CSI fingerprint positioning method based on PD array in VLP systems with signal blockage

Abstract

In visible light fingerprint positioning, the line of sight (LOS) signal between the photodetector (PD) and the LED may be blocked by randomly moving people or objects, resulting in degradation of positioning accuracy. To solve this problem, this paper studies a fingerprint positioning method based on PD arrays and channel state information (CSI). The proposed method leverages the spatial arrangement of the PD array to constrain multiple CSI fingerprint matching operations, rather than relying on a single PD for fingerprint matching. Two algorithms are proposed: the PD array minimum matching error (PAMME) algorithm and the PD array LOS path selection (PALS) algorithm. The PAMME algorithm leverages the spatial relationship between multiple PDs to perform multi-point matching, calculating cumulative matching errors to mitigate the limitations of single PDs in fingerprint matching. Building on PAMME, the PALS algorithm estimates the LOS signal, selecting signal combinations with the smallest matching error and removing interference from reflection paths, further improving positioning accuracy. To reduce computational complexity in multi-PD fingerprint matching, the particle swarm optimization (PSO) algorithm is integrated into the method. A segmented search strategy with nonlinear variation factors and Gaussian perturbation is introduced to avoid local optima. In a 4 m × 4 m × 3 m indoor multi-path simulation environment, where two LOS signals are randomly blocked, the PAMME and PALS methods achieve average positioning errors of 0.5 cm and 0.21 cm, respectively. This represents error reductions of 64% and 85% compared to single PD-based CSI fingerprint positioning. Additionally, the proposed PSO strategy optimization reduces the time complexity of PAMME by 94% and PALS by 50%, with minimal increases in positioning error. The simulation results demonstrate that the proposed multi-PD fingerprint positioning method achieves excellent positioning performance with a moderate increase in computational complexity. This highlights the method’s potential and advantages, offering new insights and approaches for indoor fingerprint positioning research.