摘要
针对AI模拟飞行研究,提出应用平台Gym与Flightgear模拟飞行软件相结合,构建AI模拟飞行训练平台。通过平台优化,可不断增加飞行动作的难度系数,重新设计奖励机制与神经网络,实现由AI操控模拟飞行软件向AI反馈训练数据的交互闭环。训练结果验证了该训练平台的有效性。
Aiming at the high research significance of AI simulation flight research,it is pointed out that the application platform Gym is combined with Flightgear flight simulation software to construct an AI simulation flight training platform.Through platform optimization,the difficulty coefficient of flight action can be continuously increased,and the reward mechanism and neural network can be designed to realize the interactive closed-loop feedback of training data from AI control simulation flight software to AI.The training results verify the effectiveness of the training platform.
作者
刘剑超
董斐
林亚军
俞艺涵
姚杰
Liu Jianchao;Dong Fei;Lin Yajun;Yu Yihan;Yao Jie(No.91475 Unit of PLA,Huludao 125000,China;No.92543 Unit of PLA,Changzhi 046000,China)
出处
《兵工自动化》
2023年第6期38-40,共3页
Ordnance Industry Automation
