Research On Global Localization Of AGV Navigated By LADAR In Feature Map

Automatic guided vehicle(AGV)has been developed rapidly in intelligent manufacturing and logistics system and its global localization is one of the research hotspots in autonomous navigation.Markov localization algorithm is a global localization method based on probability distribution,which is widely used.It can solve the problem of multi model and nonlinear so it has received extensive attention and research.At present,there is no applicable probabilistic global localization method in feature map.Based on the relevant models of AGV Markov localization method,the global localization problem in feature map is studied.Aiming at the problem that it often appears that the correlation between the sensor observation and the feature data of the map is not unique when using Markov positioning algorithm in feature map in global positioning of AGV.This paper presents a new method of Markov location calculation without data association.By using the Gauss kernel function,the sparse features in the environment are combined to form a smooth dense curve,and the observation model of Markov is calculated by comparing the similarity of the two dense curves obtained from the observation and the prediction of the algorithm.At the same time directly using electronic compass sensor AGV attitude information,which makes the algorithm reliability focus on solving AGV discrete position without 3D data and calculate the AGV pose,based on conventional Markov positioning method in a symmetrical environment failure problem reduces the amount of computation.The validity of the global localization method was verified by simulation analysis.According to the problem of large amount of calculation and low efficiency of Markov localization algorithm,this paper presents a new method of variable resolution discretization of the planar grid based on the four fork tree model and the direct attitude of the AGV is got from the electronic compass information.The computation efficiency and the rate of convergence of the extreme value of the algorithm are improved by reducing the computation of the grid region which is usually zero in the map.The simulation results show the effectiveness of the variable resolution grid discretization method.Compared with the fixed resolution Markov localization method based on Gaussian kernel function,the proposed method is more efficient for global localization of AGV.Finally based on the global positioning experiment of the AGV in semi-closed environment,the trajectory estimated by the method was compared with the trajectory estimated by extended Kalman filter method.Though initial position of AGV is unknown,the accuracy of the AGV trajectory estimated by the method proposed is still higher and the positioning result is more effective.