CAREER: Efficient Distributed Resource Allocation Algorithms, with Application to Wireless Communications

尽管确实需要精通算法、经济和管理等领域的工程师，但几乎没有机会将这些领域添加到本已拥挤的工程课程中。因此，工程教育工作者有责任使现有课程多样化，尽可能地纳入这些主题。因此，有趣的是，某些类型的工程问题自然而然地引起了这些问题。例如，在网络和通信领域，经常会遇到这样的系统，它们将受益于分散经营，甚至可能是完全分散的或市场型的“经营”，在这种经营中，分散的行为者被允许独立运作，不受中央强加的限制。然后，工程师必须确保“市场”导致期望的(例如，公平的或更常见的，高效的)操作解决方案。该项目涉及无线通信网络的研究，并将算法设计、经济和管理方面的问题整合到工程课程中。功率控制是现代无线通信网络的重要组成部分，在基于IS-95或码分多址的网络中至关重要。这个项目的研究针对的是功率控制系统，与传统的功率控制系统不同，它不遵循接收功率平衡或信干比平衡算法。相反，ne3w系统的目标是在给定服务质量的情况下将能耗降至最低。它们还与它们的前辈不同，因为它们动态和协作地进行优化-当干扰级别足够高时，发射机开始降低其功率。相比之下，基本上所有现有的功率控制系统都要求节点在面临增加的干扰时增加功率，在某些情况下达到所有相互干扰节点的饱和点。基于新方法的功率控制系统在提供更好的稳定性和相当的吞吐量方面具有显著优于传统系统的潜力，而平均功率要低得多。虽然好处是全系统范围的，但从发送方的角度来看，导致这些好处的行动纯粹是自私的，不需要与其他节点进行明确的协调。这项研究旨在利用这些原理开发改进的功率控制算法，更广泛地说，开发适用于各种工程问题的高效分布式资源分配算法。

While there is a definite need for engineers well versed in areas such as algorithms, economics, and management, there is little opportunity to add these areas to an already crowded engineering curriculum. Hence it is incumbent upon engineering educators to diversify the existing curriculum to incorporate these topics wherever possible. It is of interest, then, that certain kinds of engineering problems give rise to these subjects naturally. For example, in networking and communications it is not uncommon to encounter systems that would benefit from decentralized operation, and possibly even fully distributed or market-type" operations, in which the distributed actors are allowed to perform independently and free of centrally imposed constraints. The engineer must then must ensure that the "market" leads to the desired (e.g., fair or, more commonly, efficient) operational solution. This project involves research in wireless communication networks and the integration of algorithm design, economic, and management issues arising therein into the engineering curriculum. Power control is an important element in modern wireless communication networks, and is essential in networks based on IS-95 or code division multiple access. This project's research is directed toward power control systems which, unlike conventional power control systems, do not follow a received-power balancing or signal-to-interference ratio balancing algorithm. Instead, the ne3w systems target minimum energy consumption for a given quality of service. They also differ from their predecessors because they optimize dynamically and cooperatively - a sufficiently high interference levels, a transmitter begins to decrease its power. By contrast, essentially all existing power control systems call for nodes to increase power when faced with increased interference, in some cases to the point of saturation at all mutually interfering nodes. Power control systems based on the new approach have the potential to outperform conventional systems significantly, in terms of offering better stability and equivalent throughputs at considerably lower average powers. While the benefits are system-wide, the actions that lead to them are purely selfish from the standpoint of the transmitter, and require no explicit coordination with other nodes. This research aims to exploit these principles for the development of improved power control algorithms and, more generally, for efficient distributed resource allocation algorithms applicable to a variety of engineering problems.