Improved RTAB-Map Algorithm Based on Visible Light Positioning

• Published in: IEEE access Vol. 13; pp. 136854 - 136863
• Main Authors: Hunag, Po-Hao, Liu, Si-Chuan, Yang, Cheng-Hsiung
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 2025; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Accuracy; Algorithms; Cameras; Decoding; Kalman filters; Laser radar; Light emitting diodes; Light sources; Lighting; Line of sight; Localization; Location awareness; Modulation; robot localization; Simultaneous localization and mapping; Template matching; visible light; Visible light communication
• ISSN: 2169-3536; 2169-3536
• DOI: 10.1109/ACCESS.2025.3590961

Visible Light Positioning (VLP) is an emerging technology that enables highly accurate indoor localization, especially in areas where GPS is not effective, such as inside buildings. VLP makes use of existing lighting systems but typically requires a high density of Light Emitting Diodes (LEDs) and a clear line-of-sight (LOS) between the light sources and the receivers. It is also crucial to recognize that sensors can have limitations, and their readings may not always be precise. This research presents an improved RTAB-Map algorithm based on visible light positioning. By using a custom modulator, we send signals to the LEDs to create a striped pattern. A rolling shutter camera captures these LED signals and uses template matching techniques to accurately locate the region of interest (ROI). Within this ROI, we identify the LED contours and extract the pixels from the central row into a one-dimensional array for decoding. The decoded information is then compared to a pre-existing database. Finally, we apply the solvePnP algorithm to determine the three-dimensional position and orientation, which is integrated with Kalman Filters to combine the data into the RTAB-Map algorithm for mapping purposes. This multi-sensor localization system shows strong and accurate capabilities for robot localization and navigation, even in scenarios with limited LEDs or power outages. Experimental results reveal that the proposed method achieves an average accuracy of 2.5 cm and an average positioning delay of about 5.3 microseconds.