Research Of Indoor Positioning Technology Based On ZigBee Network

GPS-based location services have brought great convenience to people's lives.With the improvement of quality of life,indoor positioning technology has emerged from time to time.However,GPS signals are weak and difficult to penetrate buildings and space obstacles,which is no longer suitable for indoor positioning.Nowadays,thanks to the rapid development of short-distance communication technology,indoor positioning technology has ushered in opportunities.ZigBee technology is a low-cost,low-power,high-reliability short-range wireless communication technology.Compared with other short-range wireless technologies,ZigBee has the advantages of self-organizing network,large network capacity,and good performance in the field of the Internet of Things solutions.In the traditional indoor positioning method,the indoor environment is complex and variable.It is difficult to accurately measure relevant positioning parameters,resulting in low indoor positioning accuracy.In recent years,domestic and foreign scholars have used fingerprint positioning technology to improve indoor positioning accuracy.Therefore,this paper explores a high-precision,stable indoor positioning system based on ZigBee technology.Indoor positioning methods based on ranging technology have always been the focus of industry research.This method is simple to calculate and easy to implement,but the results are not unsatisfactory.The indoor environmental obstacles are dense,the multipath interference suffered by the signal during the propagation process is serious,and the RSSI(Received signal strength indication)fluctuation is irregular,which makes it impossible to accurately convert the RSSI into distance information during the positioning process.Therefore,a large number of experiments have been carried out in this paper to explore the effects of antenna angle,walls,obstacles and human body on RSSI fluctuations.Based on this,an improved maximum filtering algorithm was proposed,and experiments show that it is better than before improvement.Moreover,Positioning experiments were conducted in this paper under three different environmental conditions.The experimental results show that the method has poor stability and is not suitable for indoor positioning in complex environments.Specifically,the average positioning accuracy is low at 1.21 m in a complex environment of 4.4m*7m.Aiming at the problem that the indoor positioning method based on ranging technology has low positioning accuracy in complex environment,an indoor positioning method based on BP neural network was proposed in this paper.In the offline phase,the method establishes the mapping relationship between the RSSI data of the sample points and its position coordinates,and then performs network training.In the online phase,the RSSI data of the target node is collected,and then the position coordinates are estimated using the trained network.The experimental results show that the proposed method improves the positioning accuracy of most target nodes,but the stability is poor.Specially,in the 4.4m*7m complex environment,the maximum positioning error is 2.2289 m,the minimum positioning error is 0.2895 m,and the average positioning accuracy is 1.1407 m.Aiming at the defects of the indoor positioning method based on BP neural network,an improved method was proposed by using K-Means clustering algorithm in this paper.In detail to be specific,the training set and the test set are preprocessed using the K-Means clustering algorithm before network training,and the cluster center is dynamically determined according to the distance between the test set data.The experimental results show that the improved indoor positioning model has a maximum positioning error of 1.0272 m,a minimum positioning error of 0.084 m and an average positioning error of 0.5037 m in the 4.4m*7m complex environment.Comparing with the improvement,the positioning accuracy is improved by 0.637 m,and the target node with positioning error less than 1m accounts for 95%,and the target node with positioning error less than 0.5m accounts for 60%.