Research On Real-time Monocular Camera/IMU/GNSS Fusion Intelligent Vehicle Positioning Algorithm Considering CRVE

Real-time and accurate position information is the basis for intelligent vehicle to realize autonomous driving.Compared with a single sensor,the multi-sensor fusion positioning can improve the positioning accuracy,but it will lead to the increase of data volume and thus reduce the real-time performance.Therefore,it is of great significance to ensure the real-time performance of the multi-sensor fusion positioning of vehicles.When multi-sensor fusion localization algorithm based on SLAM(Simultaneous Location and Mapping)framework is optimized by using fixed-size sliding window,Characteristics of Vehicle Driving Environment(CRVE)increased suddenly,resulting in an increase in positioning time.In this paper,Monocular Camera/IMU(Inertial Measurement Unit)/GNSS(Global Navigation Satellite System)integrated positioning algorithm considering CRVE is proposed,which can improve the real-time positioning of vehicles.The main work of this paper is as follows:(1)Dynamic sliding window optimization algorithm considering CRVE.Firstly,the SLAM algorithm framework and sliding window optimization principle used in this paper are introduced,and CRVE is defined.Then,the CRVE of several common road scenarios is analyzed,and the influence of CRVE on back-end optimization error and time consumption is studied.In order to improve the real-time performance of positioning,a dynamic sliding window optimization algorithm considering CRVE was proposed to solve the problem that the CRVE suddenly increased when using fixed-size sliding window optimization,which would lead to the time-consuming increase of positioning.Then,in order to dynamically change the Size of the sliding Window(SW)during the optimization process,a marginalization method to change SW was derived.Finally,the open data set KITTI was used to carry out comparative experiments.The results show that,compared with the fixed sliding window optimization algorithm,the positioning time consumption of the proposed algorithm in the Sequence＿03 and other 4 data sets is reduced by an average of 8.67%,and the positioning accuracy is similar.(2)In order to realize the intelligent vehicle positioning in the GNSS out-of-lock environment such as tunnel,interchange or urban shelter.Based on the dynamic sliding window optimization algorithm,from four aspects of the visual front end,IMU pre-integration,initialization and optimization of the back end,A Dynamic Window Visual Inertial Odometry System(DW-VINS)based on CRVE was designed.(3)When GNSS signal is trusted,the IMU measurement equation is taken as the state equation,and the GNSS measurement equation is taken as the observation equation.GNSS/IMU data are fused based on the extended Kalman filter.The fusion result is used as the initial pose of the new key frame.On the basis of DW-VINS,Dynamic Window Visual/IMU/GNSS Integrated Locating System(DW-VLS)with CRVE in mind was designed.(4)The intelligent vehicle platform was built,and three typical roads of straight line,broken line and curve were selected for experiment,and DW-VLS and VINS Fusion were compared.The experimental results show that DW-VLS can reduce the localization time in the environment with CRVE suddenly increasing,so as to achieve the purpose of improving the real-time positioning and ensuring the positioning accuracy.