Publication Date
2012
Document Type
Dissertation
Committee Members
Yan Liu (Committee Co-chair), Pratik Parikh (Committee Member), Xiuli Qu (Committee Member), Shaojun Wang (Committee Member), Xinhui Zhang (Committee Co-chair)
Degree Name
Doctor of Philosophy (PhD)
Abstract
This research was motivated by the desire of a Fortune 100 retail company to track its customers' behaviors through the implementation of a low-cost indoor location system using received signal devices. Indoor location through received signal strength (RSS) is of low costs, yet it suffers from poor accuracy due to lack of direct line-of-sight, multi-path ratio propagation, and various interferences. To improve its location estimation accuracy, therefore, we propose a mathematical programming based propagation approach for RSS based location systems. The core of our approach is an optimization model that minimizes the sum of squared errors of signal strength. In particular, the model integrates the two steps in a traditional optimization model: (a) regression analysis to find the relationship between the signal strength and distance, and (b) minimization of sum of squared error of the distance into one holistic model. By doing so, it provides a more direct and effective model for solving the RSS based indoor location problem. In addition, the optimization model is preceded by a Kalman filter model to smooth noise in the original data, and the model is expanded to link multiple time intervals into a multi-step optimization model that constrains the distance between successive steps to satisfy a feasible walking speed limit at a grocery store. To reduce computational time, an enhanced simulated annealing algorithm with dynamic neighborhood is developed to quickly find near optimal solutions. Our computational results showed that compared to traditional approaches, our approach is able to provide more robust and accurate solutions, reducing location estimation errors by 20% to 30%, and is computationally more efficient. Our approach will enable the RSS based indoor location system to be expandable to other stores without significant costs.
Page Count
112
Department or Program
Ph.D. in Engineering
Year Degree Awarded
2012
Copyright
Copyright 2012, all rights reserved. This open access ETD is published by Wright State University and OhioLINK.
