Indoor positioning with multi-beam CSI of commercial 5G signals

Abstract

The fifth-generation (5G) network has been deployed for a vast number of users. The advanced capabilities of 5G technology have opened up opportunities for accuracy positioning and navigation. However, when it comes to indoor positioning using commercial 5G signals, there are persistent challenges. One particular challenge arises from the fact that in numerous indoor scenarios, there is only one base station (called gNB) heard from the receiver. This limitation makes the traditional geometric methods difficult to be applied indoors for 5G positioning. To solve the problem of indoor positioning with single 5G gNB, we propose a fingerprinting method based on the multi-beam of 5G downlink signals. This method utilizes the multi-beam Channel State Information (CSI) and employs an Extreme Learning Machine (ELM) for dimensionality reduction, aiming to improve both the accuracy and efficiency of indoor positioning. To assess the effectiveness of this method, field tests were conducted in indoor scenarios. The results demonstrate that, by taking the advantages of multi-beam property in 5G signals, it is able to achieve CSI positioning in single-gNB scenario, and the positioning accuracy over 94% while improving the positioning efficiency.