Research On Mobile Robot Localization Algorithm Based On Global Information And The Feature Map

Autonomous localization technology is one of the difficult research areas in mobile robotics,and the method of using global information(such as ultrasonic localization system)to fuse its internal sensors for localization is the most common,but it is prone to cumulative errors when global information is lost(in blind areas).SLAM(Simultaneous Localization And Mapping)based localization method does not require global information,but relies on the surrounding environment features and requires complex laser processing to eliminate accumulated errors.Therefore,the combination of the two methods in real-time requires a balance between stability and real-time.In this paper,a mobile robot positioning algorithm based on global information and feature map(GF-SLAM,Global information and feature-based SLAM)is proposed.When the global information is stable,EKF(Extended Kalman Filter)positioning is performed,and a feature map is established at the same time,it is converted to F-SLAM(Feature-based SLAM)for particle filtering and continuous output pose when the global information is lost.The details of this paper are as follows:(1)According to the principles,models,and advantages and disadvantages of the odometer,IMU(Inertial Measurement Unit),lidar,and ultrasonic positioning system,the output data of each information source is determined,and the conversion relationship between the coordinate systems in the algorithm is analyzed.(2)The EKF positioning model based on global information is established,and the global information state is judged according to the EKF optimized pose and the dead reckoning pose difference.Simulation experiments verify that EKF fusion multi-sensor positioning can greatly improve positioning accuracy.(3)A new feature map construction method is proposed.After feature extraction,a Kalman filter is established separately for each feature in the environment to track,which greatly reduces the state matrix dimension and memory usage,and eliminates low-quality features to ensure the map quality.(4)The adaptive improvement on the traditional particle filter obtains the particle filter positioning algorithm under the feature map.The EKF pose at the previous moment initializes the particle swarm,calculates the matching particles to reduce the particle swarm matching time,and determines the validity after obtaining the pose estimation to ensure positioning stability.(5)In the experiment,two different simulation scenarios were built and compared with several classic SLAM algorithms.The results show that F-SLAM performs at an intermediate level in several algorithms,The two positioning methods in GF-SLAM can achieve seamless switching,and continue to mapping and positioning in blind areas,the introduction of global information can quickly eliminate the cumulative error in F-SLAM,and improve the overall positioning accuracy.The mobile robot positioning algorithm based on global information and feature map proposed in this paper realizes the effective use of various sensor information.The algorithm is simple and effective,requires low computing power,and has strong real-time performance.It can be widely applied to low-cost mobile robot positioning,and also great practical value and research significance.