Research On Key Technologies Of Inertial/visual Integrated Navigation

Since the 21 st century,many types of automated guided vehicles(AGV)and unmanned aerial vehicles(UAV)have begun to enter people's lives.With the development of technology and the increasing demand for application scenarios,the accuracy and autonomy of their navigation capabilities have raised higher requirements,and it is impossible to use a single sensor to complete different navigation tasks.Therefore,the combined navigation method of inertial information and visual information was studied.Rely on their high autonomy,this integrated navigation can work normally under the harsh external conditions and various types of interference,it has better navigation accuracy and ability because it didn't need external equipment to provide information.The article first introduces the basic knowledge of inertial navigation and visual navigation.Leading to the theory of inertial pre-integration to solve problem that the frequency difference between the two sensors.Deducing the pre-integration model and error transfer equation,so that the two types information can be efficiently integrated.Next,the feature matching method based on visual features is studied.The results show that the optical flow method has higher efficiency.Then a simple visual navigation algorithm is designed.Experimental analysis shows that the stereo vision algorithm designed in this paper can be used for navigation,but the accuracy is low,and the real geographic position cannot be obtained.For the problems of the above navigation methods,this paper uses a nonlinear optimization method to fuse inertial information with visual information to improve navigation performance.Derive its error equation and build model of the fusion problem,and study the key technology which used in the integrated navigation algorithm.By analyzing the error of the system through semi-physical simulation,the direction of optimization is brought up.At the end of the article,based on the conclusion of Chapter 4,the deviation of IMU is also estimated.At the same time,in order to reduce the impact of inaccurate external parameters,external parameters are also placed in the optimization model.Due to distance from different feature points in the three-dimensional space to camera is different,the different weights are given to different feature points to get more accurate state estimation results.The experimental results show that above methods effectively improves the accuracy of the integrated navigation system.In addition,Chapter 5 analyzes the characteristics and existing problems of the commonly used nonlinear optimization algorithms,improves the algorithm,and proposes a new optimization algorithm based on the dogleg algorithm to solve the navigation tasks.The experimental results show that the algorithm in this paper further improves the accuracy and solving efficiency.Through the experiments of sports cars,the algorithm of this paper has practical significance.