Experimental investigation of environmental interference mitigation and blocked LEDs using a memory-artificial neural network in 3D indoor visible light positioning systems

Environmental interference and blocked light-emitting diodes (LEDs) often happen in the received signal strength (RSS)-based indoor visible light positioning (VLP) systems, while few solutions to these problems exist. In this paper, we proposed a novel deviation-correction algorithm named memory-artificial neural network (M-ANN) in the 3-dimensional (3D) indoor RSS-VLP system. By memorizing and utilizing the features of signal strength conversion between adjacent test moments, M-ANN can adapt to different test environments in the positioning process. Also, with the help of a designed genetic algorithm (GA) module, M-ANN can efficiently search and retrieve the missing data from an offline simulation database to prevent the VLP outage caused by the blocked LED. The experimental results in a test region of 0.6×0.6×0.8 m3 demonstrate that the proposed M-ANN can significantly mitigate the impact of environmental interference, and it can still maintain relatively high-precision positioning even in the case of blocked LEDs. The average positioning error of 1.04 cm, 2.89 cm, and 3.53 cm is experimentally achieved in the situation of environmental interference, one blocked LED and two blocked LEDs, respectively.