【作者】 黄帆；

【导师】 杨大成；

【作者基本信息】 北京邮电大学 ， 通信与信息系统， 2011， 博士

【摘要】 正如以往每一代新系统出现时那样,为了达到更高的传输速率和服务质量,蜂窝移动通信系统在演进到第四代的过程中也采用了很多新技术。其中协作多点传输/接收技术(CoMP, Coordinated Multi-Point transmission/reception)备受瞩目。CoMP技术发展的推动力包括网络回传能力的日益强大、广泛应用的分布式天线技术、越来越小的蜂窝小区以及由此导致的日益严重的小区间干扰等等。协作通信基于多用户多入多出(MU-MIMO, Multi-User Multiple-Input Multiple-Output)和虚拟MIMO技术,应用于由多个基站(BS, Base Station)和多个移动台(MS, Mobile Station)组成的系统。现代蜂窝移动通信系统大多基于OFDM (Orthogonal Frequency Division Multiplexing,正交频分多址)技术,因而可以忽略小区内的干扰。为了进一步提高系统吞吐量,降低蜂窝系统固有的小区间干扰成为了一个重要的研究课题。传统的方法大多是为不同的站点分配不同的资源(如频率、时隙、空间资源),从而减轻相邻站点间的干扰。由于资源利用率高,如果小区间干扰控制得好,全频率复用可以获得更高的系统容量和空间利用效率。CoMP技术基于让协作基站复用所有时频资源的思想,让协作基站过来的潜在的干扰信号变成有用信号并协助传输。另外,CoMP技术还可以很好地加强对盲区的覆盖。MU-MIMO技术首次在蜂窝移动通信网络中得以应用,得益于预编码技术的成熟以及多天线的广泛应用。多个用户的数据信号经过预处理后,在多个相互正交的子空间中传输,相互之间不会有干扰。通过虚拟MIMO技术,多个基站联合起来,共同组成虚拟MIMO天线阵,一起为用户服务,则会有空间分集的效果。本论文的研究和学术贡献都基于上述CoMP系统,并且涉及到了四个研究点,其中每个研究点都在论文当中有详细的论述。第一个研究点是多小区天线校准问题。多个协作BS需要共享MS的信道信息。以下行链路为例,为了进行预编码,协作BS需要了解准确的下行信道信息。一般来说,BS可以为此采取两种方式：通过MS的反馈；利用TDD (Time Division Duplex,时分双工)系统中上下行链路的信道互易性,由上行信道精确估计出下行信道信息。当上下行信道传输的信号占用同一个频带时,如果上下行信号的时间间隔小于信道的相干时间,那么可以认为它们在空间中经历的衰落相同。因此信道互易性得以实现。后一种方法不需要反馈并且没有时延。从这个角度看,在TDD系统中采用CoMP具有天然的优势。实际上,信道互易性只能适用于无线空间传播信道。在硬件实现方面,发射机和接收机中的非理想模拟电路会对信号进行非线性放大,其效果参数与硬件本身的工艺水平和周围的环境(如温度、湿度等等)密切相关。所以电路的参数可以说是时变的。这等效于在信号处理方面给传输信道引入了失真。此外,它对上下行信道的影响通常不同,这一点可能影响信道的互易性。在实际的系统中,BS和MS的每一根天线都有独立的发送和接收链路,可以将其视为模拟放大器。这样,在信号处理的时候,每一根天线的信号都要乘以一个复放大系数——它将改变传输特性,以此来模拟对链路的影响。这个系数随环境的改变而改变。因此就需要进行“天线校准”来消除发送和接收链路之间的失配,即人为引入额外的系数,补偿物理信道的不完美,使发送和接收链路的特性一致,从而保证信道互易性。以往的研究主要针对单个小区内部的失配和校准。这类天线校准只适用于在单个BS或MS的多根天线之间进行。与之不同,在CoMP系统中,由于一组BS间需要联合处理,因此还要考虑多个BS或MS之间的失配问题。相应的,CoMP系统中的天线校准问题也比传统系统中复杂。这个研究点正是研究CoMP系统中的天线校准问题。论文建立了一个数学理论模型,给出了天线校准问题的数学描述。此外,论文也分析了CoMP系统BS端和MS端因天线未校准带来的影响,以及在线性预编码过程中校准的一些性质。论文提出了一个切实可行的新方案,它可以在一组基站内实现基站间校准。这个方案需要的反馈量较小,同时保持了较好的可行性。第二个研究点是在单基站功率约束条件下的多小区最优下行波束赋形算法。这个问题的研究目标通常是使多小区下行吞吐量最大化,这是一个受限的非凸最优化问题。解决这个问题需要较多的技巧。迄今为止,学术界对单小区多天线无线系统中的下行波束赋形问题已经进行过大量的研究。解决这类问题的一个主要工具是上下行链路的对偶性。有的文献利用拉格朗日对偶理论证明,在波束赋形问题和容量区域问题中,单天线功率受限的多天线下行信道的对偶是带有对角受限的不确定噪声的上行信道。针对多小区协作系统的类似问题的研究不多,其研究目标是使多小区系统吞吐量最大化。有文献提出了一些贪婪算法,它在图形化的结构中,将降低干扰和资源分配直观地分开以解决问题。有文献考虑了在多小区环境中SINR (Signal to Interference and Noise Ratio,信号干扰噪声比)受限的条件下上下行波束赋形的对偶性,证明了在每个子信道只对应一个用户的多入单出(MISO, Multiple-Input Single-Output)系统中最优下行波束赋形问题等同于对偶上行系统的最小均方误差(MMSE, Minimum Mean Square Error)波束赋形问题。它同时还证明了对偶上行问题和原始的下行波束赋形问题能够达到同样的最大吞吐量。论文考虑的对偶上行问题是一个最小最大组合最优化问题。因为这个对偶问题是非凸最优化,所以难以找到全局最优解。论文在这一研究点的主要贡献是提出了一种高效的新算法,通过寻找满足对偶问题最优化必要条件的解,提高系统的和速率。数值仿真的结果还验证了该算法的收敛性。第三个研究点是对BD (Block Diagonal,块对角化)预编码低复杂度实用算法的研究。BD算法是MU-MIMO系统中一种有效的预编码技术,对该系统的应用起了非常关键的作用。在这种预编码算法里,MU-MIMO信道被分解成若干个并行独立的单用户MIMO信道,每个MS的预编码向量(或矩阵)都在其他MS的信道矩阵零空间里面生成。因此,可以在发送端预先消除用户间的干扰。但是,在实际应用时,以往的算法当中,BD预编码都需要对信道矩阵进行特征值分解(SVD, Singular Value Decomposition)运算。当信道矩阵的维数较大时,如在协作传输的场景中,基于SVD的BD算法的计算复杂度将非常大。实际上,移动台的接收天线数通常较小,因为它的体积受限。在当前的B3G演进中,2根接收天线几乎是下行的标准配置。有研究表明,每个MS仅配置了2根天线的协作多点传输/接收系统能够满足国际电信联盟(ITU, International Telecommunications Union)对IMT-Advanced系统上、下行频谱效率的目标要求。基于MS端只有2根接收天线的前提条件,论文提出了一种低复杂度方案以在BD预编码时替代SVD过程。该方案计算出来的特征向量的性能效果与SVD的结果完全一致,而计算复杂度却有明显的降低。例如,当发送天线数为20时,本方案计算特征向量和整个方案的运算量分别是1500和53000次浮点运算,而基于SVD的方案则分别需要16000和78000次浮点运算。第四个研究点是对CoMP系统中协作簇划分方法的研究。在实际系统中需要考虑信令开销问题。在CoMP系统中,多个基站联合起来同时服务若干个用户。若系统采用集中式的信号联合处理,则各基站需要与中央处理器交互大量的数据以及信道信息；若系统采用分布式的信号处理,则各基站之间需要相互交互这些数据和信道信息。由于在未来移动通信系统中的数据传输速率大大提高,所以这个交互的数据量是非常庞大的。因此,为了减少开销,应当限制协作的基站数,但也同时会降低CoMP的效果。论文的研究目标是设计一种折衷的方案,能够在尽可能地增强协作能力的同时尽量降低系统复杂度,也可以实现网络的平滑演进。CoMP系统协作簇的划分是传输增益大小的重要影响因素,也是一个热门研究点。本文只考虑基站端的协作。论文把相互协作的一组基站称为“协作簇”,它们共同服务一组MS。本研究点的系统模型采用在实际网络中得到广泛应用的扇区化的六边形小区结构,并且采用较成熟的静态的分簇方法。论文研究了集中式和分布式协作两种情形。集中式的发送天线都配置在同一个站址,而分布的发送天线则分布在多个站址当中。目前公认的效率较高并且实现简单的静态分簇方法是把相邻的三个扇区划分成—个协作簇。本论文提出了一种半动态的分簇方法,即“多层协作簇”分簇方法,弥补了静态分簇法的缺陷,同时复杂度的提升也不大。针对集中式方法,论文也有相应的改进方案,以最大限度地提高系统性能。另外,方案中还考虑了资源分配问题,也使得本方案相对于传统方法更实用。更多还原

