UAV Swarm Confrontation Autonomous Decision-Making Method

Large-scale,low-cost UAVs to penetrate and attack in swarm are one of the important combat modes of future warfare.They have the advantages of good concealment,strong maneuverability,and low economic cost.Traditional defense methods are facing huge challenges.UAV swarm attack operations pose a severe threat to the protection of high-value targets.The use of UAV swarm to counter enemy UAV swarm in a "group-to-group" approach has the advantages of agility,efficiency,and cost symmetry,and is an anti-swarm technical route with great development potential.From the perspective of UAV swarm confrontation,this thesis conducts research work related to UAV swarm intelligent confrontation and autonomous game,aiming at the problem that pre-programming and remote real-time control methods cannot adapt to the complex battlefield environment.Aiming at the problem that traditional mathematical analysis,expert system and other methods cannot make efficient and accurate decision-making in the face of uncertain battlefield environments,the study uses deep reinforcement learning algorithms to train UAVs based on real-time battlefields.The main research contents are as follows:(1)According to the characteristics of the number and scale of UAV swarm,research the battlefield situation assessment and target allocation methods under the conditions of swarm confrontation,and solve the key technologies such as complex three-dimensional situation assessment and effective target allocation plan in the process of swarm confrontation.Build a UAV swarm confrontation model and simulation platform for training and verifying maneuvering strategies.(2)Aiming at the single-aircraft-to-single-aircraft UAV air combat scenario,construct the UAV three-dimensional air combat Markov decision process model,and study the UAV autonomous maneuver decision method based on deep reinforcement learning algorithms.Optimization methods improve the autonomous decision-making learning effect of UAV in confrontation scenarios.(3)Aiming at the problem of multi-UAV coordinated countermeasures,establish a multi-machine coordinated countermeasure learning task mathematical model,study the modeling of multi-machine cooperative reward function,and multi-machine based on multi-agent reinforcement learning algorithm.Carry out multi-aircraft confrontation simulation and tactical analysis.(4)Aiming at the scenarios of UAV swarm penetration and interception,the Monte Carlo method is used to conduct simulation tests to analyze the impact of different UAV numbers,different UAV performances and different swarm tactics on the interception effectiveness,and summarize the operational rules.The thesis explores the application of deep reinforcement learning in the autonomous decision-making field of UAV air combat,and the research results provide theoretical support for the realization of swarm confrontation.By building a simulation platform to analyze the UAV swarm combat process and summarize multi-aircraft cooperative tactics,it has important reference value for subsequent research and formulation of UAV swarm confrontation strategies.