Research On Simultaneous Localization And Mapping Of Self-driving Vehicle In Unknown Environment

The automobile industry has emerged the trend of development,which is intelligent,connected.As its specific object,self-driving vehicles will inevitably become the protagonist of the future automotive industry.At present,the domestic and foreign car companies,the Internet giant,research institutes have invested enormous efforts to research self-driving vehicles.Autonomous navigation is one of the most basic and important functions of self-driving vehicles.In order to realize the autonomous navigation in the unknown environment,it is necessary to realize the accurate positioning and map building of the environment simultaneously.This interrelated process is known as simultaneous localization and mapping(SLAM).With the deepening of SLAM research,seeing that its important theoretical and practical value,many scholars agree that simultaneous localization and map building is considered to be one of the key technologies for the realization of self-driving vehicles.In this paper,we research on the simultaneous localization and mapping of self-driving vehicles in unknown environment.The main contents are as follows:Firstly,this paper expounds the technical connotation of the SLAM,discusses the development status of SLAM and the key problems need to be solved.Secondly,the model to solve the SLAM problem of self-driving vehicle is discussed and established,including self-driving vehicle body model,environment model,sensor model and data association model.On the basis of the above model,a general mathematical model for the SLAM problem of the self-driving vehicle is presented,the rationality of the model is proved by the simulation experiment.Thirdly,the mathematical model of SLAM algorithm based on extended kalman filter(EKF-SLAM)and SLAM algorithm based on unscented kalman filter(UKF-SLAM)is established.The advantages and disadvantages of EKF-SLAM algorithm and UKF-SLAM algorithm are analyzed theoretically.The simulation experiments were carried out for two algorithms by MATLAB.Twenty independent repeated experiments were carried out to verify the feasibility of the two algorithms and the superiority of the UKF-SLAM algorithm.Then,the SLAM algorithm based on particle filter is discussed.According to the principle of particle filter,established the mathematical model of FastSLAM algorithm.On the basis of FastSLAM algorithm,adaptive resampling technique and UKF are introduced,A new algorithm ARUFastSLAM is proposed,which is introduced adaptive resampling and unscented kalman filter into basis FastSLAM algorithm,and given its the mathematical model.The two algorithms are verified by simulation experiments,and the experiments of 10,20,and 100 particles are designed respectively.Each group has 20 independent repeated experiments.And the results show that the ARUFastSLAM algorithm is effective.Furthermore,the standard test set Victoria Park Dataset is used to verify the estimation accuracy of ARUFastSLAM.Finally,the GUI:interface is designed by graphical interface design tool of MATLAB,which is used for the self-driving vehicle SLAM algorithm simulation.It makes complex simulation program presented in the form of simple human-computer interface interaction,and improves the efficiency of the algorithm.