18 / 2024-07-30 16:25:01

Research on UWB Non Line of Sight Error Recognition Based on Deep Learning“AITC 2024+空天之星”

Indoor positioning;,Ultra Wide Band (UWB),Non Line of Sight (NLOS),Deep learning

Based on the advantages of high transmission rate, strong resistance to multipath interference, and high positioning accuracy, Ultra Wide Band (UWB) is widely used in the field of indoor positioning. However, due to the severe impact of Non Line of Sight (NLOS) errors on UWB, its positioning accuracy is limited. At present, deep learning algorithms have been widely applied in NLOS error recognition. In response to the problem that deep learning requires multiple feature dimensions and identifies a single source of NLOS error in the process of NLOS error recognition, this paper takes the difference between the total received power and the first path received power, as well as the distance value, as the feature values, and evaluates the recognition performance of multiple deep learning methods for UWB NLOS errors and NLOS error sources using less dimensional information. Firstly, this article evaluates the UWB ranging performance under different obstructions (doors, walls, glass, human body). Secondly, the received power of UWB in Line of Sight (LOS) and NLOS environments was analyzed. Finally, the UWB NLOS error and its source were identified using the difference between the total received power and the first path received power, as well as the distance value. The experimental results show that both the Random Forest (RF) method and the Back Propagation (BP) neural network algorithm have an accuracy rate of over 98% in identifying NLOS errors under four types of occlusion, and the RF method can identify the source of NLOS errors with an accuracy rate of over 97%. The above conclusion can provide support for UWB NLOS error recognition and NLOS error modeling.