A Novel Robust Visible Light Positioning Based on Sparsity-Adaptive Compressed Sensing

The fingerprint-based visible light positioning (VLP) method offers high-precision indoor positioning with centimeter-level accuracy, making it well-suited for applications like robot navigation. However, the sensitivity of photodiodes (PDs) to natural light compromises the system's robustness. To address this challenge, we construct a robust received signal strength (RSS) database that leverages the properties of pulse width modulation (PWM) technology to minimize interference from natural light. This robust RSS database serves as the foundation for a novel, low-complexity, sparsity-adaptive compressed sensing (CS) algorithm, which enables accurate signal reconstruction without requiring prior knowledge of sparsity. Experimental results show that the proposed VLP system effectively mitigates the impact of natural light and maintains consistent positioning performance across varying natural lighting conditions. With a grid spacing of 20 cm, the proposed RSS database construction method achieves an average positioning error (APE) of 1.36 cm in VLP, marking a 75.54% reduction compared to the traditional RSS database construction method. Furthermore, the proposed VLP system maintains an APE of 2.15 cm even at a 40 cm grid spacing. This demonstrates that our system achieves excellent positioning accuracy while maintaining low computational and storage complexity.