群有关。然而,需要注意的是,所有这些散射体相关的参数实际上完全是通过测量获得的，而不是预先定义的随机分布的散射/集群。因此，所有的SCM,COST2100和WINNER模型归类为随机参数模型更合适。接下来,我们将对广泛使用的伪几何计量模型做一个简要介绍。
WINNER channel model: In WINNER channel models [24], the basic principle is that for every link the large-scale parameters, fox example, angular spreads, are taken from a map. In that way the correlation properties of those parameters are matched with those observed in measurements. However, the small-scale parameters,for example, AoA and AoD, are randomly drawn from a distribution, independently for each link. This means that even close-by links have independent values for for example AoA and AoD, which is of course not the case in reality. This spatial inconsistency is not problematic with the quasi-stationary modelling of WINNER (drop concept), but it has an impact on performance with for example multi-user multi-input multi-output (MIMO) case. The spatial inconsistency also means that the WINNER approach does not handle time evolution very well. New set of parameters are randomly drawn at each location of a mobile, and there is no smooth transition between two locations. This means that dynamic simulations are problematic. Interpolation between two locations is of course possible. The interpolation can be done by drawing random small-scale parameters, such as cluster delays, powers, directions and so on, to two UE locations and linearly interpolating parameter values in between locations. A problem may result because interpolated values are always between the original values, and thus with interpolation all distributions become narrower. WINNER models also do not specify transitions between different propagation environments (urban, rural, outdoor, indoor and so on) or between LoS and NLoS, which also create spatial inconsistency and unrealistic transients.
WINNER 信道模型：在WINNER 信道模型中每一个环节的基本原则是,大规模的参数,例如角差,都来自映射。那些参数的相关属性以这种方式与观测值相匹配。然而,每一个独立链接的小规模参数,例如AoA和AoD,都是随机分布的。这就意味着即使靠近链接也有独立的值,例如AoA和AoD,当然现实中并非如此。这种不一致不影响似稳的WINNER模型,但它对有些情况下的性能有影响,例如多用户多输入多输出(MIMO)的情况。这种不一致也意味着WINNER模型不能很好地处理时间演化。新设置的参数在每个位置随机移动，而且在两个位置之间没有平稳的过渡。这意味着动态模拟是有问题的。两个位置之间插值是可以实现的。两位置之间的线性插值可以通过规划如集群延迟和力的方向等小规模随机参数来实现。由于总是在两个原始数据之间插值,因此可能导致插值分布变得越来越窄。WINNER模型没有指出不同传播环境(城市、农村、户外、室内等)之间以及仿真结果之间的转换，这也导致了空间不一致以及不切实际的转变。
3GPP D2D channel model: In December 2012,a study item on LTE Device to Device Proximity Services was established in 3GPP[42].Several channel models Were proposed for D2D in 3GPP RAN1 meetings in Malta, Chicago,and Fukuoka during the first half of 2013 and the agreement was as follows: symmetric angular spread distribution and dual mobility corrections; direction of travel (velocity vector) independent and random; Doppler is determined by path AOA/AOD; uniform AoA spread of 104°.
3GPP D2D信道模型：2012年12月,一项关于LTE服务设备方面的研究项目成立了3GPP。2013年上半年在马耳他、芝加哥以及福冈RAN1会议上，提出了几种3GPP D2D信道模型,协商结果如下：对称角的扩散分布以及双重流动性修正;运动(速度矢量)的独立性和随机方向;多普勒由AOA /AOD路径决定;相同的AOA都以104°传播。
COST 2100 channel model: The COST modelling approach is as such not necessarily aimed
to D2D, because COST channel models are designed with one end of the link fixed [41]. The channel model could be composed of a set of randomly drawn clusters with all the parameters drawn from probability distributions which are extracted from channel measurements. Cluster would have visibility regions as in the COST model. Each cluster would be coupled to a subset of other clusters. If two radios enter visibility regions of coupled clusters the radio signal propagates interacting with the clusters. A consistent model has a proper number of `active' clusters with proper characteristics for each possible set of locations of transceivers. Some clusters can also be moving [43].
