ITR: A Cross-layer PHY/MAC Solution for Ad-hoc Networks with Multiple-element Arrays

Georgia Tech Research Corporation, Atlanta GA

Investigators

Abstract

Ad-hoc wireless networks lack the infrastructure of traditional wireless networks; for example, they have no base stations or switching centers. Every node in an ad-hoc network can act as a forwarder or router to the other nodes. While ad-hoc networks were conceived primarily for military and emergency relief applications, more recently, such networks are finding applications in regular wireless packet data environments because they offer convenient deployment, improved coverage, reduced energy consumption, and higher network capacity than traditional infrastructure networks. Concurrent with the expansion of ad hoc network applications, multiple-input multiple-output (MIMO) links have drawn tremendous interest for the extremely high data rates they can support in the absence of interference. A MIMO link has multiple antennas at the transmitter and multiple antennas at the receiver. The multiple transmit antennas are used to transmit multiple parallel streams of data in the same channel; signal processing on the outputs of the multiple receiver antennas separates the parallel streams. In the presence of interference, closed-loop MIMO (CL-MIMO), which requires that the transmitter have channel state information (CSI), is known to significantly outperform open-loop MIMO, which does not use CSI. This project investigates how MIMO can be used in ad-hoc wireless networks to increase their throughput. Previous work by the authors has shown that network throughput can be increased by up to 70% by allowing some CL-MIMO links to interfere with each other, in other words, by allowing space-division multiple access (SDMA) for CL-MIMO links. Control of the number of streams is the key to this problem, because the same parallel stream feature that enables a MIMO link to carry such high data rates also makes interference coming from a MIMO transmitter especially degrading to an unintended receiver. Such sophisticated antenna technology necessitates appropriate developments at the different layers of the network protocol stack. Previous work by the authors has shown that existing multiple access control (MAC) protocols, such as the MAC for the IEEE Standard 802.11, are incapable of attaining the maximum throughputs possible with SDMA and CL-MIMO links. The main objective of this project is to determine a cross-layer physical-layer/medium-access-control solution for ad-hoc networks with MIMO links. "Cross-layer" means that more than the traditional amount of information is shared between the physical and MAC layers. For example, a receiver can estimate, at the PHY layer, how many more streams (either desired or interfering) it can tolerate without being overwhelmed. This information can then be used by the MAC protocol to allow or disallow additional co-channel streams to be transmitted in the area. Specific physical layer research activities include study of (i) the dynamics of the joint-link adaptive process, including the effects of asynchrony of packets and the constraints imposed by the MAC protocol, (ii) various learning opportunities for the transmitter in order to speed convergence, (iii) how nonlinear receiver processor algorithms effect stream control, (iv) realistic channel models, and (v) alternative joint optimization algorithms. Also, the MAC layer research activities in this project are focused on (i) formulation of the medium access control problem in ad-hoc networks with CL-MIMO links, and identification of the key optimization considerations, (ii) design and development of medium access control algorithms and schemes that help achieve the benefits of stream-control, address its drawbacks in the target environment, incorporate the key optimization considerations, and use the unique flexibility offered by the PHY layer in its operation, (iii) molding the resulting MAC protocol to operate within practical and deployable protocol frameworks, and (iv) the development of algorithms through a prototype implementation that