Research On The Construction Of Semantic Map Based On Cloud For Service Robot

For the environment map,the degree of anthropomorphic description is an important factor to restrict the intelligent navigation of the robot.Self-building capability in complex dynamic environment is an important indicator that determine the intelligent level of service robot.Based on the cloud-based environmental semantic framework,this paper constructs the item base,the ontology knowledge base and realizes the semantic map,which apply the original items,region and other functional awareness to the service task.Aiming at the problem that the robot can not obtain the semantic of all kinds of object in the intelligent service task of complex environment.In this paper,the cloud and construction of semantic map are combined to form cloud-based environmental semantic framework.Although the cloud platform can provide us with rich image data,in order to enable the robot to serve a diverse,personalized environment,we need to further add the sample data.Based on Caffe,we can obtain the semantic information.Support vector machine(sSVM)and point cloud are used to classify the semantic base and feature base.Finally,the semantic-CVFH base is realized.The robot downloads the required semantic-CVFH base from the public cloud to the private cloud according to the work scene(eg,the home kitchen),and used for object recognition in this work scene.In order to optimize the form of semantic map,items are divided into identification items and attribution items(including explicit attribution and hidden attribution).By querying the semantic-CVFH base to obtain the semantic,the identification items and the explicit attribution items are stored in identification base and attribution base.By processing the point cloud,locative relationship of Identification-Attribution is realized.The ontology knowledge base is designed to solve the problem of hidden attribution and room function.The ontology model based on human,object and room is established,the instance attribute,object attribute and instance are defined.Using cloud resource to improve the model and persistent storage to form ontology knowledge base.The hidden attribution items are stored in attribution base to achieve the association with semantic map.Based on the cloud environmental semantic framework,the construction of semantic map is completed.The robot provides structured environment information(including grid map and self localization).After obtaining semantic of the items at a physical coordinate,the identification item is inserted into the current position and the Identification-Attribution relationship is recorded by XML,thereby forming semantic map.The proposed method is closely related to the working environment of the robot.The rich data of cloud and robot map closely related to transform the robot map more knowledge and humanization.This work has important scientific significance and practical value for deepening and perfecting the research of the cloud robot and accelerating its development and application.