关键词: 超宽带, 神经网络, 自调节卡尔曼滤波, 定位算法

Abstract: A UWB positioning algorithm was proposed based on neural networks and self-adjusting Kalman filters for improving the positioning accuracy of current a certain three-line automatic driving rail transport system vehicles. The UWB tags and base stations were used to collect large amount of distance information between tags and various base stations and collect the actual locations of the corresponding tags, and the neural network was trained. The distance information between the tags and various base stations was sent to the centralized control center server through the network during the real-time positioning stage, and the real-time locations of the UWB positioning tags were obtained by the optimized neural network. The self-adjusting Kalman filter was used to improve the accuracy of the real-time tag positions furtherly. A set of UWB tag moving trajectories containing inclines, straight paths, and curves were designed for simulation based on the actual vehicle operation, and a UWB positioning system was built, the moving trajectories of the tags were designed, the UWB positioning algorithm combining the neural network and self-adjusting Kalman filter was verified through experiments. The results show that the maximum positioning error of the positioning algorithm combining neural network and self-adjusting Kalman filter is as 223.58 mm, and the average positioning error is as 43.16 mm, and the root mean square value of the positioning errors is as 42.06 mm. The positioning algorithm proposed combining the neural network and self-adjusting Kalman filter has the advantages of higher accuracy, better real-time performance and stability compared with the three-point positioning algorithm, Kalman filtering algorithm, and neural network algorithm, and the current positioning requirements of the three-line rail transports may be fulfilled.

Key words: ultra-wide band（UWB）, neural network, self-adjusting Kalman filter, positioning algorithm