Research On Multi-sensor Information Fusion Based Localization And Navigation System For Intelligent Vehicle

With the development of artificial intelligence and computer technology,smart vehicle automatic driving technology has also received widespread attention in the academic research and engineering circles.The core basic technology of smart vehicle autonomous operation is autonomous positioning and navigation functions.However,in the operating environment with complex interference,and the sensors used have their own advantages and disadvantages,multi-sensor data fusion positioning and navigation technology has become one of the important research directions in the field of smart vehicle.For complex indoor and outdoor scenes,the lack of scene features,insufficient lighting,obstruction and other factors will affect the positioning of smart vehicle,so this paper mainly studies the multi-sensor fusion method applicable to different indoor and outdoor scenes to achieve robust positioning of smart vehicle navigation mission.In order to facilitate and quickly verify the multi-sensor fusion algorithm,a simulation platform for smart vehicle and experimental environments is built based on Gazebo software,which can realize sensor data collection and smart vehicle control decision-making functions.First,according to the design requirements,describe the kinematics model of the smart vehicle,and establish the camera coordinate system of the sensor,the lidar coordinate system and the vehicle body coordinate system,and then analyze the imaging model of the depth camera and lidar,the data collection principles of GPS and IMU,and the time synchronization relationship of each sensor,and complete the error calibration and data collection experiment of each sensor.In order to solve the problem that GPS signal occlusion affects the positioning accuracy of smart vehicle in outdoor scenes,this paper uses the measurement results of GPS and IMU sensors to design an adaptive extended Kalman filter algorithm,and combines GPS differential positioning technology to improve the positioning accuracy and robustness of the system.Then design a predictive tracking model for trajectory tracking to complete the outdoor navigation task,and finally combine the lidar sensor for environmental perception,and complete the program design and outdoor experimental testing of the perception layer,decision layer and control layer of the smart vehicle,and analyze the experimental results to produce positioning errors and verify the multi-sensor fusion positioning method.Aiming at the problem of autonomous positioning of smart vehicle in the indoor environment without GPS signals,a tightly coupled positioning method for visual inertial navigation based on nonlinear optimization is designed.First,build a SLAM system of pure visual feature point algorithm,including ORB feature point extraction and matching,PNP pose estimation and back-end graph optimization algorithm.In order to improve the positioning accuracy and robustness,a tightly coupled least squares error objective function of visual inertial navigation is constructed and a method for solving pose is given.A public data set was used to compare the experimental result data of the positioning trajectory and posture curve of pure visual and visual inertial navigation fusion,and the tight coupling fusion method of visual inertial navigation was verified on the physical platform.At the same time,considering the problem of weak light environment affecting visual feature extraction and matching,the design of a fusion lidar positioning mapping module program can switch to laser-assisted positioning and construct a occupancy grid map for real-time navigation tasks in the case of weak light conditions.