COST2100信道模型:COST模型方法不一定旨在D2D,因为COST模型设计的一端固定连接。信道模型可以由一系列集群和所有从信道测量中提取出来的概率分布参数随机组成。在COST信道模型集群有它的可见区域。每个集群将被耦合成其他集群的一个子集。如果两个收音机输入可见区域耦合的集群，无线电信号的传播将与集群交互。一致的模型在每个收发器的位置有适当数量的“活跃”特色集群。其中一些集群也是可以移动的。
The GBSM is derived from some predefined stochastic distribution of the scatterers/clusters by applying the fundamental laws of wave propagation. Such models can be easily adapted to diverse scenarios by modifying the stochastic distribution and properties of scatterers/clusters and the shape of the scattering region. GBSMs can be further classified into regular-shaped GBSMs (RS-GBSMs) [44一9] and irregular-shaped GBSMs (IS-GBSMs) [50, 51] depending on whether scatterers/clusters are placed on regular shapes, for example, two-sphere and two-cylinder, or irregular shapes. Its direct involvement of scatterers/clusters renders GBSM, one of the most promising candidates for D2D channel modelling. However, compared with MBPGM, GBSM is a bit more complicated, which blocks the development of D2D GBSM.
GBSM来源于一些遵从应用波传播基本规律而预定义的随机分布的散射/集群。这样的模型可以通过修改随机分布和散射的特性/集群以及散射区域的形状后很容易地适应不同的情况。几何随机模型根据散射/集群放置的常规形态，进一步分为形状规则的几何随机模型和形状不规则的几何随机模型，例如，二维球、二维圆以及不规则形状。直接参与的散射/集群呈现出GBSM,这是D2D信道模型最有前景的模型之一。然而,与MBPGM相比,GBSM有些复杂,这阻碍了D2D GBSM的发展。
4 D2D channel simulation and analysis 4 D2D信道的模拟与分析
In this section, we choose GBSM as an example to show some important channel properties for various D2D scenarios. Both RS-GBSM and IS-GBSM are considered in these simulations. Basic parameters used in simulations for RS-GBSM are as follows: both transmitter and receiver have two antenna elements, the carrier frequency is set to 5.25 GHz, because that some of traditional measurements on D2D channels for vehicular systems are in this band, and the scatterers are symmetrical around the transmitter and receiver, while basic parameters used in simulations for IS-GBSM are as follows: both transmitter and receiver have two antenna elements in the form of uniform antenna array, the centre frequency is 5.25 GHz, and the distance of the antenna is O.SA, different antenna distances are compared in the simulation of RS-GBSM, whereas in the simulation of IS-GBSM, we choose its mid-value to make more reasonable compansons.
在本节中,我们选择GBSM作为一个例子来展示D2D场景一些重要模型属性。这些模拟同时考虑了RS-GBSM和IS-GBSM。用于模拟RS-GBSM的基本参数如下:发射机和接收机都有两
个天线元素,载波频率设置为5.25 ghz,因为在D2D中用传统方法测量车辆系统也在这个分支中,散射对称地分布在发射机和接收机周围,而IS-GBSM用于模拟的基本参数如下:发射机和接收机都有两个形式统一的天线阵,中心频率为5.25 ghz,天线的距离是0.5 l,不同的天线距离在RS-GBSM模拟中被拿来相互做比较,而在模拟IS-GBSM时,为了做更合理的比较，我们选择它的中值。
4.1 D2D channel properties for RS-GBSM 4.1 D2D RS-GBSM信道属性
Based on Figs. 2-6, we can observe that the spatial correlation decreases with the increase of the distance between the transmitter and the receiver. As shown in Fig. 2a, the spatial correlation of three-dimensional (3D) two-sphere model is stronger than that of the two-dimensional two-ring and 3D two-cylinder models, which caused by the rich scatterers around the transceiver. In Fig. 2b, the spatial correlation of LoS case is stronger than the NLoS case, and the spatial correlation increases with the Ricean K-factor. From Fig. 3, it can be noted that the increasing of the scatterers radius around the transceiver directly causes the decreasing of spatial correlation. Because under the same scatterers distribution, when the scatterers radius is expanded, the scatterers density is decreasing and leads to low correlations. As illustrated in Figs. 4a and b, the distribution of elevation angle significantly influences the spatial correlation properties. Compared with the case of uniform distribution, because of the truncated Gaussian and Laplace distribution concentrate on the elevation angle in a narrow range, the spatial correlation decreases rapidly, and gradually levelling off, meanwhile, the spatial correlation of cosine distribution is relatively large. As demonstrated in Figs. 4c and d we can observe that when the angle spread is large, because of the multipath reflection and scattering, the angle spread at the receiving antenna is broadened, which caused the spatial correlation decreasing. As shown in Fig. 5a, the speed of the transceiver has significantly influences on the power spectrum density (PSD). Fig. 5b clearly proves that the high velocity of UEs causes obvious spread of PSD. While as Fig. 6 illustrates, the other factors such as the horizontal and elevation angle distribution and angle spread have less influence on the PSD properties.
