An experimental approach to RFID system performance prediction model

Radio frequency identification (RFID) is increasingly applied for advanced manufacturing systems as part of ubiquitous systems, implementation of which is the task of the ubiquitous system engineer. The present method, used by almost all the ubiquitous system engineers, is based on trial-and-error methods. To reduce costs, prediction of RFID performance in ubiquitous systems is a promising approach. Our previous research has proposed a Computer-Aided Ubiquitous System Engineering (CAUSE) framework by which a ubiquitous system can be designed, analysed and verified on the computer. By developing an architecture and prototype, we showed its usefulness in the ubiquitous system engineering task. In order to be practically effective, however, the accuracy of the CAUSE needs to be improved. For such a purpose, however, the conventional model-based approach is hard to apply because only a few factors have been taken into consideration. In this paper, we investigate a heuristic approach for dealing with realistic problems based on experiments, by proposing a procedure for the experiments and neural network methods. Experimental results show the effectiveness of the experimental method compared with the conventional model-based approach.