A Real-Time Indoor Positioning System Based on Wi-Fi RTT and Multi-source Information

Abstract

In order to solve the problem of locating pedestrians in indoor environments, an indoor real-time high-precision positioning system based on smart phones was constructed. Aiming at the non-line-of-sight and multipath problems in the wireless signal-based indoor positioning technology, a method using deep convolutional neural network (CNN) to learn the nonlinear mapping relationship between indoor spatial position and Wi-Fi-FTM ranging information is proposed. At the same time, a fingerprint gray-scale construction method combined with a specific AP location is designed for pedestrian location prediction. Considering the large fluctuations and poor continuity of fingerprint-based positioning results, a particle filter positioning algorithm with adaptive update of state parameters is proposed, which improves the degree of freedom and positioning accuracy of pedestrian positioning. Finally, a large number of tests were conducted in an indoor test environment of about 800 m². Compared with the traditional fingerprint positioning method, the fusion positioning algorithm based on the CNN network improves the positioning accuracy by about 30%. Compared with the millimeter-level precision optical dynamic calibration system, the particle filter fusion method has 94.2% results less than 1 m, and the average positioning error is 0.41 m. Moreover, in the real indoor environment of a commercial supermarket, it is further verified that the positioning system has high precision and high continuous positioning performance in practical applications, and has great application and promotion value.