Haibin Chen, Rongqing Zhang＊, Wenjun Zhai, Xiaoli Liang, Guojuan Song

1 Institute of Electronics Chinese Academy of Sciences, Beijing 100190, China

2 Institute of Electronics Chinese Academy of Sciences, Suzhou 215123, China

3 Colorado State University, Fort Collins, CO 80523, USA

Abstract: Cellular vehicle-to-everything(C-V2X) communications is regarded as a promising and feasible solution for 5G-enabled vehicular communications and networking. In this paper, we investigate the pilot design and channel estimation problem in MIMO-OFDM-based C-V2X systems with severe co-channel interference due to spectrum reusing among different V2X communication links. By using zero-correlation zone (ZCZ)sequences, we provide an interference-free pilot design scheme and a corresponding time-domain (TD) correlation-based channel estimation (TD-CCE) method. We employ the ZCZ sequences from the same family set to be designed as the TD pilot symbols and guarantee the pilot sequeneces for neighboring V2X communication links are code-division multiplexing (CDM). The co-channel pilot interference of the deisgned pilot symbols can be effectively eliminated by exploiting the provided TD-CCE method. Simulation results indicate that the accuracy of channel estimation can be effectively improved by the proposed scheme, whose performance is close to that of the non-interference situation.

Keywords: C-V2X; MIMO-OFDM; zero-correlation zone sequences; pilot design; channel estimation

I. INTRODUCTION

Vehicle-to-everything (V2X) communications is regarded as an essential and promising concept in providing real-time and highly reliable information flows to enable safe, efficcient, and environmentally conscious transportation services [1]. Cellular V2X (C-V2X) is developed within the 3rd Generation Partnership Project(3GPP) and considered as the prominent technology that can achieve the V2X requirements and pave the way to connected vehicles and autonomous driving in a more efficient manner [2]. Supported cellular networks and facilities, C-V2X communications includes many communication modes, such as vehicle-to-vehicle (V2V) and vehicle-to-pedestrian (V2P)communications via direct device-to-device(D2D) communications [3], and vehicle-to-infrastructure (V2I) and vehicle-to-cloud (V2C)communications via enhanced cellular links.

As effective and promising techniques,orthogonal frequency division multiplexing(OFDM) and multiple-input multiple-output(MIMO) are widely applied in current cellular networks to improve the spectrum and energy efficiency. We can envision that in future C-V2X systems, MIMO-OFDM will still be the key applied technique. However, due to various communication modes and complicated network topology [4][5], it will be more challenging to employ the MIMO-OFDM technique in C-V2X systems with complex and severe interference. Channel estimation is one of the most chanllenging issues in the MIMO-OFDM-based C-V2X system. The C-V2X system can be considered as a multicell environment with multiple active transmitters concurrently. Pilot interference due to simultaneous V2X transmissions will de finitely reduce the accuracy of channel estimation.Hence, effective pilot design that has strong interference robustness to result in high channel estimation accuracy is essential to realize the potential performance of MIMO-OFDM in C-V2X communications. However, most current pilot design and channel estimation methods in vehicular networks does not take this multi-cell situation into consideration [6].

In the literature, there are a few researches focusing on the improved pilot design in a multi-cell system [7][8]. In [7], a Chu sequences [9] based pilot design was provided for a multi-cell MIMO-OFDM system,which can effectively reduce the inter-cell interference. In [8], a general optimal pilot design scheme was proposed, in which by using a least square (LS) channel, the worstcase MSE is minimized. To emlimate the co-channel pilot interference entirely, the sequences used for pilot design should have both ideal impulse auto-correlation (AC) and zero cross-correlation (CC) properties. However, according to [10], we cannot realize both which actually cannot realize both AC and CC properties simultaneously. Therefore, although both the above pilot designs in [7] and[8] can be employed in MIMO-OFDM-based C-V2X systems to reduce the co-channel pilot interference and achieve an improved performance of channel estimation, the co-channel pilot interference cannot be eliminated. Especially, in a complicated interference scenario like C-V2X systems that includes various and spectrum-sharing communication modes, the cumulative co-channel pilot interference will be much higher, leading to signi ficatly reduced channel estimation performance.

Fig. 1. Pilot cluster reuse pattern for C-V2X communications.

In this paper, we focus on the optimal pilot design for MIMO-OFDM-based C-V2X systems in order to eliminate the co-channel pilot interference from a TD perspective. By exploiting the correlation properties of ZCZ sequences (ideal impulsive AC value and ideal zero CC value within a certain correlation zone), we propose an interference-free pilot design and pilot cluster reuse scheme using ZCZ sequences, in which the ZCZ sequences from the same family set are designed as the TD pilot symbols. Moreover, we also provide a corresponding TD-CCE method. Simulation results verify the efficiency of the proposed interference-free pilot design and the provided TD-CCE method in eliminating the co-channel pilot interference and thus increasing the accuracy of channel estimation in an MIMOOFDM-based C-V2X system.

