Particle Filter-Based SAR Approach and Trajectory Optimization for Real-Time 3D UHF-RFID Tag Localization
Matthias Gareis, Patrick Fenske, Christian Carlowitz, Martin Vossiek
Abstract
The availability of three-dimensional (3D) product maps is among the most important factors for smart warehouse environments connected to the Internet of Things. Therefore, online access to the data is a key feature that can only be provided by real-time localization systems. With the well-known synthetic aperture radar technique in combination with ultrahigh frequency (UHF) radio-frequency identification (RFID) tags, good localization results have been achieved, by labeling tags on objects and measuring the tag responses on generated trajectories, for example in warehouses, with a mobile robot platform equipped with a UHF-RFID reader. The location estimation of the tags is thus made offline after the whole synthetic aperture radar (SAR) trajectory was driven and all data are available, which can't provide online availability of the object positions. We introduce a particle filter-based real-time localization estimation technique that involves estimating tag positions during the recording of the incoming tag signals. We first evaluate the impact of trajectories according to lateral and radial resolution and show the capability of real-time signal processing on tag counts and the amount of tag reads. We then compare our novel localization technique to the well-known grid-based approach in terms of accuracy and computational load and show the localization error improvement of our technique over the filter iteration steps by generating a 3D product map of objects labeled with standard UHF-RFID tags. The 3D root mean square error (RMSE) achieved is 2.86 cm and the computation time is reduced by a factor of more than 50.