratinia, T. E. Hunter, and A. Hedayat, Cooperative counication in wireless networks, IEEE Coun. Mag., vol. 42, no. 1, pp. 74-8, Oct. 24. [2] E. Telatar, Capacity of ulti-antenna Gaussian channels, Eur. Trans. Telecoun., vol. 1, no. 6, pp , Nov./Dec [3] A. I. Akin, H. Ilhan, and Ö. Özdeir, elay selection for DF-based cooperative vehicular systes, EUAIP J. Wireless Co. and Networking 215, 215:3. DOI /s [4] A. Bletsas, A. Khisti, D. eed, and A. Lippan, A siple cooperative diversity ethod based on network path selection, IEEE J. elect. Areas Coun., vol. 24, pp , Mar. 26. [5] N. Noikos, T. Charalabous, I. Krikidis, D. Vouyioukas, and M. Johansson, Hybrid cooperation through full-duplex opportunistic relaying and ax-link relay selection with transit power adaptation, in Proc. IEEE Int. Conf. on Co. (ICC), Jun [6] M. Xia and. Aïssa, Fundaental elations Between eactive and Proactive elay-election trategies, IEEE Counications Letters, vol. 19, no. 7, pp , Jul [7] N. Noikos, E. T. Michailidis, D. Vouyioukas, and A. G. Kanatas, Mobile-to-Mobile Counications via tratospheric elays: elay election and Perforance Analysis, in Proc. IEEE International Conference on Counications (ICC) 215, London, UK, 8-12 Jun [8] D.-. hiu, G. J. Foschini, M. J. Gans, and J. M. Kahn, Fading correlation and its effect on the capacity of ultieleent antenna systes, IEEE Trans. Counun., vol. 48, no. 3, pp , Mar. 2. [9] Ε. T. Michailidis, P. Theofilakos, and A. G. Kanatas, A 3-D Model for MIMO Mobile-to-Mobile Aplify-and-Forward elay Fading Channels, in Proc. 6th European Conference on Antennas and Propagation (EuCAP) 212, Prague, Czech epublic, 26-3 Mar [1] Final draft ETI E V1.1. (29-11) Intelligent Transport ystes (IT); European profile standard for the physical and ediu access control layer of Intelligent Transport ystes operating in the 5 GHz frequency band. [11] "IEEE P82.11p/D3., Draft Aendent for Wireless Access in Vehicular Environents (WAVE)," Jul. 27. [12] A. G. Zajic, G. L. tüber, T. G. Pratt, and. T. Nguyen, Wideband MIMO obile-to-obile channels: Geoetry-based statistical odeling with experiental verification, IEEE Trans. on Vehic.Technol, vol. 58, no. 2, pp , Feb. 29. [13] Y. Yaada, Y. Ebine, and N. Nakajia, Base station/vehicular antenna design techniques eployed in high capacity land obile counications syste, ev. Electr. Coun. Lab., vol. 35, no. 2, pp , [14] D. Gesbert, H. Bölcskei, D. A. Gore, and A. J. Paulraj, Outdoor MIMO wireless channels: Models and perforance prediction, IEEE Trans. Coun., vol. 5, no. 12, pp , Dec. 22. [15] A. Abdi and M. Kaveh, A space-tie correlation odel for ultieleent antenna systes in obile fading channels, IEEE J. on elec. Areas in Counic., vol. 2, no. 3, pp , Apr. 22. [16] J. D. Parsons and A. M. D. Turkani, Characterisation of obile radio signals:model description, in Proc. Inst. Electr. Eng. I, Coun., peech, Vision, vol. 138, no. 6, pp , Dec [17] A. G. Zajić and G. L. tüber, Three-diensional odeling, siulation, and capacity analysis of space-tie correlated obile-to-obile channels, IEEE Trans. on Vehic. Technol., vol. 57, no. 4, pp , Jul