The remaining of this paper is organized as follows. In Section II, we describe the considered MIMO-OFDM-based C-V2X system. In Section III, by using ZCZ sequences, we propose an interference-free pilot design and pilot cluster reuse scheme. In Section IV, we further provide a corresponding TD-CCE method.Simulation results are given in Section V, and the conclusions are drawn in Section VI.

II. SYSTEM DSCREIPTION FOR MIMO-OFDM-BASED C-V2X COMMUNICATIONS

As shown in figure 1, we consider an MIMO-OFDM-based C-V2X system, which mianly includes two communication types, i.e., downlink V2I communications from the BSs (or RSUs) to the vehicles and V2V communications. The BSs and the vehicles are equipped withNTandNRantennas (generallyNT>NR),respectively. Hence, in downlink V2I communications, there areNTtransmit antennas andNRreceive antennas, whereas in V2V communications, the number of transmit and receive antennas are bothNR. We further assume there areNBsources in the network, including the BSs and the transmitters in V2V communications. Note that the same channel estimation procedure is employed at each receive antenna in MIMO-OFDM systems. Therefore, for simplicity, in the following, we only take one receive antenna of a downlink V2I communication link as an example for further derivations.The similar expressions of V2V communication links can be easily obtained by replacing{t, NT} with {r, NR} in the derived results.

The TD transmitted symbol with the cyclic prefix at thet-th transmit antenna of thev-th source (assumed to be a BS in the derivations)can be expressed as

whereNgis the length of the cyclic prefix.In order to eliminate the effect of pilot interference, we assumewhereLdenotes the maximum channel multi-path delay andτmaxdenotes the maximum propagation time dfference among the adjacent sources [12]. NT, NR, and NVrepresent the sets of indices {0,1,...,NT? 1}, {0,1,...,NR? 1},and {0,1,...,NV? 1}, respectively. The channel response vector between thet-th transmit antenna of thev-th source and ther-th receive antenna of the targeted vehicle, denoted by, can be given as

Given perfect timing synchronization and frequency offset compensation at the receiver,the TD received symbol at ther-th receive antenna can be represented as

III. INTERFERENCE-FREE PILOT DESIGN AND PILOT CLUSTER REUSE USING ZCZ SEQUENCES

In this section, we introduce a simple and typical class of binary ZCZ sequences [11], which is exploited to design the TD pilot symbols.And then, we further propose an interference-free pilot design and pilot cluster reuse scheme for MIMO-OFDM C-V2X communications.

Note that the introduced class of binary ZCZ sequences is merely used as a feasible example in this paper, while many other classes of ZCZ sequences, e.g., ZCZ sequences given in [12]-[14], are also feasible to be applied in our proposed interference-free pilot design scheme.

According to [11], we can obtain

Fig. 2. AC value of ZCZ sequences as an example.

Fig. 3. CC value of ZCZ sequences as an example.

wherenandmare nonnegative integers,andL0is the initial sequence length during sequence construction. The correlation characteristics ofC(N,M,Z) can be represented as

Figs. 2 and 3 show the correlation characteristics ofC(1024,16,65), in which the sequence construction parameters in (6) are set asL0= 2,m= 3, andn= 3. From Figs.2 and 3, we can obviously there is a zero correlation zone for both AC and CC values,where the length of the ZCZ is equal toZ.

Considering that the ZCZ characteristic of ZCZ sequences can be exploited to eliminate the pilot interference, the TD pilot symbols are proposed to be designed by the ZCZ sequences from the same family set, where the sequence length is equal to that of the OFDM data symbol. And CDM is applied for the pliot cluster resue of neighboring V2X communication links. Speci fically, the pilot cluster reuse pattern in a C-V2X system is illustrated in figure 1, where at leastQ= 4 ZCZ sequence clusters to guarantee the pilot resue diversity among adjacent or overlapping communication areas.

As for multi-antenna communications,such as anMIMO system, the pilot sequences assigned to different transmit antennas are different ZCZ sequences but within the same pilot cluster, which means the pilot pattern among different transmit antennas is also CDM. Therefore, we have that each designed pilot cluster should contain at least NT different ZCZ sequences.

As for the V2I communications, any two neighboring BSs (or RSUs) should be allocated with different ZCZ sequence clusters. As for the V2V communications, the allocated ZCZ sequence cluster of a speci fic V2V communication link should be different from: 1) the ZCZ sequence clusters assigned to the BSs whose communication coverage the destination vehicle falls into; 2) the ZCZ sequence clusters assigned to the V2V communication links that are within two-hop effective communication distance of the source vehicle. Note that since the ZCZ sequence clusters assigned to the V2V communications are dynamic due to the mobility of the vehicles, whereas the ones allocated the BSs are settled at the time when the BSs are deployed. If there is no new BSs or RSUs added in the network, their allocated ZCZ sequence clusters may not change.

