| With more and more service robots coming into our work and life,people have higher and higher requirements for the quality of service of robots.Therefore,the construction of environmental information representation model suitable for robot service tasks has become a hot research topic.The environmental information involved in the task of service robots has the characteristics of diversity,semantics,hierarchy and relevance.However,the existing representation methods of environmental information,such as predicate logic and ontology,emphasize the concepts and conceptual relationships in the environment,not only is it difficult to express the above characteristics of environmental information,but also there are some problems such as low visualization of knowledge representation,poor human-computer interaction,insufficient knowledge reasoning ability,slow knowledge search speed,which can not meet the needs of intelligent and personalized services.This paper designs an indoor environment information representation model based on knowledge graph,uses knowledge graph technology to mine,analyze,construct,draw and display knowledge,and organizes storage in the form of robot knowledge base to enhance the generality,systematicness and intuitiveness of environmental information representation,and to improve reasoning ability and retrieval efficiency.The main work of this paper includes the following aspects:A knowledge graph-based representation model of environmental information is designed,which integrates the environmental information perceived by the robot ontology with the semantic description information such as the functional attributes acquired by the Internet to form a structured and logical reasoning knowledge base of service robots,including the conceptual layer and the data layer.The data sources in the knowledge base are wide and heterogeneous,and there are differences in entity semantics and knowledge system between different datajsources.A multi-modal entity alsources. fusion based on vector representation and an architecture based on How-net are designed to unify the descriptive information of indoor environment,such as features,objects and contexts,and achieve the automatic representation of environmental information system.In order to mine the hidden high-level knowledge in knowledge base,a knowledge reasoning mechanism based on Markov logic network is designed.In this paper,we present a method of obtaining path features of knowledge graph by random walk,and inference rules in knowledge base are learned by path features.Learn the weights of inference rules in Markov's logical network to improve the accuracy of reasoning.The Markov inference model is used to refine the knowledge graphs of relationships,semantics,types,etc.,predict and mine the relationship between information,improve the coverage of knowledge graphs,and realize the development of knowledge base.In the face of the robot service task demand,the concept of semantic map is introduced.Robot can obtain structured environmental map,according to the knowledge base based on knowledge graph,semantic map is generated by associating the semantic information of objects with structured map.Since the objects in the knowledge base are connected to each other in the form of a relationship,when the service robot retrieves the objects in the knowledge base,not only the semantics and attribute information of the object but also other objects associated with it can be obtained.Information can enhance the robot's task decomposition and mission planning capabilities to achieve intelligent services.The method proposed in this paper combines the robot and human service requirements closely.The introduction of knowledge graph technology not only enables the knowledge base of service robots to have the ability of structuring and hierarchical expression of environmental information,but also has the "association"similar to human brain,laying a foundation for the intelligent tasks of robots.This work has important theoretical and practical value for realizing robot knowledge base construction and intelligent service. |
PDF Full Text Request