Genetic algorithm optimized grid-based RF fingerprint positioning in heterogeneous small cell networks

In this paper we propose a novel optimization algorithm for grid-based RF fingerprinting to improve user equipment (UE) positioning accuracy. For this purpose we have used Multi-objective Genetic Algorithm (MOGA) which enables autonomous calibration of grid-cell layout (GCL) for better UE positioning as compared to that of the conventional fingerprinting approach. Performance evaluations were carried out using two different training data-sets consisting of Minimization of Drive Testing measurements obtained from a dynamic system simulation in a heterogeneous LTE small cell environment. The robustness of the proposed method has been tested analyzing positioning results from two different areas of interest. Optimization of GCL is performed in two ways: (1) array-wise calibration of the grid-cell units using non-overlapping GCL and (2) creating an overlapping GCL to cover of whole simulation area with different rectangular grid-cell units. Simulation results show that if sufficient amount of training data is available then the proposed method can improve positioning accuracy of 56.74% over the conventional grid-based RF fingerprinting.