Characterizing and Improving the Non-Collaborative and Collaborative Localization Problems

This thesis focuses on the least-squares formulation of the non-collaborative and collabo- rative position location problems. For the non-collaborative problem, characterization encompassing the number of minima and the causes thereof is provided. Based on these efforts, we propose an improvement to the existing modified parallel projection method (MPPM), the reflected parallel projection method (RPPM). We show that the global minimum to the non-collaborative objective function can nearly always be found using the non-optimal reflected parallel projection method (RPPM).

For the collaborative position location problem, we provide a characterization that shows a heavy tail of root-mean-square (RMS) error due to a small percentage of simulated node/anchor layouts when solved by the iterative parallel projection method (IPPM). We provide an identification technique that successfully identifies most layouts that show large RMS error followed by a proposed solution to improve the accuracy in those problematic layouts.

Finally, we report the findings of a measurement campaign that validates our Gaussian model for line-of-sight (LOS) noise and shows that, for these particular measurements, non-line-of-sight (NLOS) noise is difficult to accurately model and can be very large.

This research was supported by a Bradley Fellowship from Virginia Tech's Bradley Department of Electrical and Computer Engineering, made possible by an endowment from the Harry Lynde Bradley Foundation.