A modified technique in RFID networking planning and optimization

Radio Frequency Identification (RFID) system is acknowledged as a mature technology often deployed in large scale tracking applications. Implementation issues like cost and effectiveness necessitated research on RFID Network Planning (RNP). The solution typically inspired by nature includes the use of Genetic Algorithm (GA), Bacteria Foraging Optimization (BFO) and Particle Swarm Optimization (PSO) Algorithm. In this research, PSO algorithm was used in the optimization process as it was considered as a very useful, efficient and well known algorithm. However, there are no parameters settings of PSO that fits all. This issue becomes more significant if PSO is used for solving complex optimization problem such as the RFID Network Planning (RNP). Any variation made to the values of PSO parameters (number of iterations, number of swarms, inertia weight value and correction factor value) will result in a huge difference to the output of the optimization process. In addition, RFID tag coverage optimization comes with another set of parameters to be considered such as the number of RFID readers, number of RFID tags and working space area. RFID tag coverage optimization is also considered as a high dimensional optimization process. To reduce the complexity of the optimization process, this research focuses on developing a method to determine the optimum setting for PSO parameters. Two sessions of Design of Experiment (DOE) analysis were embedded in the optimization process. Initially, the objective function was developed by elaborating the mathematical model of RFID tag coverage optimization. In order to get the general settings of PSO parameters, several RNP scenarios were generated by the first session of DOE and a Matlab code was developed for each scenario. For the second session of DOE, the results from the PSO optimization of each RNP scenario were analyzed using Minitab 16 software and the optimum settings of PSO parameters were identified. From here, the general settings of PSO parameters that can be applied to all scenarios are proposed. For the purpose of validation, the RFID tag coverage optimization using PSO and DOE combinations was tested against two variants of PSO. The comparison tests were done for all RNP scenarios and from the experiment results, the combination of PSO and DOE manages to perform better compared to other PSO variants in the test of objective function value eventhough not the fastest. As a conclusion, the proposed method (PSO and DOE combination) can be considered as a robust and efficient optimization system because it manages to generate high quality results in overall RNP scenarios. Additionally, the spread of the generated results is small.