Research And Application Of Autonomous Cooperative Planning Technology For Multi-Unmanned Platforms Based On Domain Knowledge

Multi-unmanned vehicles cooperative operation has the advantages of good spatial distribution and high fault tolerance,and is of important application value in military and security fields.Under the complex operational environment,the cooperative operation of multi-unmanned vehicles is faced with such problems as task complexity,combination diversity and real-time response,making the research on the cooperative operation technology of multi-unmanned vehicles become a very challenging subject.Nowadays,artificial intelligence based on domain knowledge is regarded as the future development trend because it is similar to human thinking mode.Based on multi-unmanned vehicles,this article focuses on the study of cooperative mission planning mechanism based on domain knowledge.Firstly,we construct the domain knowledge model of multi-unmanned vehicles cooperative operation.Secondly,we study the rough decomposition of high-level combat command based on rules and the detailed decomposition of subtask based on timing planner.Finally,the transformation method between domain knowledge representation and domain description file of timing planner is proposed.The main research work and contributions of this paper are as follows:(1)The ontology-based knowledge representation and modeling of the domain of multi-unmanned vehicles cooperative operation.In this paper,ontology is adopted to represent and model the knowledge in the field of unmanned vehicles cooperative operation.The ontology is divided into four categories,namely entity ontology,environment ontology,task ontology and state ontology.Knowledge representation structure design and instantiation are carried out respectively.In addition,this paper proposes a method to evaluate the scalability and responsiveness of the constructed robot knowledge base.The evaluation results show that the constructed ontology in this paper has good scalability and responsiveness.(2)The rough decomposition of high-level combat mission instructions based on rules.On the basis of the domain ontology,this paper proposes to introduce SWRL rules into the knowledge base to build a rule base,so as to make up for the lack of representation of rule-based knowledge in OWL.Finally,based on the domain ontology knowledge base and rule base,the typical combat sub-task set obtained from the rough decomposition of the high-level task and the associated information of the combat execution unit are deduced.(3)The subtask decomposition based on timing planner.In this article,POPF task planner is used to decompose subtasks in detail to generate the final atomic action sequence.This paper proposes a method of generating POPF domain description file based on domain ontology knowledge,which overcomes the problem that traditional domain description file needs to be written manually.Finally,based on POPF task planner,the subtasks were decomposed in detail,and Turtle Bot3 was used for physical experiment to verify the usability of the domain knowledge ontology constructed and the correctness of the domain knowledge transformation algorithm.(4)The autonomous cooperative task planning experiment of multi-unmanned vehicles based on typical unmanned combat missions.The multi-robot patrol virtual simulation framework based on ROS and Stage was used to conduct the autonomous cooperative planning experiment of multi-unmanned vehicles for the combat scenarios including patrol and strike tasks.In the experimental process,the corresponding optimization and scheduling methods were proposed for the possible operational conflicts.