Meanwhile, to effectively eliminate the co-channel pilot interference, in a practical C-V2X system, the length of the constructed ZCZ sequences should be equal to or a little larger than the cyclic pre fix length.

IV. TD-CCE METHOD

In this section, to fully exploit the advatages of the ZCZ sequences based pilot design scheme,we further provide a TD-CCE method, which can eliminate the co-channel pilot interference effectively and efficciently. In addition, we also derive the theoretical normalized MSE with the proposed TD-CCE method to prove the efficiency in eliminating the co-channel pilot interference and the strong interference robustness of the designed pilot using ZCZ sequences.

The detailed procedure of the proposed TDCCE method is illustrated in figure 4, where a cross-correlator is utilized at the receiver to processLshifted local ZCZ sequences with the received pilot symbol, resulting in the corresponding channel impulse response of each channel. Because of the existence of both zero AC and CC zones of the applied ZCZ sequences, the multi-path interference as well as the pilot interference can be effectively eliminated through the CC processs. Hence,the proposed TD-CCE method can fully exploit the advantages of the ZCZ sequences and achieve a more accurate channel estimate in MIMO-OFDM-based C-V2X systems.

The TD-CCE channel estimator at ther-th antenna of a vehicle in a certain communication area, e.g.,v= 0, can be given as

Hence, the normalized MSE of the TDCCE channel estimator at ther-th antenna of the vehicle can be derived as

Fig. 4. TD-CCE method.

According to (11), we can obtain that the effect of the co-channel pilot interference in channel estimation can be entirely eliminated using the proposed ZCZ sequences based pilot design and the TD-CCE method, in which the received SNR is the only factor that affects the normalized MSE.

V. SIMULATION RESULTS

In this section, we evaluate the proposed interference-free pilot design and the TD-CCE method in an MIMO-OFDM-based C-V2X system as shown in figure 1, whereNB= 9(including six BSs and three vehicles as transmitters) and each BS is equipped withNT=4 transmit antennas, while each vehicle is equipped withNR= 2 transmit or receive antennas. The vehicles are randomly distributed in the investigated scenario. We consider the BS locating at (0, -50 m) as the serving transmitter, whereas the other transmitters (including the other BSs and the source vehicles) as interfering sources for the multi-cell channel estimation performance evaluation. The investigated desitination vehicle is located at (-100 m, 0 m). The ITU channel model A for vehicu-lar test environment given in [15] is employed to model the multipath channel fading. The other simulation parameters are given in table 1.

Table I. Simulation settings.

In the simulations, we use the ZCZ sequences in the family setC(1024,16,65) as the TD pilot symbols and make sure that each antenna of different transmitters in the considered scenario is assigned a different ZCZ sequence from the same family set. Moreover,we employ the optimal frequence-domain multi-cell CDM (FD-CDM) pilot design and the LS channel estimation method introduced in [8] for performance comparison. In the optimal FD-CDM pilot design scheme, the maximum allowed CC value is set to 0.004 to make sure that there are sufficient Chu sequences for the pilot cluster reuse in the investigated scenario.

In figure 5, we compare the proposed TDCDM pilot design and the optimal FD-CDM pilot design in terms of the normalized MSE.From figure 5, we can clearly find that the proposed scheme has a significant performance gain over the optimal FD-CDM scheme for channel estimation in C-V2X systems. Besides, we also employ the non-interference case as the performance baseline. We can find that the proposed TD-CDM pilot design and TD-CCE method can achieve very approaching performance compared with this non-interference case. This veri fies that the co-channel pilot interference can be effectively elimated by our proposed scheme in C-V2X systems with complicated interference scenario.

In figure 6, we simulate the normalized MSE performance of a moving vehicle when it passes by the BS located at (0 m, -50 m).The x axis officgure 6 denotes a series of the x locations of the investigated vehicle moving along the line (x, -20 m). Then, we have that larger x means the closer to the coverage zone edge of the BS, resulting in more interference from neighboring BSs. From figure 6, we can find that the proposed TD-CDM scheme has much stronger robustness to the interference compared with the optimal FD-CDM scheme which suffers from a signi ficant channel esti-mation performance degradation at the zone edge. But when the interference is weak (e.g.,at the zone center of a BS), the optimal FDCDM scheme has better normalized MSE due to the exactly pefect AC feature of the used ZC sequences.

VI. CONCLUSIONS

In this paper, from a TD perspective, we focused on the optimal pilot design and channel estimation for MIMO-OFDM-based C-V2X systems in order to eliminate the co-channel pilot interference. By applying a TD-CCE method, we proposed to utilize ZCZ sequences for interference-free pilot design and pilot cluster reuse. Due to both zone-contrainted perfect AC and CC characteristics of the ZCZ sequences,the co-channel pilot interference can be effectively eliminated. Simulation results have demonstrated the efficiency of the proposed scheme applied in the C-V2X system.

Fig. 5. Normalized MSE vs. SNR.

Fig. 6. Normalized MSE vs. locations of a moving vehicle.