High-Precision Positioning And Orientation Determination And Application Of Simultaneous Mapping For Vehicle Scenarios

High-precision positioning,orientation determination and high-definition maps are the key components to realize autonomous vehicle navigation under urban synthesis scenarios.The GNSS/SINS integrated navigation system can provide continuous and stable positioning services in most open-sky scenes,but its performance is severely degraded in complex scenes due to the IMU level.In recent years,with the gradual construction and improvement of global navigation satellite systems,the available observation signal resources are rapidly increasing.How to make full use of multi-frequency signals and current ground-based reference network resources under new opportunities,and tap the unique motion information in-vehicle applications to suppress the accumulation of navigation errors,are of great importance to achieve fast,continuous and robust high-precision positioning and fixing.In addition,the conventional high-definition mapping method relies on high-precision trajectory input,and the real-time mapping accuracy is inferior and there are many manual interventions,which is difficult to meet the needs of rapidly changing urban scenes.In this paper,we study the method of high-precision positioning and simultaneous mapping for complex urban scenes to realize fast and high-precision positioning and mapping of vehicle intelligent terminals in urban scenes.The main research contents of this paper are as follows.In this paper,we focus on the problems in the current stage of vehicle positioning and mapping,and research the method of high-precision positioning and simultaneous mapping for complex urban scenes,to realize the fast and high-precision positioning and mapping of vehicle intelligent terminals in urban scenes.The main research contents of this paper are as follows.(1)A unified model of GNSS/SINS tight coupling based on original observations is described from the state domain and observation domain.A fused navigation model assisted by multi-source on-board motion information is constructed,and it is pointed out that the difficulty in the application of motion constraints lies in the accurate calibration of installation errors.Therefore,the GNSS/SINS tight-coupling-based autonomous calibration method for the spatial relationship of on-board sensors is proposed,and the effectiveness of the algorithm is verified by simulation analysis and on-board experiments.(2)Based on multi-frequency multi-constellation GNSS observation resources,a new multi-frequency PPP/INS tightly coupled positioning method for continuous decimeter-level demand is proposed,which achieves single-epoch ambiguity resolution by the long wavelength of ultra-wide lane/wide lane combined observations,and further tightly fuses IMU observations to obtain continuous decimeter-level positioning.The effectiveness of the algorithm is verified using two sets of onboard experiments in typical urban scenes,and a certain degree of continuous precision positioning can be maintained even in extreme environments such as tunnels and woods,meeting the demand for fast and precise positioning in-vehicle applications.(3)For the demand for fast centimeter-level positioning,a new method of PPP-RTK/SINS tightly coupled fast precision positioning with vehicle-borne information augmentation is proposed,which significantly improves the ambiguity convergence efficiency based on the atmospheric correction constraint extracted from the existing ground-based reference network resources,further improves the ambiguity resolution rate and position availability with INS assistance,and uses the typical wheel odometer and non-holonomic constraints of in-vehicle terminals and other augmentation information to improve GNSS degradation scenario system positioning performance.Finally,the effectiveness of the algorithm is verified by four typical scenarios: open-sky road,GNSS-complex road,under elevated and tunnel.(4)Targeting the pain point of high cost,low efficiency and slow update of urban high-precision mapping,a large scale 3D mapping method MSF-PM which is compatible with high-precision positioning filter estimation and pose graph optimization is proposed,which completes the simultaneous construction of the back-end map while realizing the front-end high-precision positioning,and uses multiple threads parallel strategy to improve the algorithm efficiency.The data collection and algorithm validation of the comprehensive urban scenes are carried out based on the self-developed multi-sensor integration platform.The results demonstrate that the MSF-PM method can achieve stable and continuous high-precision pose estimation even under the extreme conditions of severe satellite occlusion and complete interruption.In addition,the maps generated based on the optimized high-precision poses and laser point clouds not only have high-precision information on the global reference frame,but also have better relative accuracy and overall consistency.The MSF-PM method significantly improves the accuracy,consistency and production efficiency of large-scale urban 3D maps,which can meet the application requirements of rapid update of high-precision maps and real-time crowdsourced mapping.