3D Wall Environment Mapping And Localizaiton Of Biped Wall-climbing Robots

Climbing robot has a wide range of application prospects in high difficulty,high intensity and high risk tasks.At present,the main limitation of climbing robots is that they failed to perform autonomous navigation well.In this paper,the mapping technology and localization technology in autonomous navigation of climbing robot have been studied.The mapping technology consists of the mapping and the parametric expression map of 3D wall environments.The localization technology tracks the pose of climbing robots.The main contents are as follows:(1)The mapping algorithm of climbing robots based on multi-sensor information fusion is studied.In order to eliminate the registration error caused by the degradation of Li DAR measurement,this paper proposes a mapping algorithm integrating multi-sensor information.The result of IMU integration is taken as the initial value of the point cloud registration algorithm,so as to obtain the advantage of combing two kinds of sensing information.Experimental results showed that the algorithm could effectively overcome the degradation of Li DAR measurement,which means that the algorithm has the good robustness.(2)The parametric expression map of 3D wall environment that would be applied to path planning of climbing robots is studied.In the parametric expression map,the contour points of the wall is used to represent the wall,as the contour points can accurately describe the pose and shape of the wall.This paper presents an algorithm to convert point cloud into the parametric expression map.The algorithm consist of point cloud preprocessing,plane segmentation,fitting mathematical model of planes,extraction and smoothing of the wall contour points.Besides,three indexes are proposed to evaluate the accuracy of the parametric expression map.Experimental results show that the parametric expression map can accurately describe the pose and shape information of the walls in the real environment,and the proposed algorithm has higher accuracy than other algorithms.(3)The localization algorithm based on particle filter is studied.In this algorithm,laser odometry is used to predict the pose of climbing robot at the next moment.Andthen the point cloud registration is used to update the weight of particle swarm,which can correct the prediction results.Theoretical analysis shows that the estimated pose based on laser odometry is more accurate in the prediction stage.In the correction stage,the data structure of Octo Map improves the calculation efficiency of point cloud registration.The experimental results showed that the algorithm could track the trajectory of climbing robot with the the average localization error 0.2m.