Monte Carlo Tree Search Method And The Application In Autonomous Mission Planning For Stealth Aircraft

The research on the planning method of the Online Autonomous Mission Planning for stealth aircraft combat is very important for improving the combat capability of the stealth aircraft.Because of the stealth aircraft need to maintain stealth capability in the course of tasks,the stealth aircraft should not transmit radio waves,and the use of airborne radar should be reduced,but with the development of the air defense weapon,the battlefield threat environment exists a stronger risk and uncertainty,and the operational task constraints have more obvious diversity,which is a dynamic and uncertain environment.This paper is based on the groud terget tasks of stealth aircraft,and attempts to resolving the autonomous mission planning problem.Based on the theory and method in the autonomous mission planning of stealth aircraft,aiming to the limit of Monte Carlo Tree Search in the problem,the research of the model and Monte Carlo Tree Search algorithm for the autonomous mission planning of stealth aircraft.This article mainly carries on the following research:1.Autonomous mission planning Analysis and modeling of stealth aircraft.This analysis is carried out from two aspects: mission profile and task environment.Based on the analysis of the uncertainty in the operational process of stealth aircraft,the sequential decision-making characteristics of stealth aircraft are summarized.Considering the characteristics of autonomous mission planning of stealth aircraft,the online planning of stealth aircraft is modeled,and the autonomous mission planning model based on Partially Observable Markov Decision Processes(POMDP)is constructed,which lays the foundation for the further research.2.The Monte Carlo Tree Search algorithm is systematically studied,and an improved Monte Carlo Tree Search algorithm for autonomous mission planning of stealth aircraft is proposed.The paper introduces the Monte Carlo Tree Search algorithm,expounds the characteristics of the algorithm,and analyzes the problem that the Monte Carlo Tree Search algorithm,Partially Observable Monte Carlo Planning(POMCP),dependents strongly on the number of Monte Carlo simulation.Aiming at this problem,the POMCP-RAVE algorithm is designed by combining the All Moves As First(AMAF)heuristic information.The improved algorithm is tested on the benchmark problem Battleship and Rocksample,the average return is averagely 11.99% and 17.12% higher respectively in the two peoblems with same simulations,and the effectiveness of the improved algorithm is verified.3.The study of application of the autonomous mission planning of stealth aircraft is carried out,the applicability and effectiveness of the improved Monte Carlo tree search algorithm is verified.Based on the actual background of the stealth aircraft,the specific task of stealth aircraft is set up.The POMDP model is used for autonomous mission planning of stealth aircraft,and the improved Monte Carlo tree search algorithm,POMCP-RAVE,is applied to plan the problem.The efficiency of the proposed algorithm is verified by the experimental results,and the applicability and effectiveness of the proposed model and the improved algorithm are verified.The simulation data show that the improved model and algorithm are reasonable and effective,and it is the effective framework for solving the problem of the autonomous mission planning of stealth aircraft,and put forward a new method for the improvement of mission planning technology of stealth aircraft.