RFID indoor positioning using RBFNN with L-GEM

As pervasive computing becomes more popular, the importance of context-aware applications increases. Physical location of user is important to context-aware pervasive application providers. RFID is one of the most widely adopted wireless positioning technologies. Compared to other wireless technologies, e.g. GPS and WLAN, RFID is particularly suitable for indoor positioning. Existing methods usually assume a constant environment for the application field. However, this may not be true in many cases. For example, warehouse may have different goods yielding different interference to RFID signal in different days. This paper proposes a new method to estimate locations of objects based on RFID. The indoor positioning with RFID reader based on the received signal strength and passive UHF tags as reference tags. A Radial Basis Function Neural Network (RBFNN) trained via a minimization of the Localized Generalization Error (L-GEM) is adopted to learn the object location based on received RFID signals. The L-GEM provides an estimate on the generalization capability of the RBFNN which is important to locate future unseen samples correctly in different yet similar environments. Simulation experiments show that the proposed method outperforms existing RFID based indoor positioning method.