Research On Positioning And Detection Methods For Mobile Wall Polishing Robot

With the adjustment of the market economy structure,the real estate industry is facing the pressure of transformation.Not only is the demographic dividend weakening,but the employment outlook of the labor force has undergone tremendous changes.For example,jobs such as wall sanding,which is a heavy task and in a harsh environment,are unattended.Therefore,it is particularly important to introduce unmanned wall grinding robots into the construction industry.At present,mobile operation robots equipped with lidar have been widely used in the construction process.Due to the limitation of the accuracy and hardware performance of lidar itself,as well as the cross influence of obstacles and moving objects,the robot cannot be accurately positioned and it is difficult to adjust the motion posture in time..Therefore,based on the static data of the laser total station,this paper proposes a pose detection strategy for point cloud data fusion,and uses the high-precision overall data of the static total station to fuse the local and low-precision data of the laser radar in the robot.data,and complete the pose detection of the wall grinding robot through a new algorithm.The main research contents include:Firstly,the relationship between point cloud matching and posture correction is illustrated.The interfusion of point cloud data in different coordinate systems is achieved by rotation and translation,and the correspondence between the two point clouds is found;i.e.the correspondence between the current posture of the robot and the coordinates of the working area.Next,the relevant point cloud processing technique is proposed to provide optimisation for the subsequent matching algorithm.Next,a point set weight matching method is proposed for improving the positional recognition capability of the wall sanding robot,where the filtered datum features are combined in a certain way and the point set weights are used to solve for the matching error between total station data and Li DAR data.Finally,the method is validated by simulation experiments,which show that the method has a better matching effect.Then,a matching method with vector weights for fast fusion of high-precision static globaldata and low-precision dynamic local data is proposed.First,a fast preliminary matching of total station data and Li DAR data is achieved by convolutional matching of a sampling network,then a systolic interest interval is constructed for adaptive search of baseline features and mapped to the corresponding area of Li DAR data,and finally,the posture parameters of the wall sanding robot are obtained by vector weight matching.Finally,taking the work execution process of a wall-sanding robot as an example,a wallsanding environment simulation experimental platform is designed and given a selection of the equipment to be carried on the experimental platform,and finally the wall-sanding environment simulation experimental platform is built independently.The experimental platform is used to verify the feasibility of the vector weight matching method.The experimental results show that the vector weight matching method can achieve a positioning accuracy of ±7mm and an attitude control recognition accuracy of ±1.4° in a 6m×8m×2m experimental space,breaking through the actual scanning accuracy of Li DAR,reducing the scanning error of Li DAR and greatly improving the efficiency and economic benefits of mobile robot position recognition.