Non-line-of-sight Obstacle Recognition And Error Compensation For UWB Indoor Positioning

In recent years,indoor positioning technology has been widely used.Ultra-wideband(UWB)stands out in the field of indoor positioning with its unique advantages.However,in a complex environment,UWB signal propagation is easily blocked by Non-Line-of-Sight(NLOS)obstacles,resulting in NLOS errors.In this paper,PCA-FCM fuzzy clustering obstacle recognition method,LS-MSE algorithm is used to optimize anchor layout,and FCM-KNN algorithm is used to compensate the positioning error,which solves the problem of low accuracy of NLOS channel identification and complex indoor environment positioning.Main tasks as follows:(1)Aiming at the problems of traditional NLOS obstacle recognition method,low recognition accuracy and application scenarios that need to be specifically considered,a fuzzy clustering recognition method based on PCA-FCM is proposed,which achieves 92%recognition accuracy.Firstly,according to the difference between the line of sight(LOS)signal and the non-line-of-sight signal,six channel characteristic parameters such as signal strength and mean excess delay are selected;Secondly,in view of the principal component analysis(PCA)The advantage of the model in feature extraction.The PCA model is used to reduce the dimensionality of the feature parameters.Finally,FCM algorithm is used to optimize the objective function after dimensionality reduction to obtain the clustering center Degree of membership,thereby improving the accuracy of NLOS signal recognition.(2)Aiming at the problems that traditional anchors placement algorithms do not consider obstacles and insufficient coverage of anchors'signals,anchors layout algorithm based on Least Square(LS)and Mean Squared Error(MSE)is proposed.Firstly,different groups of anchors use LS for intra-group connection.Secondly,select a group of anchors with a smaller MSE value.Finally,arrange the anchor at a point where the MSE value is small.The positioning accuracy of the algorithm under NLOS environment is less than 0.5m~2,and the time to install the base station is also greatly reduced.(3)For the traditional K-Nearest Neighbor(KNN)positioning algorithm,the measurement value fluctuates greatly,resulting in inaccurate positioning accuracy.A positioning algorithm combining FCM and KNN is proposed and verified by experiment.Firstly,divide the indoor environment into grid areas,collect the RSSI of each point,and establish an offline database;Secondly,to prevent the label from detecting the signal of each anchor,build a dynamic fingerprint database;The dynamic database corresponding to RSSI,and preliminary m position points;Finally,the FCM algorithm is used to cluster the m position points,and the average value is used to obtain the final positioning coordinates.Experiments show that,compared with the traditional KNN algorithm,the positioning algorithm proposed in this paper has higher accuracy.There are 27 figures,10 tables,and 80 references in this paper.