Research On Autonomous Task Planning And Execution Of Service Robot Based On Natural Language

As service robots are starting to perform everyday manipulation tasks,complete knowledge storage and autonomous planning capabilities are critical for robots to provide intelligent service tasks.The design of the project is to improve the knowledge storage of the service robots by acquiring the natural language information for home services in the Internet and combining the advantages of the smart space,which improves service decision-making ability.Focusing on complex tasks,this paper proposes a framework for robot autonomous planning and execution based on natural language tasks.Firstly,build a natural language service information database so that the robot has the ability to obtain service information independently.Then the robot control instruction structure is designed to use the natural language understanding algorithm to train the constructed natural language service information database as a corpus,extract the characteristic information and generate the robot control instructions.On this basis,the semantic knowledge expression is introduced into the autonomous planning field,and the STRIPS standard ontology model is constructed to integrate and organize the information in the smart space for providing a unified interface.The control instruction is converted into a robotic action execution sequence,and the experiment is performed on the Unity simulation platform.(1)Construction of Natural Language Service Information Database When assigning a complex task,the robot needs to obtain natural language information from the network to complete this task.In this paper,the improved topic-focused crawler algorithm is used to realize the autonomous acquisition of the knowledge required by the service robot for the operation task.The LDA topic modeling is used to map the subject and content of the acquired information.Utilize a relational database as a storage tool and set up a dynamic information management mechanism to implement the construction of a natural language service information base.(2)The robot control instruction is designed to realize the conversion from the natural language information to the robot control instruction,which parses the task in the information database through the natural language understanding algorithm.Based on the semantic parsing and phrase chunk analysis methods,the key information of the natural language task is extracted using a method combined rule matching and conditional random field algorithm.Through the semantic similarity analysis,the key information is mapped to the robot action lists,in which the acquired natural language service information is converted into a robot control instruction.(3)Semantic ontology technology is used to conceptualize the knowledge level of information.Construct intelligent space ontology modules—environment ontology and task ontology to abstract the underlying knowledge of context awareness into upper layer knowledge.Use JESS inference engine based on SWRL rules to mine hidden knowledge.On this basis,a recursive search algorithm is used to combine the action execution sequences of service tasks to realize autonomous planning of service tasks in a smart space.Finally,the planning result is mapped to the action function of the executing agency,and a simulation platform including a mobile robot with a robotic arm and a smart space environment is built on the Unity software to verify the feasibility of the method.