QRLA: A Quick and Robust Localization Approach Based on RFID for IoT Indoor Applications

The device-free passive localization has been received much attention in the past few years. An important issue with this technique is that the packet containing the received signal strength readings may be lost and negatively affects the localization accuracy. To address the challenge of packet loss, we propose a fingerprint-based packet-loss resistant method that combines a missing RSS value processing algorithm with a support vector machine model. To evaluate the proposed method, comprehensive experiments are conducted in various settings in which the maximum packet loss rates of links reach 22.63% and samples reach 28.80%. The proposed method outperforms previously proposed linear discriminant analysis and Naive Bayes algorithms, achieving localization accuracies at least 94.86% in an empty room and 93.32% in a cluttered room. Further simulations show that the proposed method works well even for extreme conditions when approximately 50% of original samples are discarded. The experiment results also reveal that the proposed method is insensitive to different human activities including standing and walking, thus reduces training cost.