Preliminary Analysis of RFID Localization System for Moving Precast Concrete Units using Multiple-Tags and Weighted Euclid Distance k-NN algorithm

Abstract

This paper presents two RFID localization methods based on a k-NN algorithm for multiple moving tracking tags attached to a concrete masonry unit (cinder block). This work uses passive RFID tags for localization and seeks to provide rapid wireless analysis for future smart infrastructure projects where precast concrete modular structures are moved during transport and assembly. The RFID localization system uses four reader antennas, four tracking tags, and 28 reference tags in a realistic indoor assembly environment. Results show average error in the direction of movement as low as 10.5 cm. Increasing the number of nearest neighbors in the k-NN algorithm is shown to reduce error in all coordinate directions. Increasing k from 4 to 6 is shown to reduce error by 4 cm or 10%. The localization environment is analyzed, and reference tags 22, 9, 5, and 8 around the moving cinder block are seen most commonly as nearest neighbors. A modified k -NN algorithm, described here as a weighted Euclidian distance k -NN algorithm is presented that reduces total error from 41.1 cm to 32.5 cm.