Research On UKF-SLAM Of Multi Sensor Fusion Based On Observation Likelihood

Since the 21st century,with the development of science and technology,the commercialization potential of mobile robots has been further explored,and more and more enterprises have focused on this market.Mobile robots begin to show practical value in service industry,logistics industry,manufacturing industry and other industries,which requires mobile robots to have certain intelligent and autonomous behavior ability.Autonomous navigation is the basis of autonomous behavior ability of mobile robot,Simultaneous localization and mapping(SLAM)is a key technology for mobile robot to realize autonomous navigation in complex and changeable environment without prior map.As a solution to realize robot self localization and environment map construction,SLAM has been a research hotspot in robot field since it was proposed.Slam system consists of front end and back end.The front end of the system is responsible for sensor data processing and data association;the back end of the system involves information fusion,positioning and mapping of multiple sensors.In this paper,aiming at the data reliability problem of the front-end displacement sensor,data fusion between odometer and inertial measurement unit(IMU)is carried out;Aiming at the problem that the increase of the number of map features affects the efficiency of observation feature data association,this paper proposes a grid map based statistics method of observation feature potential association objects,which can effectively reduce the calculation of traversal matching process.The slam system back-end based on Unscented Kalman filter(UKF)solves the problem that Kalman filter(KF)is not suitable for nonlinear system,and has higher nonlinear approximation accuracy and less computation than extended Kalman filter(EKF).However,as a slam system based on Kalman filter algorithm,it also has the problems of time-consuming increase and system robustness decrease with the increase of state vector dimension.This paper proposes an improved ukf-slam back-end algorithm based on idea of Rao Blackwellised,which is an improved ukf-slam method based on observation likelihood.It splits the positioning and mapping process of ukf-slam,uses the matching degree to screen the observation features with better quality to participate in the state updating,and uses the observation likelihood to optimize the posterior state in the iterative process of state updating.Simulation results show the effectiveness of the improved method.Compared with ukfslam,the improved ukf-slam method based on observation likelihood reduces the state update time by 73.2%,and improves the positioning accuracy by 25.3%.Moreover,the algorithm time is no longer affected by the number of map features,and is proportional to the number of observation features at t time.