Research on downhole visible light modeling and localization based on point cloud technology

Abstract

In this paper, we reconstruct the downhole visible light channel model based on point cloud technology. To make up for the defect that the reflective surface element is a virtual surface, the planar projection method and the idea of point-by-point insertion are combined to turn the point cloud data on the wall surface into a triangular mesh topology. In the meantime, the point cloud data of obstacles is utilized to determine the precise range of shadows cast by light. Subsequently, shadow coefficients are defined to derive new expressions for channel gain. In addition, the localization in a 6 m × 4 m × 3 m space is achieved through the utilization of the particle swarm algorithm optimized BP neural network (PSO-BP) localization algorithm. In scenarios where the walls exhibit unevenness and barriers are present, the root mean square positioning error is found to be 6.52 cm when taking primary reflection into consideration. Conversely, when solely relying on direct power, the positioning error increases to 9.92 cm. Consequently, it is important to consider the non-line-of-light (NLOS) link when conducting downhole visible light localization.