Arthur Charléty, Morgane Magnier, Mathieu Le Breton, Laurent Baillet, Eric Larose, Ludovic Moreau

Radio-Frequency Identification (RFID) shows great potential for earth-sciences applications, notably for landslide surface monitoring at a high spatio-temporal resolution with long-term robustness to meteorological events (rain, fog, snow). The ability to localize RFID tags using Unmanned Aerial Vehicles (UAV) in a Synthetic Aperture Radar (SAR) approach, would offer new possibilites for monitoring inaccessible terrains, even under vegetation and snow.
To that end, an onboard measurement system was built that allows Global Positionning (GPS) tracking of an RFID reader antenna, in order to perform real-time SAR measurement acquisition of RFID tags on the ground. Three antenna tracking methods were compared. In addition, Markov-Chain Monte-Carlo (MCMC) optimization was used to estimate tag position and characterize the solution, even in non-convex cost function scenarios. Two cost functions were compared, based on different RFID-phase processing approaches.
Real-time SAR-RFID localization yielded a centimeter accuracy in the horizontal plane, with lower resolution in the vertical direction. The Post-Processed Kinematics algorithm proved to best fit antenna tracking. The unwrapped-phase based cost function provided more convex solutions, at the cost of a lower accuracy compared to the complex-phase cost function. MCMC proved to be computationally efficient in SAR-RFID optimization, with enhanced results concerning the shape and orientation of the main localization errors.