【Abstract】 Just the same as each new generation system, to achieve higher transmission rate and better QoS (Quality of Service) performance, cellular mobile communication system adopts many latest technologies when it evolves to the 4th generation currently. Among them, Coordinated Multi-Point (CoMP) Transmission/Reception technology attracts lots of attentions. Some of the backgrounds to apply CoMP are powerful backhaul transmission ability, widely used distributed antenna technology, shrinking cellular size and severe inter-cell interference. Based on multi-user multiple-input multiple-output (MU-MIMO) and virtual MIMO techniques, coordinated communication can be realized within multiple base stations (BSs) and multiple mobile stations (MSs).The cellular mobile communication systems nowadays are mostly based on orthogonal frequency-division multiplexing (OFDM) technology so that intra-cell interference can be neglected. In order to further improve the system throughput, to lower inter-cell interference, which is inherent to the cellular systems, is one of the main issues for research. Traditional approaches mostly mitigate interferences from neighboring sites by partitioning resources, e.g. frequency, time slot, or space resources, for different sites. However, full spectral reuse can allow the system to achieve higher network capacities and spatial resource utilization efficiency if the inter-cell interference is effectively managed. CoMP is based on the transmission of cooperative BSs by reusing the entire frequency spectrum, where the potential interference across cooperative BSs will act as assist transmission. Additionally, CoMP can strengthen the covering of the dead zone.Thanks to mature precoding techniques and widely application of multiple antennas, MU-MIMO is exploited in cellular network for the first time. The precoded signals of multiple users served by the same BSs are transmitted in orthogonal subspaces without interfering each other. With virtual MIMO techniques, cooperative BSs communicate with users via a virtual MIMO antenna array to take the advantage of the spatial diversity.The research background and the contributions of this thesis are based on CoMP system. Four research points are envolved and each of them is detailed discussed in the thesis.The first research point is multi-cell antenna calibration problem.Channel information of MSs is required to be shared among multiple coordinated BSs. Take the downlink for example. In order to precode for MSs, the cooperative BSs need to be aware of the exact downlink channel information. Generally, there are two ways for BSs to achieve this:through the feedback from the MSs; by predicting the exact downlink channel information from the uplink using the channel reciprocity of downlink and uplink in time-division duplex (TDD) system. Channel reciprocity can be realized due to the fact that when occupying the same frequency band, signal experiencing the same fading in the space between downlink and uplink channels if the time interval between them is less than the channel coherence time. The latter one doesn’t need feedback and is delay free. From this point of view, there is a natural advantage to apply CoMP in TDD system.Actually, channel reciprocity can only apply to wireless space propagation channel. In hardware realization aspect, the analog circuits in transmitter and receiver would act as non-linear amplifiers to the signal, and the coefficients of this effect are closely related to the state of the art of the hardware itself and properties of the surroundings (e.g. temperature and humidity and etc.). So the coefficients of the circuits are time-variant. This effect is equivalent to adding a distortion to the transmission channel in terms of the signal processing. In addition, the effects usually would not be the same to downlink as that to uplink channel, so that the channel reciprocity might be affected.In realistic