Wireless networks have had a profound impact on the way people work and live. With the advent of mesh networks and WiMax, wireless multihop networks could have a similar impact by providing the last piece of a ubiquitous networking infrastructure. However, wireless multihop networks are subject to a capacity limit, as set forth in the classic work by Gupta and Kumar. Of different technologies proposed for throughput enhancement, the use of MIMO links is especially promising. This project considers how to use MIMO resources optimally to achieve network-wide performance goals. MIMO resources can be used to increase single-link performance through array and diversity gains and spatial multiplexing of streams, or to activate otherwise conflicting links simultaneously through interference suppression. The proposed research characterizes the inherent tradeoffs among these capabilities. The PIs are addressing the problem through both formal optimization techniques and algorithm design and analysis.
Algorithms for evaluating feasibility of a set of MIMO links, for scheduling streams across a MIMO network, and for routing of flows are being developed. Our problem formulations account for important MIMO characteristics that were ignored in prior work on MIMO networks. From these more accurate models, the PIs are designing MIMO-aware algorithms that will enable major throughput gains. In addition to the formal analyses and new algorithms that will result from this research, they are integrating MIMO models and algorithms into the ns-3 network simulator. This is expected to spur additional research into the benefits of MIMO from the broader wireless networking community.
Multiple-input, multiple-output (MIMO) communication involves the use of multiple antenna elements at both the transmitter and receiver of a wireless communication. The multiple communication paths that result from MIMO can be used for various purposes, e.g. increasing signal strength by exploiting path diversity, achieving higher data rates through parallelism (multiple communication streams), and cancelling interfering signals coming from competing transmissions. The primary goal of this project is to develop and evaluate scalable techniques for optimizing overall wireless network performance by exploiting these capabilities of MIMO links, with a primary focus on interference cancellation. The two major activities carried out under this project were: 1) design, analysis, and experimental evaluation of new algorithms related to MIMO interference cancellation, and 2) implementation of a MIMO software-defined radio testbed for experimenting with new MIMO algorithms and techniques. Significant outcomes resulted from both of these activities. Novel MIMO algorithms were designed that substantially improved the state-of-the-art in MIMO communication and networks, a MIMO testbed was developed and used to evaluate MIMO processing techniques, and the testbed technology was transitioned for use within the Department of Defense. The project also contributed to the educational development of four graduate students, three of whom received their PhD degrees during the period of the award and two of whom are underrepresented minorities. Some technical highlights of the project include: - identification of two types of MIMO interference cancellation (IC) techniques, referred to as unilateral IC and bilateral IC - comparison of unilateral IC and bilateral IC, which demonstrated the clear superiority of bilateral IC from the perspective of aggregate performance - development of a new unilateral IC technique that handles cyclic link dependencies and has a performance within about 25% of bilateral IC, compared to the more than 50% performance degradation of prior unilateral IC approaches - development of a novel bilateral IC algorithm that jointly optimizes MIMO weights and numbers of communication streams allocated to links, which achieves higher throughput than prior approaches and is able to converge quickly to good solutions under high interference conditions - evaluation of the unilateral IC feasibility problem, proving that the general problem is NP-complete but identifying several special cases of the problem that can be solved efficiently - development of a high-accuracy time and frequency offset estimation algorithm for OFDM MIMO that avoids large time and frequency spreads by using frequency-domain symmetric correlation.
