Lane-level Positioning And Map-matching Based On Multisource Information Fusion

With the rapid development of traffic,high precision positioning and map-matching technology plays a more important role in intelligent transportation system.Nowadays,lane-level positioning and map-matching system depends on LIDAR,high-def camera and DGPS,which are expensive and easily affected by environment.So,these systems can not be accepted by the consumers soon.A lane-level positioning and map-matching system is designed in this paper,which improves location precision by low-cost sensor fusion.Meanwhile,this system can complete lane-level map-matching based on commercially available road network maps.The main research contents are summarized as follows:(1)A Global Navigation Satellite System(GNSS)/Inertial Navigation System(INS)and Magnetometer(MAG)integrated navigation system is proposed and implemented in this paper.With a low-cost SPS GPS and MEMS IMU,this navigation system can reach the request of lane-level localization.The Madgwick algorithm is used to improve the accuracy of attitude estimation,by fusing accelerometers,magnetometers and gyroscopes.The error state equation of the integrated navigation system is implemented based on an extended Kalman filter.(2)An artificial intelligence module is designed to improve navigation system performance during GNSS outage periods(such as tunnel and overpass).Long and short-term memory(LSTM)networks are trained to learn the relationship between position error and dead-reckoning output when GNSS available.When GNSS outage occurs,LSTM networks predict and correct errors of the integrated navigation system to improve location precision.(3)An efficient lane-level map-matching algorithm is proposed in this paper.This algorithm can determine the lane on which the car is driving,depending on integrated navigation system and commercially available road network maps.First,a curve-to-curve road-level map-matching algorithm is implemented based on a hidden Markov model.Then,a least squares regression is used to calculate the error of each lane.The lane with minimum error is the desired one.In view of the above design,every part of the system is simulated and analyzed.This system is also tested in an urban area.In viaduct environment,positioning error of the designed integrated navigation system is 1.04 meter under confidence level of0.5.The accuracy of lane-level map-matching algorithm reach 96.49%.High-accuracy of lane-matching can promote the development of lane-level path guidance.