Research And Application Of Method Combining Domain Knowledge In AI Planning

AI(Artificial Intelligence)Planning is an important part of Artificial Intelligence.AI Planning and current popular areas such as machine learning and deep learning belong to different branches of Artificial Intelligence,which is Artificial Intelligence in the traditional sense.In recent years,the field of AI planning has made new research progress in heuristic search method and planning model design method,further promoting the development of AI planning.However,even the current advanced methods like heuristic are not necessarily effective for complex large-scale planning problems.In addition,AI planning,as an extremely applied field,is still not widely used in reality,and the main reason is that the design process of planning model is relatively complicated.In summary,this thesis focuses on the above two core issues,exploring the multi-heuristic methods in different specific fields,and combining the knowledge engineering to research the design method of planning model and to use the planning model to solve the real world problems.The specific work content is as follows: 1.The method based on heuristic search is the current advanced planning method.This thesis proposes a method to solve the problem by using the advantages of multiple heuristics,namely the multi-heuristic fusion method,after analyzing the problems of the single heuristic method in the process of searching.In the method,a strategy of setting the heuristic queue priority is proposed for the heuristic selection problem.The next state to be extended is selected according to the priority of different heuristic queues,and then the planning solution is searched quickly in the planning process based on the greedy idea.And the algorithm of multi-heuristic fusion in planning process is given.Finally,it shows that the method can effectively solve large-scale problems and has wider applicable fields than a single heuristic through a large number of experiments and comparisons in the basic test fields of the International Planning Competition.2.The primary problem for AI planning for practical applications is to design the planning model.Considering the difficulty of designing planning model with existing knowledge,the thesis proposes the design method of planning model based on domain knowledge.After studying the relevant characteristics of theparking planning problem,the specific APPG(AI Planning for Parking Guidance)method was proposed according to the concept of AI planning.Finally,the parking planning model is designed according to APPG method,and the validity of the model is proved by experiments.3.There is no research and application of AI planning method to guide parking navigation.This thesis analyzes the current situation in the field of parking,and proposes a variety of methods to expand the parking planning model to meet the practical application requirements,and then simulates the different scenarios to verify the validity of the extended planning model.Finally,the proposed multi-heuristic fusion method is used for experimental comparison in the extended planning model,and the optimal solution of parking planning model under different scenarios is analyzed.At the same time,the planning ability of multi-heuristic fusion method in solving large-scale problems is verified.