Search

Browse

• Communities
• Authors
• Year
• Titles
• Subjects

My Space

• Login

Indoor RF localization with algebraic methods

Peppas Spyridon

Full record

---

Summary

This work investigates the problem of localizing batteryless, passive, ultra-low cost RFIDtags, under the presence of non-white noise that multipath introduces inside buildings. Two recently proposed grid-based methods are utilized to tackle the aforementioned problem with radio signals. The first method is borrowed from the through-the-wall radar imaging community and this thesis adapts it to the case of monostatic interrogation of Gen2 tags, with commercial RFID readers. The method exploits the strength of the received signal (RSSI) and its phase, expressed in a linear system that incorporates the impact of known reflectors.The second one is a phase-based method, which puts forth a ”differential mitigation” scheme based on maximum likelihood estimation (MLE).For the first method, it is found that certain ambiguities need to be resolved, relevantto the wavelength, which can be mitigated based on the fundamental theory of compressive sensing. Specifically, it is observed that the so-called sensing matrix, can lead to a successful reconstruction by properly selecting the carrier frequency or spreading the measurements, with respect to the coherence of its columns or its blocks. As for the second method, a selection of a phase measurement is needed; it is found that if more than one phase measurements are randomly selected, then a more accurate estimate of the target is achieved. According to simulations, the random selection of a subset of the phase measurements as a reference can provide up to 50% improvement, concerning root mean squared error.This thesis includes both simulation and experimental results using a commercial Gen2RFID reader at UHF, installed on a robotic platform. Experimental results with the available equipment show that, given static poles, mean absolute localization error in the order of 10 cm for indoor 2D localization is achieved. On the other hand, when the RFID reader is placed on a mobile robotic platform, it is shown that mean absolute localization error in the order of 19cm is possible. Finally, a proof-of-concept is offered, by the modeling of a single reflector, which leads to about 19% 3D localization error improvement, compared to the case where the reflector is ignored. Further examination of this finding perhaps opens a fruitful new research direction.

Available Files

Services