Research On Integrated Positioning Strategy Based On Multi-sensor Information For Intelligent Vehicles

As an important part of environment perception for intelligent driving,intelligent vehicle positioning is one of the enabling technologies for decision,trajectory planning and motion control.It is also one of the key technologies for intelligent transportation system.And it has become a hot research topic on intelligent vehicles.Global Navigation Satellite System(GNSS)is currently the most widely used positioning method,but its signal often drifts and is even blocked completely in places such as high-rise buildings,tunnels,overpasses,dense avenues,etc.Therefore,the integrity and continuity of positioning can't be guaranteed.Besides,its update frequency and positioning accuracy are both quite low.Many other positioning methods have been proposed,for example,differential technology,Dead Reckoning(DR),Inertial Navigation System,Visual Odometry,Ultra Wide Band(UWB),Map Matching,etc.They are complementary to GNSS to some extent,but the critical problems,such as error accumulation,hardware cost,map accuracy,matching efficiency,etc,still remain to be solved.So any independent positioning method has its inevitable drawbacks,and it's difficult to meet practical application requirements.Therefore,this paper carries out the research of integrated positioning strategy based on multi-sensor information for intelligent vehicles,and an integrated positioning algorithm is established with GNSS,DR,UWB and visual map matching.Besides,both off-line simulation tests and real vehicle tests are implemented to verify the proposed integrated positioning strategy.The research contents are as follows:(1)Research on independent positioning algorithmThe working principle and error sources of three independent positioning methods including GNSS,DR and UWB are analyzed.And their independent positioning algorithms are studied respectively.Firstly,the independent positioning algorithm of GNSS is established based on the current statistical model.Secondly,the independent positioning algorithm of DR is established based on optimal estimated heading angle.Thirdly,a two-stage positioning algorithm of UWB is established.Finally,all three independent positioning algorithms are verified respectively based on MATLAB/Simulink.(2)Research on fusion positioning algorithm of GNSS/UWB/DRA fusion positioning algorithm based on GNSS,UWB and DR is established.The spatial coordinate transformation equation and the Lagrangian three-point interpolation method are adopted to realize multi-sensor data synchronization.The Federated Kalman Filter is adopted to establish a multi-sensor fusion positioning algorithm with adaptive information distribution coefficient.The sensor error models of magnetic compass,gyroscope and odometer are analyzed,and Kalman Filter equation of multi-sensor fusion positioning algorithm is established by indirect method.Finally,the established multi-sensor fusion positioning algorithm is verified based on MATLAB/Simulink-Car Sim co-simulation platform.(3)Research on integrated positioning strategy based on visual map matchingAn integrated positioning strategy based on visual map matching is proposed.Firstly,the time sequential pose map of driving environment is designed which includes map reference frame block and map reference pose block.Secondly,based on the polar geometry constraint relationship,the Random Sample Consensus algorithm is adopted to solve the camera extrinsic parameter matrix through feature extraction and matching.Thus the pose of intelligent vehicle can be estimated.Besides,the results are optimized by using graph optimization method.Finally,based on KITTI dataset,the proposed integrated positioning strategy is verified on MATLAB/Simulink-Visual Studio co-simulation platform.(4)Real vehicle test verification and analysisA real vehicle test platform is built according to testing requirements.The accuracy evaluation and calibration work of relevant experimental equipments is implemented.The time sequential pose map of real vehicle test scene is constructed and the actual location coordinates of UWB base stations are measured.Finally,the real vehicle test and analysis are carried out based on the real vehicle test platform.The results of real vehicle test show that the proposed integrated positioning strategy for intelligent vehicles can effectively improve the positioning accuracy.