Analysis of localization error and its effect on geographic routing in wireless networks

This dissertation focuses on several topics related to two fundamental functions of wireless networking, namely wireless localization and wireless routing. First, we investigate the error of location-estimating for several widely-applied localization algorithms. The experiments on localization by radio frequency identification (RFID) devices show that RFID is a feasible technology for wireless localization due to its satisfactory localization accuracy and low infrastructure investment; and that the support vector machine (SVM) method has the highest localization accuracy, while the nearest-neighbor method seems to be the best choice with respect to the trade-off between localization accuracy and computation complexity. The simulation on multi-hop localization shows that the centralized methods have higher accuracy than the distributed ones; and that the distance-based methods perform better in low-noise, low-landmark-ratio or short-communication-range cases than the connectivity-based ones.;Next, we conduct a comprehensive study on the capacity of wireless localization. The formulation and fundamental properties of the Cramer-Rao lower bound (CRLB) of localization error are derived. It is found that the optimal landmark placement in various regions can be obtained by several empirical rules; that the area coverage of the optimal landmark layout is smaller when the ranging error is distance-proportional than distance-independent; and that the localization capacity improves as the target distribution used shows higher level of centering in a region. In addition, a hybrid multilateration-based localization approach is proposed, which is shown to be effective in location-estimating.;Lastly, we study geographic routing in location error-prone environments. The effect of location error on routing performance is first statistically analyzed and then investigated by Monte Carlo simulation. The results show that location error increases path length, decreases delivery rate, and reduces network throughput. Furthermore, we propose a new data forwarding strategy, threshold-probability forwarding, to improve the robustness of geographic routing against location error. It is shown that the proposed strategy significantly increases the delivery rate in a network and reduces the length of a route; only slightly increases the number of transmissions along a route; and has negligible effect on the network throughput.