Research On Noise Adaptive SLAM Algorithm For Mobile Robots

With the continuous development of robot technology,various robots gradually play a role in people's daily work and life.Among them,mobile robots bring great convenience to people with their flexibility and mobility.In order to perform the assigned task autonomously,the mobile robot should have the function of autonomous navigation.Positioning and environment map construction are two basic problems that need to be solved to realize autonomous navigation.At the same time,Simultaneous Localization and Mapping(SLAM)technology is an effective method to deal with these two basic problems.SLAM has a lot of achievements after decades of development,but as demand continues to increase,SLAM still has many problems to be solved.This paper focus on the SLAM problem of mobile robots.When the observation has a wild value point or the observed noise parameter is unknown or changed,the performance of the traditional SLAM algorithm will decrease.This paper proposes an adaptive SLAM algorithm based on the variational Bayes method.The new algorithm combined variational Bayes method with cubatrue Kalman filter for state estimation,which effectively improves the accuracy of SLAM.The main contents and results of this paper are as follows:1.The mobile robot navigation system used in this paper was modeled.The SLAM problem of mobile robots is analyzed,described and modeled through probability theory,which transforms SLAM into a state estimation problem.At the same time,the traditional SLAM algorithm based on cubature Kalman filter is introduced.2.Aiming at the problem that the traditional SLAM algorithm is degraded when the observed noise parameters are unknown or changing,a variational Bayesian cubature Kalman filter SLAM algorithm based on the inverse Wishart distribution is proposed.The inverse Wishart distribution is used to model the observed noise parameters,and then the nonlinear variational Bayes filter is utilized to estimate the joint posteriori probability of the mobile robot state and the unknown observation noise parameter.Through simulation experiments,the positioning accuracy of the algorithm is greatly improved when the observed noise parameters are unknown or changed.3.When there is a wild point in the observation,the observed noise will not satisfy the Gaussian distribution because of the presence of the heavy tail characteristic.At this time,the performance of the traditional filter-based SLAM algorithm is degraded.In this paper,a variational Bayes cubature Kalman filter SLAM algorithm based on Student-t distribution is proposed.The algorithm considers the existence of the heavy tail characteristics of the observation,and uses the Student-t distribution to model the observed noise,because the Student-t distribution is a generalized Gaussian distribution,and its tail is heavy than the Gaussian distribution.Then the cubature integral method is used to approximate the mean and variance of the nonlinear transformation,and the variational Bayes method is used to estimate the posterior probability of the state of the mobile robot and the observed noise parameters.The effectiveness of the algorithm is verified by simulation experiments.The last chapter of this paper summarizes the whole thesis,and looks forward to the research prospects of positioning at the same time.