首页|基于深度强化学习的多无人机空战机动策略研究

基于深度强化学习的多无人机空战机动策略研究

扫码查看
原文链接
  • 国家科技期刊平台
  • NETL
  • NSTL
  • 万方数据
面对敌方空中力量的来袭,具有自主协同、灵活机动能力的无人机是参与空中作战的重要力量.面向多无人机协同高制胜率的对抗作战任务需求,并根据空战目标数量划分,重点对多无人机对单目标协同空战机动策略和多无人机对多目标协同空战机动策略展开研究.本文主要分析了空战过程中的关键战场要素,基于多机机动特性,建立无人机运动模型.根据无人机火控特点,分析无人机状态变化规律,建立无人机对敌攻击模型和动态对抗模型;针对多无人机对单目标自主协同空战问题,提出基于专家规则和强化学习相结合的多机自主机动策略.仿真结果表明,所提算法可以完成态势实时变化的多机对单目标空战任务.在作战双方数量相同的前提下,若敌方不具备智能机动行为,我方制胜率为 100%;即使双方采用相同的策略,如果我方数量多于敌方数量,我方仍有大的胜率,体现了协同策略的有效性.
Research on Multi-UAV Air Combat Maneuver Strategy Based on Deep Reinforcement Learning
In face of the incoming attack of enemy air power,UAVs with autonomous coordination and flexible maneuvering capability are an important force to participate in air combat.Facing the demand of confrontation combat mission with high winning rate of multi-UAV coordination,and based on the number of air combat targets,we focus on the research of multi-UAV to single-target coordinated air combat maneuver strategy and multi-UAV to multi-target coordinated air combat maneuver strategy.This paper mainly analyzes the key battlefield elements in the process of air combat,and establishes the UAV motion model based on the characteristics of multi-machine maneuver.According to the fire control characteristics of UAV,analyze the change rule of UAV state,establish UAV attack model and dynamic confrontation model against the enemy;for the problem of multi-UAV to single-target autonomous coordinated aerial combat,put forward multi-autonomous maneuver strategy based on the combination of expert rules and reinforcement learning.The simulation results show that the proposed algorithm can accomplish the task of multi-aircraft aerial combat against single target with real-time change of situation.Under the premise of the same number of combatants,if the enemy does not have intelligent maneuvering behavior,our victory rate is 100%.Even if both sides use the same strategy,if our number is more than the enemy,we still have a large victory rate.This demonstrates the effectiveness of the coordinated strategy.

air combat strategyreinforcement learningautonomous mobilitymultiple machine collaborationsituation assessment

雷毅飞、王露禾、贺泊茗、胡劲文、徐钊、吕明伟、徐港

展开 >

西北工业大学,陕西 西安 710129

航空工业沈阳飞机设计研究所,辽宁 沈阳 110034

空战策略 强化学习 自主机动 多机协同 态势评估

国家自然科学基金航空科学基金航空科学基金陕西省重点研发计划项目中国博士后科学基金

618033092019ZA053008201855530342020ZDLGY06-022018M633574

2024

10.19452/j.issn1007-5453.2024.03.013

航空科学技术
中国航空研究院

航空科学技术

影响因子：0.24
ISSN：1007-5453
年,卷(期)：2024.35(3)
  • 17