Research On A Vehicle Positioning System Based On Radio Frequency Identification, Vision And Ultra Wide-band Communication

One of the major challenges for high-level self-driving is to comprehensively perceive the road environment.Existing intelligent self-driving cars are equipped with a large number of high-performance sensors,leading to a unacceptable cost for the majority.Rapid development of connected vehicle offers another way,in which vehicles can directly obtain position of others via wireless short range communications.In this setting,accurate and reliable vehicle positioning plays a significant role in propelling self-driving,reducing traffic accident,and improving transportation efficiency.Directed to the shortages of existing positioning approaches,a new positioning system based on Radio Frequency Identification(RFID),vision,and Ultra Wide-band(UWB)was proposed in this paper,aiming to realize lane-level-accuracy positioning in dense urban environment.In this system,vehicle positioning was divided into two steps:lateral positioning,and complete positioning.Lateral positioning was performed using RFID,vision,and a lane judging algorithm.After lateral positioning,longitudinal position of the vehicle could be found by GPS(with low accuracy),or vehicle-infrastructure(V-I)UWB communication and a single-anchor positioning algorithm(with high accuracy).Relative position between vehicles could also be found via vehicle-vehicle(V-V)communication and a V-V positioning algorithm.To study the proposed system's feasibility,the positioning accuracy of single-anchor V-I positioning algorithm and V-V positioning algorithm were analyzed grounding on their error sources and error models under typical working conditions.It was demonstrated that V-I/V-V distance was the key factor to positioning error,and these two algorithms could achieve decimeter-level accuracy when Roadside Units(RSU)were arranged well and a safety distance between vehicle was kept.Taking Shanghai as an example,the proposed system's cost for the whole city was compared with that of high-performance GPS.It turned out that the positioning system was economical and suitable for large-scale implementation.The lane judging program and control software for this system were developed based on OpenCV and C++.A prototype was built with a real car,RFID reader,UWB module,and camera,and positioning functions were realized on a real road.Using this prototype,a integrated collision warning system,which could detect potential dangers from three different lanes within 200m and alarm when necessary,was design and developed.Besides,the framework and methodology of an automated bus rapid transit system based on this positioning system was proposed.In conclusion,the positioning system proposed in this paper is of lane-level lateral accuracy and decimeter-level longitudinal accuracy while much cheaper than the existing high-performance GPS approach,showing its feasibility for large-scale deployment.The development of the prototype and exemplifying applications prove the proposed positioning system's bright prospect in the field of active safety and self-driving car.