Object localization using passive RFID tags

Abstract

Passive radio frequency identification (RFID) systems are revolutionizing the indoor positioning and tracking applications. There has been substantial research on practical applications of this technology and hospitals especially trauma care units are one such area where this capability can lead to improved workflow. Our system uses the Alien RFID reader and the “Squiggle” passive RFID tags to create an effective solution for tracking various medical items. Based on the Received Signal Strength Indication (RSSI) value of the tags, we developed a localization algorithm which uses a neural network estimator to estimate the distances of the tags. To reduce the effect of noise in the RSSI values received from the reader, we accumulate data over a period of time, remove the outliers and the average the remaining RSSI values. The RSSI based estimation algorithm provides very accurate estimation when the spatial density of tags is low (about 25 tags per square meter). To improve the localization accuracy at higher spatial densities we augmented the RSSI method of estimating distances by using the number of times the tags were read or the “read-count”. We also investigated how different types of occluding materials affect the localization accuracy. Metal and Humans also can cause complete occlusion when the positioned in direct line of sight between the antenna and the tag. To overcome human based occlusion, we placed an additional ceiling mounted antenna per 10 m2. This intervention makes possible the detection (but not localization) of tags when the vertical field of view is not occluded. We also studied the effect of the material to which the tags are attached and determined the effects on localization accuracy.The software system developed using Java is designed in a modular fashion and provides interfaces to tools like Matlab so that it is easy to experiment to various other localization algorithms. We also developed an intuitive User Interface to display the locations of tags and the associated items. Once a tag is identified its associated description can be looked up in a computer database and this also can be displayed in the user interface.