Research On Improved KNN Indoor Positioning Algorithm Based On UHFRFID In Directional Radiation Scenarios

Radio Frequency Identification(RFID)operating at 900MHz has been widely used in indoor positioning such as logistics warehousing and personnel tracking.How to design higher positioning accuracy,lower system cost and stronger environmental adaptability UHF RFID indoor positioning system has become a hotspot at home and abroad.In the UHF RFID positioning system,constructing an accurate channel propagation model and designing a suitable position estimation method are two important factors affecting the positioning performance.For channel modeling,existing research is usually based on ideal conditions such as the antenna of the reader and the tag antenna being omnidirectional and the gain value is fixed,leading the performance of the partial positioning system is greatly degraded in actual work.For the position estimation method,the k-Nearest-Neighbor(kNN)algorithm is very representative,and the positioning performance in LANDMARC and VIRE positioning systems verifies its effectiveness.Nevertheless,selecting an appropriate k value is still a key issue that constrains the performance of the kNN algorithm.Based on the above background,this paper takes the improved kNN indoor positioning algorithm based on UHF RFID in the directional radiation scenario as the research content,and the key points are as follows:1.The radiation characteristics of half-wave dipole antenna and half-wave microstrip antenna are studied inwhich the received signal strength under directional radiation scene is accurately estimated.The three-dimensional coordinate transformation technique is used to deduct the antenna gain in term of the antenna pose.Estimation model and receiver signal strength estimation model suitable for multipath propagation is accurately estimated.2.Considering the disadvantages of the existing kNN algorithm,an improved kNN indoor localization algorithm based on spectral clustering is proposed.Firstly,by rotating the reader antenna,the RSSI received in each azimuth is collected and stored,and the concept of"optimal radiation azimuth"is introduced to realize the screening of the dominant signall strength.Secondly,the spectral clustering NJW algorithm in graph theory is introduced to accomplish the dynamic selection of k values,which improves the positioning accuracy and efficiency.Finally,in order to improve the positioning time consumption,a reader antenna rotation scanning method based on mapping from coarse-grained to fine-grained is proposed.Simulating results showed that,the proposed method has a good positioning performance and environmental adaptability.3.A real measurement system for verifying the actual performance of the proposed method is designed and built.Firstly,Impinj R420 with the high-performance and Freescale MCU readers are utilized to design the hardware platform of the positioning system.Secondly,waveform control motor module is used to generate PWM pulse wave that drives the stepper motor module.The rotation of the motor drives the antenna to rotate toreceive the data.Finally,the software platform of the positioning system is designed in C#language.The measured results show that the system can complete the superior screening and accurate positioning.