Research On The Tightly Coupled Single-Frequency Multi-GNSS/INS/Vision Integration For Precise Position And Orientation Estimation

With the rapid development of self-driving cars,autonomous robots,and Unmanned Aerial Vehicles(UAVs),the demand for precise position and attitude has been increasing in urban environments.The integration of Global Navigation Satellite System(GNSS)and Inertial Navigation System(INS)can provide continuous,reliable,and complete navigation solution,which has become one of the most widely used integrated navigation techniques.However,the navigation errors of the GNSS/INS integration will be increased dramatically during the GNSS outages for the low-cost Micro-Electro-Mechanical-System(MEMS)Inertial Measurement Units(IMUs),which definitely restricts the system availability.In order to navigate in GNSSdenied environments,the monocular visual-inertial system has been widely adopted in robotics community due to the complementary characteristics and low-cost hardware.However,the visual-inertial system can not provide the absolute navigation information in a global reference frame,and it suffers from navigation error accumulation.Therefore,the integration of GNSS,INS and vision could be developed to take the most of their advantages to obtain better navigation performance.Meanwhile,the rapid deployment of the multi-constellation GNSS(multi-GNSS)will greatly improve the positioning accuracy and availability in GNSS challenged environments.However,the most relevant existing research focus on the integration of the vision,INS,and the position measurement from the Global Positioning System(GPS)Real-time Kinematics(RTK).It is well-known that this kind of integration has the drawback that the information from the visual-inertial system can not be used to aid RTK.Furthermore,the positioning capability of GPS RTK is very poor in GNSS challenged environments,which seriously restrict the performance and application of the integrated navigation system.In order to overcome the limitations of the existing research and meet the demand of highaccuracy position and attitude determination in GNSS challenged environments,this thesis conducts systematic and in-depth research on the algorithm of the tightly coupled integration of the single-frequency multi-GNSS(GPS,BDS,and GLONASS),MEMS-IMU,and monocular camera.The innovation-based outlier-resistant algorithm and INS-derived relative constraint for ambiguity resolution algorithm are proposed specifically for the tightly coupled RTK/INS integration,and the complete filtering model of the tightly coupled single-frequency multi-GNSS RTK/INS/Vision integration is constructed.Finally,the performance in terms of the ambiguity resolution,positioning,velocity and attitude determination is evaluated and analyzed comprehensively by using multiple sets of field vehicular data.The main research content and contributions of this thesis are summarized as follows:1.In order to enhance the high-accuracy positioning performance of single-frequency RTK in complex kinematic environments,the tightly coupled single-frequency RTK/INS integration algorithm is proposed.To begin with,the mathematic model of multi-GNSS RTK positioning was established,and then two mathematic models of RTK/INS integration were presented and discussed,together with the ambiguity resolution with inertial aiding.Secondly,the innovationbased outlier-resistant ambiguity resolution(AR)and Kalman filtering strategy is proposed specifically for the RTK/INS integration to resist the measurement outliers or poor-quality observations.Finally,the performance of the tightly coupled RTK/INS integration is evaluated and analyzed comprehensively by using filed vehicular test data in both open-sky and sub-urban environments.The results indicate the empirical success rate of single-epoch ambiguity resolution for the tightly-coupled single-frequency multi-GNSS RTK/INS integration is over 99% even at elevation cut-off angles of 35° or 40° for short baselines in open-sky conditions,and centimeter-level positioning accuracy can also be achieved in both horizontal and vertical directions.For short baselines in sub-dense urban environments,results reveal that the outlierresistant strategy is effective to improve the ambiguity fixing rate,and the tightly-coupled single-frequency multi-GNSS RTK/MEMS-IMU integration can greatly improve the positioning accuracy and availability of GNSS RTK,and even outperforms the dual-frequency multi-GNSS RTK in terms of AR and positioning performance.It demonstrates the superior advantages that the tightly coupled integration of the single-frequency multi-GNSS RTK and low-cost MEMS-IMU has.2.The traditional method that uses INS-predicted absolute position to aid ambiguity resolution has the limitation that the correct ambiguity resolution will be impossible in case of biased system state.In this thesis,a new INS aided ambiguity resolution approach using INSderived relative constraint is proposed to overcome this limitation.To begin with,we derive the formula for calculating the INS-derived relative position increment based on the IMU preintegration theory.Then,the formula for the INS aided ambiguity resolution using INS-derived incremental position measurement is derived.Meanwhile,cycle slip detection with INS-derived phase increment is developed to make the new approach feasible.Finally,the effectiveness and performance of the proposed algorithm is validated and evaluated by a field vehicular test.The results indicate that it is effective for the cycle slip detection method to detect all the real and simulated small cycle slips like one cycle(data interval is 1 s),and the high-accuracy INSderived incremental position measurement can still be obtained in the case of biased system state,and the new method can achieve comparable ambiguity resolution performance in comparison with the the traditional method for the single-frequency GPS+BDS RTK/INS integration.3.In order to address the problem of the rapid navigation error accumulation during GNSS outages for the low-cost MEMS-IMU based GNSS/INS integration,the thesis presents two kinds of vision aided INS methods,namely the tightly coupled Vision/INS integration with known features and with multi-state constraint.We prove that the multi-state constraint Kalman filter is optimal from the perspective of least-squares estimation theory.Then the method to unify the tightly coupled RTK/INS integration model and Vision/INS integration model is presented,and finally,the two tightly coupled integration models are validated and analyzed with simulated data.The results show that the tightly coupled vision/INS integration algorithm can achieve drift-free pose estimation,and the heading accuracy is improved obviously in comparison with the GNSS/INS integration.Besides,the multi-state constraint Kalman filter algorithm greatly reduces the position and heading drift error of the INS,which effectively makes up for the drawback that the GNSS/INS integration has weak heading observability under certain conditions.4.We design and establish the hardware platform that is made up of the GNSS,IMU,and monocular camera,and the space and time synchronization between these sensors is implemented.Then,a field vehicular experiment was conducted in GNSS challenged environments to evaluate and analyze the performance of the tightly coupled RTK/INS/Vision integration in terms of position,velocity and attitude.The results reveal that both multi-GNSS and vision contribute to the centimeter-level positioning availability significantly in GNSS challenged environments.Meanwhile,the velocity and attitude accuracy can be greatly improved by using the tightly-coupled multi-GNSS RTK/INS/Vision integration,especially for the yaw angle.Therefore,the tightly coupled single-frequency multi-GNSS RTK/INS/Vision integration can be considered as an effective solution to provide precise position,velocity and attitude information in GNSS challenged environments.In summary,this thesis conducts systematic and in-depth research on the tightly coupled single-frequency multi-GNSS RTK/INS/Vision integration algorithm,build complete filtering model for the tightly coupled RTK/INS/Vision integration,and evaluate the proposed model comprehensively by multiple field vehicular tests.The proposed methods in this thesis will boost the research,application,and development of the technology for precise determination of position and attitude in GNSS challenged environments.