Knowledge Representation And Intelligent Reasoning Of Service Robots

With the deepening of artificial intelligence research, service robots are gradually walking into human life and becoming the life helper. Compared with the structured and fixed industrial environ-ment, human life environment has the characteristics of changing dynamics and unstructured. Many uncertain factors may affect the environment, for example, the daily activities of human beings, the position change of furniture items. When performing various tasks in the human living environ-ment, robot need to have the information of the various surrounding objects, including the object attributes and location, or some other related knowledge. Meanwhile, in order to understand the human instruction in natural language, robot requires the capability of knowledge-based intelligent reasoning. This paper describes a method to build the knowledge base for service robots, gives the knowledge interactive interface, and studies the service robot’s intelligent reasoning method based on knowledge base, which can be used to understand the human instruction in natural language.Specifically, the research content of this paper includes the following three parts:1. According to the service robot’s requirements when performing tasks in the human life en-vironment, we give the OWL-based knowledge representation and the construction method of knowledge base for service robots, as well as the definition of the interactive interface of knowledge base.2. we present a simultaneously active localization and object search method based on Partial Observable Monte Carlo Planning algorithm, which improves the efficiency of object search, and reduces the localization uncertainty efficiently at the same time. This method makes decision on optimal action by searching from a simulation tree which is constructed according to current observation with a certain steps simulation on the location sample and action. We proposed to calculate the selection reward of some action according to observation uniqueness of localization, entropy reduction of location probability distribution and object search’s utility function. In addition, Bayesian network is introduced to represent the conditional probability between objects and locations, thus we update the knowledge base by updating the conditional probability of Bayesian network with regard to the search result, which helps robot to adapt to the dynamic changes of object location. Simulation experiments in the grid map with a large number of self-localization ambiguity verify the effectiveness of the algorithm proposed.3. As service robots need to have the ability to understand natural language in the living en-vironment, we explore the knowledge-based natural language understanding and intelligent reasoning framework for service robots, give a pipeline of processing natural language to search structured knowledge base, which solves the problem of interaction between unstruc-tured natural language and structured robot knowledge base. The method’s performance is validated by practical applications.