Toward large-scale PSD-based visible light positioning with visual–inertial fusion

Abstract Accurate large-scale indoor positioning is essential for location-based services. Visible light communication (VLC) holds promise for supporting global indoor positioning due to its inherent advantages, including immunity to electromagnetic interference and low development cost. As a type of optical receiver, the position-sensitive detector (PSD)-based visible light positioning (VLP) system provides accurate, high-update-rate absolute positioning results. However, its positioning range is constrained to 1 m 2 due to light-emitting diode (LED) distribution and the receiver’s field of view, which is a common limitation of VLC-based positioning systems. To address this issue, a loosely coupled system is proposed in this paper through pose-graph optimization, which effectively integrated a PSD-based VLP system with visual–inertial odometry (VIO) capable of providing large-scale relative positioning. Unlike traditional VLP approaches that require dense LED deployment, our method enables large-scale indoor positioning through the sparse placement of multiple PSD-based positioning areas. Experiments were conducted in a real indoor environment to validate the effectiveness of the fusion system in large-scale areas and its capability to correct VIO-induced drift. Our proposed method provides precise position estimation in a local area of 1 m 2 , with a mean positioning error of 9.16 mm. In two large-scale indoor environments: a single room with a corridor and multiple rooms connected by a corridor, with a 1 m 2 area of sufficient LED signal coverage and other regions experiencing LED shortage/outage area, the system provides both locally precise positioning and global drift correction.