High resolution parameter estimation for millimeter wave propagation in dynamic scenarios

未来的第五代移动无线电系统将使用毫米波频率范围内的频率，带宽高达几GHz。为了实现合理的链路预算，需要非常大（“海量”）的天线阵列。这些不可避免地具有高度指向性，必须以适应的方式进行调整。对于这类系统的性能预测和评估，目前还没有可用的信道模型。这主要是因为缺乏合适的信道测量，在移动站轨迹的动态情况下，信道测量可以确定电磁波在发射端和接收端的传播方向、传播时间、多普勒频移和极化方向，并具有必要的分辨率。因此，该项目的目的是研究基于最大似然原理的高分辨率参数估计器，该估计器解决了特定天线阵列架构背景下的这一问题。参数估计器的数据模型考虑了方向、飞行时间和多普勒不再以可因式形式存在的事实。这克服了传统的窄带假设，需要在时域中对阵列进行一致的描述，这涉及到天线元件的新设计范式及其与直径和天线间距有关的排列。为了满足高度时变场景的特殊挑战，必须使用贝叶斯滤波器对传播路径进行跟踪。为此，采用了雷达中多目标跟踪的方法，这些方法适合于本文提出的估计任务。因此，我们期望的优势，如降低计算复杂性，更高的精度，和解决歧义，这可能是由于特殊的阵列设计和不可分解的数据模型。这将伴随着一种实用的方法来适应模型顺序，评估传播路径的“生命周期”，并在短期阴影的情况下对其进行持续跟踪。在这种情况下，将开发新的方法，这将为渠道建模提供新的见解。其中包括描述波传播的镜面和漫射（不可分辨）分量的模型，它们的极化模型以及在方向、时间和多普勒维度上的动态关系。最后，我们的目标是集群的统计描述，这是传统的空间和时间分辨率方法无法确定的，但通道建模需要统计描述。

Future mobile radio systems of the 5th generation will use frequencies in the millimeter wave frequency range with bandwidths up to several GHz. To achieve a reasonable link budget, very large ("massive") antenna arrays are required. These inevitably have a high directivity, which must be adjusted in an adaptive way. For the prediction and evaluation of the performance of such systems, there are no channel models available. This is mainly because of the lack of suitable channel measurements, which would allow determining the propagation directions of the electromagnetic waves on the transmitter and receiver side, their propagation time, Doppler shift, and polarization orientation with the necessary resolution in dynamic scenarios along the trajectory of the mobile station. The aim of this project is therefore research towards a high-resolution parameter estimator based upon to the maximum likelihood principle, which solves this problem in the context of a specific antenna array architecture. The data model of the parameter estimator takes into account the fact that directions, time-of-flight and Doppler do no longer exist in a factorizable form. This overcomes the conventional narrow-band assumption and requires a consistent description of the array in the time domain, which involves new design paradigms for the antenna elements and their arrangement with regard to the diameter and antenna spacing. In order to meet the special challenges of highly time-variant scenarios, the propagation paths have to be tracked by Bayesian filters. For this purpose, methods of multi-target tracking in Radar are used, which are adapted for the estimation task presented here.As a result, we expect advantages, such as reduced computational complexity, higher accuracy, and resolutions of ambiguities, which may result from the special array design and the non-factorizable data model. This will be accompanied by a practical approach for the adaptation of the model order, the evaluation of the "lifetime" of propagation paths and their sustained tracking in case of short-term shadowing.In this context, novel methods will be developed, which will allow new insights for channel modeling. These include models for the description of specular and diffuse (non-resolvable) components of wave propagation, their polarimetric modeling and the dynamic relationship in the dimensions direction, time and Doppler. Finally, we aim at statistical descriptions of clusters, which cannot be determined by conventional methods of spatial and temporal resolution, but are required for channel modeling.