Research On RFID Based Spatial Information Perception

Spatial information reflects the objects or entities spatial distribution information,mainly includes the spatial location,spatial range and spatial relationship of objects.Under the background of Internet-of-Things(Io T),how to achieve accurate spatial information perception is the key to improve our intelligence and automation level.Radio Frequency IDentification(RFID)technology is a remote identification technology which based on the electromagnetic coupling and backscatter communicating method.It is treated as an indispensable technology for IoT because of its advantages such as low cost,small volume and battery-free communication.Currently,billions of passive RFID tags are employed for labeling retail items,tracking medicines,supervising assets and security control areas.Lots of researchers are focusing on the research of RFID technology.However,passive RFID system can only detect whether the tagged objects are in or not in the surveillance region of electromagnetic radiation(about 10 meters)without the ability to give accuracy spatial location or spatial relationship for the objects.As a result,RFID system now can't provide spatial information for upper-level applications such as indoor navigation and context awareness.To solve such problems,this dissertation mainly focuses on the RFID three-dimensional spatial perception technology.Firstly,a RFID spatial perception model is proposed based on the depth analysis of the relationship between radio channel characteristic(phase,signal strength,angle of arrival etc.)and the spatial location of the tagged objects.Then a RFID spatial perception experimental platform is established,which is composed of a commercial off-the-shelf RFID reader and a mobile robot.On this basis,we additionally explore the objects' location and spatial relationship during the communication between the tag and the reader in order to provide underlying support for the spatial information based applications.The major research content and innovation are:(1)To solve the reader-tag spatial range awareness problem,we provide a multifrequency carrier waves phase difference based spatial ranging method.This method utilizes multi-frequency carrier waves to inventory the target tag.Phase differences under each frequencies pair are treated as the basic information to calculate the tag-reader spatial distance.Kalman filter is employed to calibrate the phase difference values.Phase difference fusion methods are also presented to improve the ranging precision.Experiments show that the ranging error of our method is around 16.5cm within 5 meters communication region,2cm is achieved when shrinks the reader-tag communication distance into 0.5m.(2)To solve the tag's 3D spatial location awareness problem,we provide phase interference based 3D spatial location awareness method.This method is inspired by Interference Synthetic Aperture Radar(InSAR)theory and utilizes several apertures to create multiple holographic images.On this basis,we use phase interference between different heights of apertures and combine DBSCAN clustering method to give the tags' 3D location.Comparing to existing methods,this approach can implement the localization without the development of any reference tags or auxiliary infrastructures.Only one movable reader is enough for the method.Additionally,this method reduces the location error to about 18 cm comparing to existing methods.(3)To solve the environment self-adaption spatial perception problem,a multipath radio tomography based device-free self-adaption spatial perception method is provided in order to perceive the spatial location of the untagged objects or malicious invaders(called as obstacles).This method is inspired by the radio tomography scheme and introduces the MUSIC multipath angle-of-arrival spatial spectrum estimation theory.We use the spatial spectrum estimation results to create the weight factor for each multipath channel and employ the attenuation of signal strength to establish the tomographic,which maps to the obstacle's location.The results show a locating error about 12 cm and improve the estimating accuracy about 60% comparing to classic tomographic imaging method.