Coping with Uncertainty in Mobile Wireless Networks
Sajal Kumar Das, Christopher James Rose · Kluwer Academic Publishers eBooks · 2006
With the availability of inexpensive wireless devices including sensors, and ongoing emphasis on greater integration of components, extremely large-scale wireless networks of interconnected mobile devices are inevitable in near future. It is also envisioned that almost all these varied devices will require some form of Internet access. Unfortunately, the uncertainty associated with wireless mobile networks produces unique challenges to achieving seamless integration with the Internet while provisioning end-to-end quality of service (QoS). In particular, the uncertainty in wireless channels as well as node mobility, and hence network topology, can bedevil protocols more suited to a “classical” Internet structure. Therefore, new protocols have to be designed that must be (i) robust against the uncertainty in traffic load, host mobility, resource availability and wireless link characteristics; (ii) adaptive to the network dynamics, thus making learning and prediction integral components in the design methodology; and (iii) intrinsically on-line so as to make realtime decisions based on temporal and spatial information. In order to cope with uncertainty in wireless mobile networks, we propose an overarching theoretical framework for representing relevant network information in terms of underlying entropies, entropy rates and their interrelationships. We will demonstrate how to apply information theoretic learning and prediction tools for collection and dissemination of network state information that can be used for robust and adaptive protocol design. Specifically, we will investigate the applicability of this novel framework in designing optimal mobility tracking and resource management, and also coping with uncertainty in traffic load, topology control and routing. These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.