RFID reader deployment using RBFNN with L-GEM

RFID technology has been deployed in a wide range of industries. The deployment of RFID readers is essential to the success of RFID applications. RFID is a wireless communication technology and therefore the interaction between the RFID tag and reader is easily influenced by many factors. Current RFID reader deployment based on designer experience or predefined mathematical model without field test. When the RFID system is fielded, the performance may be unsatisfactory. Therefore, we propose a neural network based method to select the best location of RFID reader deployment among possible choices in the field. The major factor affecting the performance of neural network is the architecture selection which we will select via a minimization of its localized generalization error to enhance the neural network´s generalization capability. Radial basis function neural network is adapted owing to its fast learning speed. The neural network will suggest the best location iteratively to reduce the number of cost and labor intensive random trials. Experimental results on an artificial warehouse show that the proposed method outperforms random trials.