system, every antenna located in either BS or MS has independent receiver and transmitter chains which can be seen as analog amplifiers. Thus, a complex magnification coefficient will be multiplied to each antenna which would alter the property of the transmission to simulate the effect of each chain. This coefficient varies according to the environmental changes. A procedure, therefore, which is called "antenna calibration", should be performed to compensate the mismatch between the receiver and transmitter chains, i.e., to bring in extra coefficients factitiously to reconcile the properties between receiver and transmitter chains to assure the validity of the channel reciprocity.Traditional works mainly focused on the mismatch and calibration within a single cell. In single cell case, calibrations are performed only among multiple antennas in each single BS or MS. However, mismatches among multiple BSs or MSs may also need to be concerned due to joint processing of a group of BSs in CoMP system. Accordingly, the antenna calibration problem in CoMP is even more complicated than traditional system, and it is what this research point focuses in.In this thesis, a theoretical model is set up and a complete overview of the antenna calibration problem is given.Besides, the impact of un-calibrated cases in BSs and MSs and corresponding properties of calibration from the perspective of linear precoding in CoMP system is specifically analyzed. A novel practical concept is proposed which can realize inter-BS antenna calibration within a group of BSs. In this scheme, the feedback amount is minimized while it can still achieve high practicality.The second research point is multi-cell optimal downlink beamforming algorithm with per-base station power constraints.The objective of the problem is usually multi-cell downlink throughput maximization, which is a constrained nonconvex optimization problem and it needs some skills to solve this problem.So far, there have been plenty of works on the downlink beamforming for multi-antenna wireless systems in the past. A main tool to solve these problems is uplink-downlink duality. Authors in used Lagrangian duality to prove that for both the beamforming problem and the capacity region problem the duality of a multi-antenna downlink channel with per-antenna power constraints is an uplink channel with an uncertain and diagonally constrained noise.There are not many works on coordinated multi-cell downlink beamforming problem, which aims to maximize the throughput of the systems. Some greedy algorithms were proposed in the reference, which considered this problem under a graphic framework and solved the problem by separating interference reduction and resource allocation intuitively. In a reference, the uplink-downlink beamforming duality with SINR (Signal to Interference and Noise Ratio) constraints in the multi-cell environment was considered. In that work, it was shown that the optimal downlink beamforming in MISO (Multiple-Input Single-Output) system with single user per sub-channel can be the minimum mean squared error (MMSE) beamforming in the dual uplink. It also proved that the dual uplink problem and the primal downlink problem achieve the same optimal throughput.The dual uplink problem considered in this paper is a min-max mixed optimal problem. Since the dual problem is a nonconvex optimization, it is not easy to compute the globally optimal solution. The main contribution of this paper is a novel efficient algorithm for improving the sum rate of the system through finding the solution satisfying the necessary optimality conditions of dual problem. Numerical results show the convergence of the proposed algorithm.The third research point is the study of low complexity implementation of block diagonalization (BD) precoding.BD algorithm is one of the key precoding techniques