Relative Angle Correction for Distance Estimation Using K-Nearest Neighbors

Abstract

Indoor position estimation has been cited as not being accurate due to noisy environments. Variations of the k-nearest neighbor (kNN) algorithm have been used to mitigate this noise. In this paper, the concept of using the fingerprinting process with the radiation pattern of an antenna is explored. The contents of the paper focus on finding the distance between an RFID reader and a tag. Using the relative angle of the antenna with respect to a reference point, a model is introduced. Received signal strength index (RSSI) readings from the backscattered signal of the RFID tag were measured in a 7-foot $\times \,\,4$ -foot grid. The resulting data points were placed in the relative angle correction KNN (RAC-KNN) algorithm. Both regression and classification models of the RAC-KNN were implemented. As a result, the classification model yielded an 85% prediction rate of the distance estimation and the regression model resulted in a 0.2-foot average error.