Graph Optimization Based Localization Algorithm by Fusing VLP and VO

Visible light positioning (VLP) has attracted considerable interest due to its high accuracy and low cost. However, VLP typically requires several light emitting diodes (LEDs) for positioning, which may lead to positioning failure in sparse LED layout or parts of the links are blocked. To address this challenge, we propose a novel graph optimization-based localization algorithm by fusing VLP and visual odometry (VO), termed GO-VLO. In the considered model, a few LEDs are deployed on the ceiling and broadcast positioning information, while a receiver equipped with a camera conducts GO-VLO to achieve precise positioning with limited LEDs. GO-VLO consists of a front end to process positioning information and a back end to further optimize the positioning results. In particular, VLP and VO are first used separately at the front end, thus obtaining two respective positioning results. Then, a visible-light visual odometry fusion mechanism is proposed to fuse the positioning results of VLP and VO preliminarily. To strengthen the constraints between frames and prevent location failure when VLP is unavailable, a novel graph optimization framework is further proposed in the back end, which notably takes the covisible feature points of two successive frames into consideration. Experimental results show that our proposed method attains a positioning accuracy of 4.93 cm in two dimension and 6.20 cm in three dimension under dense LED layout.