A UHF-RFID gate control system based on a convolutional neural network

In this paper, a robust, easy-to-deploy UHF-RFID system to classify transpallet actions at gates or checkpoints has been presented. The system is based on deploying a set of reference RFID tags on the floor of the checkpoint arranged in a matrix form. In addition, a single RFID reader antenna, which is over the checkpoint, is used to query both the reference tags of the system and those that are used identify goods and transpallets. When a transpallet crosses the controlled gate or checkpoint, it directly blocks the reference tags under it. Hence, reference RFID tags under the transpallet are progressively shadowed. As a consequence, if the number of readings of each reference tag is observed versus time, the movement direction of the transpallet can be inferred. This information was used to build images which are then fed to a Convolutional Neural Network (CNN) that classifies transpallet movements in incoming, outgoing or passing through the controlled checkpoint. A total of 159 measurements were acquired for different transpallet trajectories using 24 reference tags and a CNN was trained showing promising results.