Improving QoS and Management in Multi-Hop Wireless Networks.

Abstract

To meet the ever-increasing demand for wireless access to the Internet, multi-hop wireless networks have been introduced to provide fast, easy, and inexpensive wireless last-mile connectivity. By using “free” wireless links, multi-hop networks can easily be deployed to cover large areas or are being adopted for many emerging applications, including home networks and city-wide wireless clouds. On the other hand, a recent measurement study has shown that due to heterogeneous and changing environments, multi-hop networks often provide poor Quality-of-Service (QoS) and/or incur high management costs. Network users often experience fluctuation in end-to-end performance, caused by intermittent interferences or new obstacles. Furthermore, network engineers need to put considerable management efforts into identifying the interference sources or manually reconfiguring the network settings or the physical locations of nodes.

This thesis proposes novel system-level monitoring and adaptation techniques, through which multi-hop wireless networks can significantly improve user-perceived QoS and reduce human-involved network management efforts. First, it focuses on the link-quality fluctuation problem and presents distributed link-quality monitoring techniques that allow networks to accurately measure time-varying wireless link-conditions. Second, it investigates the network reconfiguration problem and proposes localized planning algorithms that enable networks to autonomously reconfigure network resources (e.g., channel, interface) on demand. Next, the thesis explores the spatial dimension of networks and proposes a mobile autonomous router system that optimizes physical placement of network nodes in heterogeneous deployment environments by using robot-based mobility. Finally, the thesis investigates the spectrum monitoring and management problem in the chaotic deployment of current wireless networks and presents a robot-based spectrum site-survey system that automates a labor-intensive spectrum monitoring task to improve both QoS and management costs.