Machine Learning Based High Accuracy Indoor Visible Light Location Algorithm

Aiming at the problem of indoor visible-light location accuracy, a visible-light indoor location method based on white light-emitting diode (LED) is proposed. Firstly, the data feature is constructed by using time difference of arrival (TDOA), which is arrived at the location point by the reference signal issued by different indoor LED. The physical coordinates of the location points are treated as labels. Use the data feature and label as input samples. Then the neural network model is trained. Finally, the location test is carried out based on the training model. The proposed method is simulated in the space region of 5m × 5m × 3m. The results show that the proposed neural network-based machine learning method can achieve the positioning error of about 1.662cm in indoor environment. The accuracy of indoor positioning is improved effectively.