Hierarchical Task Network Planning And Optimizing For COA Ontologies

In the field of artificial intelligence,the problem of plan generation is a topic that has received widespread attention.The goal of a plan generation problem is to obtain an action sequence that can complete a given task.The generation of operation plan is the application of plan generation problem in the military field.Due to the rapid development of information technology,modern warfare has shifted from traditional mechanized warfare to modern information warfare.And victory or defeat in war is no longer determined only by the scale of weapons and army,but more dependent on the correct and efficient generation of operation plans.Knowledge-Based Systems has designed the COA（Course of Action）Ontology for the U.S.Army and Marine Corps that can be used to support the generation of operation plans.With the development of multidimensional joint operations by sea,land and air,the operational planning environment has become more complex.Correspondingly,the number of instances stored in the knowledge base built on COA Ontology also increases,and the original SPARQL-based action sequence generation method of COA Ontology faces the risk of exponential explosion of computational complexity.Therefore,it is particularly important to explore more efficient methods for the generation of operation plans.Intelligent programming is an important research field in artificial intelligence,and the efficient solving of generation plans is also one of the hotspots in this field.In the field of intelligent planning,hierarchical task network（HTN）planning is a widely used planning method,and the structural and heuristic characteristics of its task network are very suitable for generating operational plans.In this paper,a method combining HTN and ontology reasoning is proposed to solve the problem of plan generation based on COA Ontology.The main content of this paper is divided into two points:1.In this paper,HTN planning technology is used to solve the problem of the generation of operation plan based on COA Ontology.This paper analyzes the relationship between the problem model and the HTN planning model,obtains the mapping relationship between the two models,proposes a series of mapping rules,and uses these rules to generate a HTN planning for COA Ontology:HTN_COA.The original SPARQL-based action sequence generation method of COA Ontology is to map all actions to SPARQL rules,and match the state on the entire rule set.However,because the information of modern warfare is very large,the domain knowledge required to complete a given task is often much smaller than the total domain knowledge.Therefore,it is not optimal to search,match,and plan across the entire rule set when completing a task.HTN planning supports heuristics and constraints,which can be used to exclude knowledge that unrelated to the given task from the planning,thereby reducing search space and reducing computational complexity.2.In this paper,a hybrid reasoning system combining ontology reasoning and HTN is proposed and implemented.This system optimizes the planning process of HTN using ontology reasoning.To support this optimization,this paper expands the new concepts and properties based on the COA Ontology,and proposes four ontology reasoning subtasks.Through experimental verification,the hybrid reasoning system proposed in this paper,which combines ontology reasoning and HTN,has a slower growth trend in planning time with the increase of the number of activities compared to the SPARQL rule matching method used in the original COA-Ontology paper.When the number of actions is greater than 500,the planning speed of the method proposed in this paper is more than twice that of the original method.