Avoidance or Alignment: Interference Mitigation Strategies for Wireless Communications

无线通信正在改变社会，使发展中地区在没有基础设施和消除障碍的情况下实现快速增长。超过10亿台移动设备(智能手机等)的激增代表着计算和电信的融合。虽然这类设备能够以非常高的传输速率运行，但实际上网络的运行速度要慢几个数量级。随着移动宽带数据流量预计在未来3年内增长18倍，这一差距正在扩大。拟议的研究计划研究关键的信号处理技术，以解决多用户干扰对频谱效率的限制。近期的研究工作旨在通过在时间、频率和空间上垂直组织传输来避免干扰。传统上，可以通过缩小网络小区大小来在空间上重复使用频率。然而，服务空间上不均匀的用户流量(热点)需要不同小区类型的重叠。由此产生的异质网络对正交性提出了挑战。多天线系统，称为多输入多输出(MIMO)系统，可以通过波束形成实现定向传输和/或通过空时编码传输对多个数据流进行多路复用。我们提出通过决策理论和优化来更好地识别未使用频谱的时间周期。然后，所获得的知识将用于避免干扰，并为新兴产品和标准带来巨大的潜在回报。首先，我们计划开发真正的双边序贯统计测试，以更有效地检测用户在时间、频率和空间上的出现和缺席。接下来，我们寻求通信开销和传感器观测能量消耗最小的分布式解决方案。一项平行调查将通过减少信令开销来解决提高频谱效率的问题。越来越多的先进方法利用信道状态信息，该信息通常被假定在短时间内是恒定的，即准静态的。在中短期内，我们提出了新的“预测性训练”方法，以应对可预测的信道变化。在这里，我们计划利用移动无线信号传输的物理特性，包括MIMO系统中的带限衰落和多普勒效应，来扩展和推广我们最近开发的信道估计方法。从长远来看，我们的目标是通过实现干扰对齐(IA)的先进信号处理来颠覆用于正交传输的干扰避免范例。这与每个干扰器花费空间维度(天线)的传统多用户干扰消除相反，或者与必须联合解码所有用户信号的多用户检测形成对比。在IA中，来自不同发射机的用户信号被处理以将干扰投影到降维子空间以恢复丢失的正交性。在低噪声水平的理想条件下，采用IA的系统有可能随着用户数的增加而成比例地增加吞吐量，即干扰成为一种好处！然而，IA的实际实现需要高精度的信道状态信息，包括定时和频率同步。我们计划研究性能限制，并提出在有限信噪比下的可实现的IA算法，该算法结合了信道估计、多定时和频偏恢复的新方法。在这里，我们计划利用我们最近在不完全同步的多个中继网的信道估计和训练方面的广泛研究，这与IA问题共享关键的“多路访问”特征。

Wireless communications is transforming society, enabling rapid growth in developing areas without infrastructure and removing barriers. The proliferation of over 1 billion mobile devices (smart phones, etc.) represent convergence of computing and telecommunications. While such devices are capable of operating at very high transmission rates, networks operate in reality at rates orders of magnitude slower. This gap is growing as mobile broadband data traffic is projected to increase by a factor of 18 over the next 3 years. The proposed research program investigates key signal processing techniques that address limitations on spectrum efficiency due to multi-user interference.Nearer term research efforts aim to avoid interference by organizing transmissions orthogonally over time, frequency, and space. Traditionally, frequencies may be spatially reused by shrinking network cell sizes. However, servicing spatially non-uniform user traffic (hot spots) requires the overlapping of different cell types. The resulting heterogeneous networks challenge orthogonality. Multiple antenna systems, known as multiple-input multiple-output (MIMO) systems, may achieve directional transmission via beamforming and/or multiplex multiple data streams via space-time coded transmission.In the shorter term, we aim to maintain user orthogonality and achieve more opportunistic and dynamic use of spectrum. We propose to better identify time periods of unused spectrum via decision theory and optimization. The knowledge gained would then be applied to interference avoidance with large potential payoffs for emerging products and standards. First, we plan to develop true two-sided sequential statistical tests to more efficiently detect both appearance and absence of users over time, frequency and space. Next, we seek distributed solutions with minimum communication overhead and sensor observation energy expenditure. A parallel investigation will address increasing spectrum efficiency by reducing signaling overhead. Increasingly, advanced methods exploit channel state information, that is typically assumed to be constant over short periods, i.e., quasi-static. In the short to medium term we propose new “predictive training” methods that account for predictable channel variations. Here, we plan to extend and generalize our recently developed approach to channel estimation by taking advantage of physical properties of mobile wireless signal transmission including bandlimited fading and Doppler effects in MIMO systems. In the longer term, we aim to disrupt the interference avoidance paradigm for orthogonal transmission by advanced signal processing that achieves interference alignment (IA). This is in contrast to conventional multi-user interference cancellation that costs a spatial dimension (antenna) per interferer, or to multi-user detection that must jointly decode all user signals. In IA, user signals from different transmitters are processed to project interference onto reduced-dimension subspaces to regain lost orthogonality. Under idealized conditions of low noise levels, systems employing IA have the potential to increase throughput proportionally with numbers of users, i.e., interference becomes a benefit! However, practical realization of IA requires highly accurate channel state information, including timing and frequency synchronization. We plan to study performance limitations and propose realizable IA algorithms at finite SNR that incorporate novel approaches to channel estimation, multiple timing and frequency offset recovery. Here, we plan to leverage our extensive recent research on channel estimation and training of imperfectly synchronized multiple relay networks, which shares key “multiple access” features with the IA problem.