for downlink MU-MIMO system. In this precoding algorithm, MU-MIMO channel is decomposed into multiple parallel independent single-user MIMO channels and precoding vector (or matrix) for a MS is computed in the null space of other MSs’channel matrices so that interference between data transmitted to these MSs is cancelled.However, the BD precoding in the previous researches requires implementations of SVD (Singular Value Decomposition) of channel matrices. When the dimension of channel matrix is large, for example, in the cooperative transmission scenario, the SVD based BD will suffer high computational complexity.In practice, the number of receiving antennas of the user equipment is usually limited due to size restraint. In the current evolution of 3G, two-receiving-antenna is a main antenna configuration of the downlink. Studies also show that when each user is equipped with only two antennas, system can achieve the target of ITU (International Telecommunications Union) downlink/uplink spectral efficiency requirement for the International Mobile Telecommunications-Advanced (IMT-Advanced) by the CoMP transmission and reception system.Motivated by this, in this thesis, a low complexity scheme is provided to substitute the procedure of SVD in BD precoding when the number of receiving antennas of MS is two. The proposed method for computing eigenvectors reaches the same effect as the SVD while its computational complexity is significantly reduced. For example, when the number of transmit antennas is 20, computational amounts of the proposed method for computing eigenvectors and the corresponding entire scheme are about 1500 and 53000 flops, respectively, while they are around 16000 and 78000 flops, respectively, in the SVD based scheme.The fourth research point is the clustering approach in CoMP system. In realistic systems, affordable amount of the overhead signals should be taken into consideration. While centralized jointly processing scheme may demand large amounts of interactive information between the central scheduler and each BS, distributed cooperative scheme requires communications among BSs. Due to the fact of significant improvement on data rate in the future mobile network, information exchanged in both centralized and distributed scheme will grow dramatically.Therefore, to reduce the overhead, the number of cooperative BSs will be limited and it will restrict the performance of CoMP. Our interest is in compromising cooperation approaches that maximize the capability of cooperation as much as possible meanwhile lower the complexity of system in order to achieve a smooth network evolution.Cooperation clustering is one key point to the transmission gain, which attracts lots of attention. The thesis only focuses on BS cooperation cluster, which is referred in particular as BS cooperation cluster in the followings. A BS cooperation cluster is defined as a set of cooperative BSs that serve the same group of MSs.In this thesis, the widely used network structure is adopted, i.e., sectorization hexagonal cell, and static clustering concept to investigate on centralized and distributed approaches. In centralized approach, transmit antennas are assembled in one site (or BS); while in distributed approach they are located in different sites. An acknowledged realization of static clustering is the cooperation among three nearby BSs. A semi-dynamic clustering approach in distributed clustering scheme that is named as "multi-layered clustering" is proposed, which aims at remedying the cooperation defects in static clustering with relatively low increment on complexity. For the centralized scheme, an evolution path is also provided to further improve the system performance. In addition, resource allocation is discussed and it will make our approach more feasible compared with the traditional one.更多还原