CIF: Small: Analog Networking: Distributed Source-Channel Approaches to Delay and Resource Constrained Communications

严重的延迟和资源（能源，带宽）的限制是典型的新兴通信和网络应用在各种科学和工程学科。这在高度受限的传感器网络中尤其如此。这样的应用强烈地促使放弃数字通信的便利性，重新考虑潜在的好处，甚至模拟通信的必要性，尽管是在分布式信源信道编码和路由的现代背景下。主要的挑战出现，包括从估计和信息论原理推导模拟网络的理论基础，专门研究涉及分布式源信道编码的各种子问题和网络场景，以及开发有效的实用算法方法，以及由于总体目标函数的高复杂性而带来的令人生畏的优化挑战optima。一般模拟网络问题是根据源节点（编码器）处的源变量和通道变量之间、中间网络节点处的接收变量和传输变量之间以及汇节点（解码器）处的接收变量和重建变量之间的模拟映射来制定的。研究是追求沿着几个具体的调查路线，包括推导延迟约束的性能界限和最优性条件，最佳和低复杂度的算法设计，全局优化技术，模拟映射的分布式源-信道编码，模拟多描述编码，扩展到源和通道的内存，和模拟网络中的最优路由。这项工作建立在一系列初步工作和结果的基础上，包括在简单设置中这种映射的最优性的必要条件，找到局部最优映射的迭代算法，以及已成功用于相关优化问题的强大的非凸优化工具（确定性退火）。

Severe delay and resource (energy, bandwidth) constraints are typical in emerging communications and networking applications within a variety of scientific and engineering disciplines. This is particularly the case in highly constrained sensor networks. Such applications strongly motivate foregoing the convenience of digital communication and revisiting the potential benefits, and even necessity of analog communication, albeit in the modern context of distributed source-channel coding and routing. Major challenges emerge, including the derivation of a theoretical foundation for analog networking from estimation and information theoretic principles, specializing to a variety of subproblems and networking scenarios involving distributed source-channel coding, and developing effective practical algorithmic approaches, as well as a formidable optimization challenge due to the high complexity of the overall objective function which is literally riddled with poor local optima.The general analog networking problem is formulated in terms of analog mappings, between source and channel variables at source nodes (encoders), between received and transmitted variables at intermediate network nodes, and between received and reconstruction variables at sink nodes (decoders). Research is pursued along several specific lines of investigation, including derivation of delay-constrained performance bounds and optimality conditions, optimal and low complexity algorithm design, global optimization techniques, analog mappings for distributed source-channel coding, analog multiple descriptions coding, extensions to sources and channels with memory, and optimal routing in an analog network. The effort builds on a body of preliminary work and results, including the necessary conditions for optimality of such mappings in simple settings, iterative algorithms to find locally optimal mappings, as well as a powerful non- convex optimization tool (deterministic annealing), which has been successfully employed in related optimization problems.