Knowledge and Strategic Learning in Multi-user Communications

多用户无线通信系统形成竞争环境，异质用户和自利用户争夺有限的频谱资源。然而，最近主导多用户通信研究的技术并不能很好地适用于不同的环境，因为它们通常假设具有类似标准的收发信机，并被动地基于对竞争对手的完全了解或不了解来选择自己的行动？协议、实用程序等。这种无源系统设计没有利用用户的智能？？并且可能导致频谱使用效率低下。相反，这项研究描述并构建了多用户通信系统，其中用户参与主动交互以划分频谱。提出了一种新的多用户通信范式，其中用户之间的交互及其所产生的绩效不仅取决于他们调整其通信策略的能力，还取决于他们基于对竞争对手和环境的了解而做出关于信息交换的最优决策的能力。这项研究有两个主要的研究方向。首先，调查人员确定知识的价值，这是当用户和资源主持人对整个通信系统具有不同数量的知识和竞争用户进行交互时可以达到的性能界限。此外，还探讨了战略用户应如何主动积累知识并提高其效用。其次，研究人员通过系统地获取其他用户的信息并部署战略学习解决方案，使他们能够预测其他用户，从而构建能够接近这些性能界限的操作算法？响应，并最终优化其传输行动。学习的价值也被量化，它捕捉到了需要各种信息管理费用和复杂性成本的各种战略学习技术的绩效收益。

Multi-user wireless communications systems form competitive environments, where heterogeneous and self-interested users compete for the limited spectrum resources. However, the techniques that have recently dominated multi-user communication research are not well suited for heterogeneous environments, because they usually assume transceivers that have similar standards and passively select their actions based on either complete or no knowledge about the competitors? protocols, utilities etc. Such passive system designs do not take advantage of the users? ?smartness? and may lead to inefficient spectrum usage. In contrast, this research characterizes and constructs multi-user communications systems, where users engage in proactive interactions for dividing the spectrum. A new multi-user communication paradigm is proposed, where the interaction between users and their resulting performance is driven not only by their ability to adapt their communication strategies, but also by their ability to make optimal decisions about information exchanges based on their knowledge about their competitors and the environment. The research has two main research thrusts. First, the investigators determine ?the value of knowledge?, which are the performance bounds that can be attained, when users and resource moderators with different amounts of knowledge about the entire communication system and the competing users interact. Moreover, how strategic users should proactively accumulate knowledge and improve their utility is also investigated. Second, the investigators construct operational algorithms that can approach these performance bounds, by systematically acquiring information about other users and deploying strategic learning solutions that enable them to forecast the other users? responses and, ultimately, to optimize their transmission actions. The ?values of learning?, which capture the performance gains for various strategic learning techniques requiring various information overheads and complexity costs, are also quantified.