Geographic location and geographic prediction performance benefits for infrastructureless wireless networks
Author(s)
Fink, Shane A. (Shane Andrew)
DownloadFull printable version (6.104Mb)
Other Contributors
Massachusetts Institute of Technology. Dept. of Electrical Engineering and Computer Science.
Advisor
Vincent W. S. Chan and Christopher C. Yu.
Terms of use
Metadata
Show full item recordAbstract
The field of infrastructureless wireless networks (IWNs) is a broad and varied research area with a history of different assumption sets and methods of analysis. Much of the focus in the area of IWNs has been on connectivity and throughput/energy/delay (T/E/D) tradeoffs, which are important and valuable metrics. When specific IWN routing protocols are developed, they are often difficult to characterize analytically. In this thesis we review some of the important results in IWNs, in the process providing a comparison of wideband (power-limited) versus narrowband (interference-limited) networks. We show that the use of geographic location and geographic prediction (GL/GP) can dramatically increase the performance of IWNs. We compare past results in the context of GL/GP and develop new results in this area. We also develop the idea of throughput burden and scaling for the distribution of topology and routing information in IWNs and we hope that this work provides a context in which further research can be performed. We primarily focus our work on wideband networks while also reviewing some narrowband results. In particular, we focus on wideband networks with non-zero processing energy at the nodes, which combines with distance-dependent transmission energy as the other main source of power consumption in the network. Often the research in this area does not take into account processing energy, but there is previous work which shows that processing energy is an important consideration. The consideration of processing energy is the determining factor in whether a whisper to the nearest neighbor (WtNN) or characteristic hop distance routing scheme is optimal. Whisper to the nearest neighbor routing involves taking a large number of short hops, while characteristic hop distance routing is the scheme by which the optimal hop distance is based on the distance dependent transmission energy and the processing energy, as well as the attenuation exponent. For a one-dimensional network, we use a uniform all-to-all traffic model to determine the total hop count and achievable throughput for three routing types: WtNN without GL/GP, WtNN with GL/GP, and characteristic hop distance with GL/GP. We assume a fixed rate system and a random and uniform node distribution. The uniform all-to-all traffic model is the model where every node communicates with every other node at a specified rate. The achievable throughput is the achievable rate at which each source can send data to each of its destinations. The results we develop show that the performance difference between WtNN with and without GL/GP is minimal for one-dimensional networks. We show the reduction in hop count of characteristic hop distance routing compared to WtNN routing is significant. Further, the achievable throughput of characteristic hop distance routing is significantly better than that of WtNN networks. We present a method to determine the link rate scaling necessary for link state distribution to maintain topology and routing information in mobile IWNs. We developed several results, with the main result of rate scaling for two-dimensional networks where every node is mobile. We use a random chord mobility model to represent independent node movement. Our results show that in the absence of GL/GP, there is a significant network burden for maintaining topology and routing information at the network nodes. We also derive real world scaling results using the general analytic results and these results show the poor scaling of networks without GL/GP. For networks of 100 to 1000 nodes, the rate scaling for maintaining topology in mobile wireless networks is on the order of hundreds of megabits to gigabits per second. It is infeasible to use such significant amounts of data rate for the sole purpose of maintaining topology and routing information, and thus some other method of maintaining this information will need to be utilized. Given the growing number of devices connected to the Internet, in the future it is likely that IWNs will become more prevalent in society. Despite the significant amount of research to date, there is still much work to be done to determine the attributes of a realistic and scalable system. In order to ensure the scalability of future systems and decrease the amount of throughput necessary for network maintenance, it will be necessary for such systems to use geographic location and geographic prediction information.
Description
Thesis (S.M.)--Massachusetts Institute of Technology, Dept. of Electrical Engineering and Computer Science, 2012. Cataloged from PDF version of thesis. Includes bibliographical references (p. 173-176).
Date issued
2012Department
Massachusetts Institute of Technology. Department of Electrical Engineering and Computer SciencePublisher
Massachusetts Institute of Technology
Keywords
Electrical Engineering and Computer Science.