基于图2 - 6,我们可以观察到空间相关性随着发射机和接收机之间距离的增加而增加。正如在图2中显示的那样,三维双球体模型的空间相关性比二维的双环和三维的两缸模型强,这种现象是由于收发器周围有丰富的散射。在图2 b中,视距情况下空间相关性的影响由于莱斯K因子系数的增加而强于非视距情况下。图3可以指出收发器周围散射半径增加直接导致空间相关性的减少。因为在相同的散射分布下，当散射半径扩大时散射密度减少,这导致相关性的降低。图4a和b说明仰角的分布显著影响空间相关性属性。与均匀分布的情况相比,由于在仰角方向上截断高斯拉普拉斯集中分布的范围，空间相关性迅速减小,并逐渐趋于平缓,与此同时,余弦分布有较大的空间相关性。图4 c和d表明，正如我们所观察到的那样，接收天线的多路径反射和散射导致空间相关性降低。图5a表明收发器的速度大大影响功率谱密度(PSD)。图5b显然证明了高速度会明显影响PSD的传播。图6说明其他因素如土地卧式锻机仰角分布和角度扩散对PSD属性影响较小。 4.2 D2D channel properties for IS-GBSM 4.2 D2D IS-GBSM信道属性
In IS-GBSM, the effect of the parameters on the statistical properties is limited. Owing to the randomness of the scatterers, we compare the results based on the same scatterers distribution. From Fig. 7a, it is clear that if we select a relatively large update time, the PSD would be changed
significantly. This is caused by the randomness of the scatterers and the time-varying modelling method. If we choose a small update time, the channel can be seen as wide sense stationary in the whole update time. We can know from Fig. 7b that the speed of the transceiver has obvious effect on the PSD properties. When the speed of the transceiver is faster, the impact on the Doppler correspondingly becomes larger, which will change the shape of the Doppler spectrum. As illustrated in Fig. 8a, the fixed scatterers have no significant effect on the PSD properties, the fixed scatterers randomly distribute around the transceiver, the distribution of the fixed scatterers does not substantially change the Doppler properties. As demonstrated in Figs. 8b and c, the density of the scatterers does not effect the PSD, because the impulse response is obtained in a small update time, the increase of the scatterers density is equivalent to doing the interpolation on the existing impulse response. Then the spatial correlation properties are illustrated in Figs. 9 and 10. The density of the moving scatterers has the most significant impact on the spatial correlation whereas the influence of other parameters such as the transceiver speed, the fixed scatterers distribution and density is not so obvious.
在IS-GBSM中，参数对统计特性的影响是有限的。由于散射的随机性,我们对有相同的分布散射结果作比较。从图7a中可以很明显看出,如果我们选择一个相对跨度较长的更新时间,PSD将显著改变。这是由散射的随机性以及时变模型导致的。如果我们选择一个小更新时间,信道可以被视为在整个更新时间内几乎是固定不变的。我们可以从图7 b知道收发器的速度对PSD的特性有明显的影响。当收发器的速度变快,对多普勒的影响也相应变大,这将改变多普勒频谱的形状。如图8中所示,固定散射PSD性质没有显著的影响,在固定散射随机分配在收发器周围，,固定散射的分布没有显著改变多普勒特性。图8 b和c说明,散射的密度不影响PSD,因为在一个小更新时间内获得脉冲响应，散射增加的密度相当于做现有脉冲响应的插值。接下来，空间相关性的属性在图9和10也展示出来。移动散射的密度对空间相关性的其他参数有很重要的影响,如收发速度、固定散射分布和密度不是那么显而易见。 4.3 Comparison between RS-GBSM and IS-GBSM 4.3 RS-GBSM和IS-GBSM的比较
From the simulation results, we can make a comparison between RS-GBSMs and IS-GBSMs. RS-GBSM can reflect the spatial correlation caused by the angle difference between the signal and the antenna array in the MIMO system, as well as the Doppler change caused by the speed of the transceiver, the horizontal angle distribution and angle spread. The IS-GBSM is based on the most real scatterers environment, which can also reflect the MIMO spatial correlation properties. The biggest advantage of IS-GBSM method is that it enables the analysis of the PSD properties caused by the speed of the transceiver, the horizontal angle distribution and angle spread. In general, RS-GBSMs are used for theoretical analysis of channel statistics and theoretical performance evaluation of D2D communication systems. Different from RS-GBSMs, IS-GBSMs intend to reproduce the physical reality and thus need to modify the location and properties of the effective scatterers of RS-GBSMs. Therefore IS-GBSMs are actually a greatly simplified stochastic version of ray-tracing method but still suitable for a wide variety of D2D scenarios by properly adjusting the statistical distributions of effective scatterers.
