Real-Time RF Mapping with Analog Backscatter Tags

Sammanfattning

Signal backscattering offers energy-efficient and scalable RF communication. Recent advancements have introduced models which utilize backscatter tags as anchors in RF frequency mapping. While these novel models achieve high localization accuracy and map RF signal propagation effectively, they require an extensive offline calibration of up to 4.5 seconds. Alternatively, novel analog backscatter tags can reflect a signal with an output frequency proportional to the incident signal strength. By decoding this frequency at the receiver, the system enables a passive channel estimation of the path attenuation between the tag and the receiver without environmental calibration. This project utilizes the custom-built analog backscatter tags to achieve a localization accuracy on par with existing models like RFIMap, while circumventing the lengthy calibration processes. The methodology of the project culminates in implementing and evaluating two localization models, the Linearized Least-Squares (LLS) model and Nonlinear Optimization, applied to different real-world and simulated datasets. The contribution of the Friis free space and the Log-Distance path loss models to signal localization is also calibrated and measured. The Log-Distance path loss model’s path loss exponent (n) was optimized to an ideal n=1.8 for the indoor Line-of-Sight (LoS) experiments conducted. Real-time simulations demonstrate the Nonlinear Optimization model’s capability to accurately localize moving RF sources. The proposed Nonlinear Optimization model produced a mean localization error of 0.136 m and a 95% localization error of 0.336 m, which performs on par with RFIMap when scaled to a comparable testbed size. Further work and implementation of machine learning models could help improve accuracy and possibly address the current lack of an RF source signal path propagation map due to insufficient environmental information.