Wireless local area networks have very promising commercial potential to be widely de- ployed and used in a variety of application scenarios, but also suffer from a number of significant engineering difficulties. For example, small cell sizes and unlicensed spectrum lead to sophisticated spatial interaction, there is a lack of common air-interface to support mobile and time-sensitive applications, and interactions among competing users, both access points and mobile stations, could form surprising and unpredictable patterns. In addition, end-to-end transmissions in a wireless LAN often traverse a hybrid network, with wireless hops as well as wired links in the backbone. This presents unique challenges in cross layer TCP-IP- MAC-PHY interactions and offers new opportunities to optimize medium access control (MAC) and physical layer (PHY) algorithms to enhance the end user experience. There is no shortage of near term ideas from both industry and academia about how to improve the delivery of wireless Internet services. What is missing is an optimization framework that captures utility at the level of networked applications, and makes it possible to figure out whether any particular incremental change at the MAC or PHY layer is positive or negative. The gap is the focus of this SGER proposal. Kelly and Low have created the counterpart of this utility maximization framework in the wired world, and it has led to a deeper understanding of TCP congestion control and to the development and field testing of the new 'FAST TCP' protocol by the high energy physics community. The key insight in this approach is to start with a given network protocol and ask: 'If this distributed protocol is the solution to an underlying global optimization problem, what is that optimization problem?' The wireless world presents a very significant challenge and the new framework needs to capture spatial interaction within a cell, operation in unlicensed spectrum, a mix of contention and scheduling at the MAC layer, and the flexibility of achieving different rate-reliability tradeoffs by coding techniques. For example, we know that TCP throughput can be improved through coordination of spatial resources, specifically through space-time coding. Embedded diversity transforms spatial diversity into a fine grained resource that can provide opportunistic communication when the channel is good and reliable communication with latency guarantees when it is less benign. We expect to see significant benefits in specific cross layer designs based on a completed framework of wireless utility maximization, but the value of the proposal is a general proof of principle that can apply to an arbitrary wireless protocol or coding technique. Project Focus and Intellectual Merit: This SGER project is a proof of concept - the particular focus is on modelling the medium access mechanisms and leveraging the new coding paradigm of embedded diversity. The particular objective in the next 12 months of SGER investigation is to identify the global optimization problem to which coding (with embedded diversity) and scheduling (for medium access) is the solution. Preliminary results have recently been obtained by the PIs on tackling interference in utility maximization problems and on designing new space-time codes with embedded diversity. We believe that by filling in two remaining holes: modelling medium access and leveraging coding techniques, a complete and widely applicable framework of wireless network utility maximization will be completed at the conclusion of the SGER project. Broader Impacts: The research activities in this SGER proposal will be fully integrated with a variety of educational activities currently pursued by the PIs. These include curriculum developments of four new, inter-disciplinary courses at Princeton, two at undergraduate level and two at introductory graduate level, active supervision of undergraduate research projects on wireless LAN, and student summer internships being arranged with our industrial partners. A unique and important aspect of this proposal is a detailed arrangement of close collaboration with five major telecom and networking companies in the wireless LAN space: AT&T Labs, Cisco, Flarion Technologies, Intel, and SBC. These interactions with the industry will further strengthen the research and education components of the project and ensure visible impacts of the intellectual contributions. A workshop on wireless LAN research, development, and deployment will be hosted at Princeton by the PIs at the conclusion of this SGER project. We will compile a list of major research issues associated with wireless LAN and a list of industrial researchers or managers interested in these topics, which may be helpful to future NSF programs on topics related to mathematical models for wireless networks.
