Research On Map Merging Method For Large-Scale Environment

Multi-robots mapping is better in a large area environment.The goal of multi-robots mapping is to share map data.Map merging technology can solve the key problem that how to construct a sharing map which is the core problem of multi-robots mapping.For different environments,there are many choices for the research objects of map merging,because the maps with different expressions focus on providing different environmental information.The methods of constructing different maps are different.And the map merging methods of different expressions have their own characteristics.Therefore,the study of map merging method for large area environment has high research value.This paper will carry out in-depth exploration on the topic of map merging methods oriented to different representation.The research object of map merging in this paper is the grid map and the hybrid topological map.The specific work is as follows:(1)The construction of the 2D topological map relies on offline map,which means little information in topology nodes.Therefore,this paper designs a hybrid topological map.The grid map is embedded into topology nodes,increasing information and facilitating the research of map merging.The grid map is to divide the environment into three regions by using sensor and mathematical calculation.This paper will introduce three algorithms of construction of the grid map and analyze their performance.In the construction of hybrid topological map,the local sub-grid maps built continuously are recognized for the semantic label by using neural network.Then,topology nodes are extracted according to the temporal continuity and semantic similarity of local sub-grid map.Finally,the hybrid topology map is composed of topology nodes and lines.(2)The commonly method for grid maps merging is based on point feature,which is inefficient.Because point features are got by the analysis of the relationship with the surrounding point pixels.However,only three pixels in grid map results in many similar feature pairs some of which are false.Driven by this challenge,a novel map merging method based on suppositional box that is constructed by vertical points and virtual lines is proposed.The paper firstly extracts vertical points selected from the intersection of the vertical line.Secondly,based on the common edge between the vertical points,the suppositional boxes in the map are constructed.Then the transformation matrix is obtained according to the matching pair of suppositional boxes.Finally,for matching errors based on the length of pairs,Kalman filter is used to optimize the transformation matrix.Experimental results show that this method can effectively merge map in different scenes and the successful matching rate is greater than that of other features.(3)The method for a topological map merging is often based on static topological structure,which may not adapt to dynamical environment.To solve this problem,this paper proposes a hybrid topological map merging method based on memory sphere model.The rich information of node in hybrid topological map can provide features and semantic label,meaning that there is no need to rely on static topological structure.At the same time the memory sphere is designed to store features and semantic label in topology node.The memory sphere is a hierarchical structure composed of multi-layer spherical nodes that describe the feature information of topology nodes.Due to the layered characteristics of the memory sphere,an improved layered matching algorithm is used to match memory spheres and merge maps.We verify the effectiveness of our method in simulation and real environment respectively.The influence of semantic label error on hybrid topological map merging is analyzed.