The invention relates to an aircraft approach planning method and device based on generative adversarial imitation learning. The method comprises the steps that historical trajectory data and historical flight state data are acquired, an expert approach planning action sequence is determined based on the historical trajectory data, an approach planning model is trained based on the historical flight state data and the expert approach planning action sequence, and the approach planning model comprises a planning module and an evaluation module. The planning module is used for generating an aircraft approach planning action sequence according to historical flight state data, and the evaluation module is used for evaluating whether a planning decision deviation degree and a trajectory conflict or not according to aircraft approach planning actions, inputting to-be-planned flight state data into the trained approach planning model to obtain the aircraft approach planning action sequence, and outputting the aircraft approach planning action sequence to the planning module. And determining a target aircraft approach plan based on the aircraft approach plan action sequence. By constructing the approach planning model and learning the expert trajectory of the real scene to perform approach decision, the efficiency and reliability of approach planning decision are improved.
本申请涉及一种基于生成对抗模仿学习的航空器进近规划方法及装置。所述方法包括:获取历史轨迹数据和历史飞行状态数据,基于历史轨迹数据确定专家进近规划动作序列,基于历史飞行状态数据和专家进近规划动作序列训练进近规划模型,进近规划模型包括规划模块和评估模块,规划模块用于根据历史飞行状态数据生成航空器进近规划动作序列,评估模块用于根据航空器进近规划动作评估规划决策偏离程度和轨迹是否冲突,将待规划飞行状态数据输入训练后的进近规划模型,得到航空器进近规划动作序列,基于航空器进近规划动作序列确定目标航空器进近规划。通过构建进近规划模型,学习真实场景的专家轨迹进行进近决策,提高了进近规划决策的效率与可靠性。
Aircraft approach planning method and device based on generative adversarial imitation learning
基于生成对抗模仿学习的航空器进近规划方法及装置
2024-07-30
Patent
Electronic Resource
Chinese
Automatic parking method and system based on generative adversarial imitation learning
European Patent Office | 2020
|