Research On Laboratory Intelligent Inspection Robot Based On ROS

Intelligent laboratory inspection robot is an important tool for the investigation of hidden dangers in the laboratory.A series of sensors such as Lidar,inertial measurement unit and camera are installed on the mobile robot to realize intelligent inspection function.The engineering realization of laser SLAM technology and the deployment of path planning algorithm have been widely used in the development and research of inspection robots.In this paper,based on ROS robot operating system under Linux,the school laboratory was taken as the test scene,laser SLAM algorithm was deployed on the mobile robot,2D grid map was built based on the optimized algorithm and the inspection function was realized by combining the path planning method.Finally,it is equipped with ultrasonic,camera and smoke sensor modules to realize intelligent laboratory inspection function.The specific research contents of this paper are as follows:(1)Based on the ROS framework,laser SLAM technology,path planning and navigation framework,and camera and other environmental perception modules,a laboratory intelligent inspection robot platform was built,which realized the high-precision planar mapping function in unknown environments.Finally,through the real-time monitoring interface of the inspection robot,real-time feedback on the site situation,including fire monitoring and smoke and flammable gas detection results,helps security personnel quickly locate the location of the safety accident and rush to the scene quickly,and finally completes the inspection task.(2)The SLAM front-end lidar data processing process is studied,and the weight factors of rotation and translation during the robot movement are introduced to guide the dichotomy process of adaptive voxel filtering.Compared with the traditional voxel filtering method,the adaptive voxel filtering method in this paper reduces the density of the point cloud again to a certain extent on the inside of the arc or on both sides of the translation area in the case of turning,and the sparse point cloud is more realistic bounding dimensions of obstacles or walls.The obtained radar frame also meets the minimum number of point clouds required for subsequent scan matching.In terms of memory consumption,compared with the adaptive voxel filtering method,the CPU usage rate and calculation consumption are reduced.(3)Based on the Ceres nonlinear optimization library,the bicubic interpolation smoothing method is used to realize the scan matching process of single-frame radar data and grid map.The SLAM front-end scan matching based on the optimization method occupies less memory than the brute force matching method in the search window,and the scan matching speed is faster.The bicubic interpolation smooth scan matching based on Ceres is the main body,and the voxel filter constitutes the SLAM front-end laser odometer.Using the EVO tool under ROS to evaluate errors such as APE,RPE,RMSE,and STD,experiments in multiple scenarios prove that the estimated trajectory of the laser odometer scheme in this paper is more in line with the inspection robot than the traditional Cartographer and Hector laser odometer methods.true trajectory.(4)Researched the SLAM back-end pose graph optimization framework,realized the back-end constraint fusion based on the Ceres and used the front-end odometer pose data,and carried out residual construction and optimization together with the traditional constraint construction method.This paper call it the FOSLAM method.Compared with the Cartographer,Karto and Hector methods in the public data set test,it shows that the FOSLAM method can improve the scan matching performance and mapping effect without adding additional algorithm complexity.Afterwards,the performance of the algorithm was tested on the robot platform in the large scene of the experimental floor.The results showed that the feedback results of FOSLAM on the front-end scan matching accuracy and the laser odometer estimated trajectory accuracy were better than traditional laser SLAM methods such as Cartographer.