根据仿真结果,我们可以对RS-GBSMs和IS-GBSMs做个比较。RS-GBSM可以反映MIMO系统中由天线阵的信号和分布式天线系统引起的空间相关性的变化，以及由收发机的速度、水平角分布和角扩散改变引起的多普勒变化。IS-GBSM基于最真实的散射环境,从而也反映了天线系统的相关性属性。IS-GBSM方法的最大优点是,它能分析由收发机的速度,水平角分布
和角扩散改变引起的PSD属性。一般来说,RS-GBSMs用于D2D通信系统的频道统计和绩效评估理论分析。不同于RS-GBSMs ，IS-GBSMs打算再现现实,因此需要修改RS-GBSMs有效散射的位置和属性。因此IS-GBSMs实际上是大大简化后的随机版本的射线追踪方法,但通过适当调整有效散射的统计分布仍适合各种D2D场景。
5 Future challenges in D2D channel measurements and modelling 5 D2D信道测量和建模未来挑战
Recently, in many channel modelling related forums, workshops and conferences, for example, COST IC1004 forum, D2D channel measurements and modelling are very hot topics and attract more and more research interests. Currently, one common understanding is that none of the existing channel models comprehensively covers the D2D scenario with adequate accuracy in all dimensions.More importantly, current modelling approaches are not sufficient in describing the unique D2D channel characteristics. The challenges discussed in this section can be considered as guidelines for setting up future measurement campaigns and proposing more realistic D2D channel models.
最近,在例如COSTIC1004论坛等这样的与造型相关的论坛、研讨会上,D2D信道测量和建模都是非常热门话题,引起人们越来越多的研究兴趣。目前已有的共识是，现有的信道模型都不能全面涵盖在所有维度上有足够精度的D2D场景。更重要的是,目前的建模方法在描述D2D信道模型的特征时是不足够。本节中讨论的挑战可以视为建立未来测量活动和提出更现实的D2D信道模型的指南。
The first challenge is how to develop a general yet easy-to-use channel model for various D2D scenarios. As shown in Table 1,in general D2D communications have ten scenarios. In fact, the scenarios can be well beyond the listed ones if detailed propagation environments are taken into account. Currently, no channel model has the ability to cover all these scenarios. For better design of D2D systems and fair comparison of different D2D technologies, it is desirable to propose general D2D channel models for various scenarios. To this end, more measurement campaigns should be built up for different D2D scenarios, discovering unique D2D channel characteristics. Based on the observation and analysis from huge measurement data, one can develop a general D2D channel model by using geometry-based modelling approach or measurement-based pseudo-geometric modelling approach. Recently, 3GPP has made their efforts on the development of D2D channel models according to the modifications of WINNER model.
第一个挑战是对各种D2D场景如何开发一个通用而易用的信道模型。如表1所示,一般来说D2D通信有十个场景。事实上,如果考虑详细的传播环境，场景可以不止列出的这些。目前,没有信道模型能涵盖所有这些场景。为了更好地设计D2D信道系统以及对不同的D2D技术进行公平的比较,应该对各种场景提出更通用的D2D信道模型。为此,对于不同D2D场景应该建立更多的测量活动，以发现其独特的信道特点。根据对庞大测量数据的观察和分析，我们可以采用线性几何造型方法或伪几何计量造型的方法开发一个通用D2D信道模型。最近,3gpp根据WINNER的修改模型努力发展D2D信道模型。
Followed up by the above-mentioned challenge, another important challenge is how to implement a channel model that can run multiple D2D scenarios simultaneously. As shown in Table 1, D2D channels should rely on different scenarios to characterise different links. So far, 3GPP D2D channel model is the only model officially announced by 3GPP group and follows the basic modelling approach of WINNER model. Therefore the current implementation of the 3GPP D2D channel model does not allow simultaneous simulation of