Indoor Robot's Navigation Methods With SLAM Techniques

The navigation based on SLAM(Simultaneous Localization and Mapping)is a research hotspot in the robot field,and SLAM is an effective and accurate method for robots to locate themselves and build environmental maps.To improve the accuracy of indoor navigation and its visualization of human-computer interaction,aiming at the low accuracy and real-time performance of traditional SLAM in indoor navigation,a real-time indoor LiDAR(Light Detection and Ranging)SLAM with BINN(Biological Inspired Neural Network)is proposed in this paper.The proposed approach has been implemented and verified using ROS(Robot Operating System)in simulation and real environments.The study on indoor navigation has a significant effect on promoting the development of robot's navigation.The main work and results are as follows:(1)Due to the influence of noise and external factors in the traditional RBPF(RaoBlackwellised)-SLAM,the motion state of the robot is difficult to be described accurately.To solve this problem,firstly,the environmental model,the robot's motion model,the radar's observation model,and the odometry model have been built by using the grid map representation,the arc motion law,the triangle ranging principle,and the likelihood field model,respectively.Then,the sensor data have been transformed into information required by the mobile robot to accurately simulate its motion and perception of the environment.Finally,the results prove that the maximum error in the straight-line and the rotation tests are 2.8%,2.22%,respectively.(2)Due to the errors caused by the odometer's reading and external factors in the traditional RBPF-SLAM,the robot positioning and mapping effect is not good.Firstly,the traditional RBPF-SLAM's principle and the suggested function's expressions have been studied.Then,the pose's difference of the robot at two adjacent moments based on laser data has been used to replace the input of the odometer in the traditional suggested function to improve the traditional RBPF-SLAM.Through ROS and self-built robot,the navigation accuracy of the improved RBPF-SLAM in the simulation and real environments have been tested and analyzed,moreover,the robot's position is located and the environmental map has been built.Finally,the results reveal that the maximum errors in positioning and mapping of the improved RBPF-SLAM are 9.40% and 10.41%,respectively,while those of the traditional RBPF-SLAM are 23.40% and 18.18%.(3)Aiming to the changes in the dynamic environment during the navigation,which will reduce the accuracy and real-time performance.Firstly,BINN has been introduced into improved RBPF-SLAM to propose a real-time LiDAR SLAM with BINN.Then,the performances of navigation in simulation and real environments have been verified and analyzed through ROS and Rviz.Moreover,the dynamic real-time path planning has been realized,and the navigation's real-time performance and accuracy have been improved.Finally,the results indicate that,compared with A* and Dijkstra,the path length,the number of path's turns and the indoor navigation time of the proposed approach are reduced by 7.09%,14.29%,6.97%,respectively.