复杂因素影响下舰载机着舰短期航迹预测方法研究

Landing signal officer (LSO) is the safety guarantor of carrier-based aircraft landing. However, LSO directs carrier-based aircraft landing on issues like poor adaptability to the environment and being easily influenced by subjective factors. Carrier Landing Approach Decision Aid, which is an alternative solution to LSO directing, has the advantage of higher autonomous and adaptation to the environment, it is the future developmental trends of carrier landing security assurance. Carrier landing short-term path autonomous prediction is the kernel of Carrier Landing Approach Decision Aid, precision short-term path prediction is the basis of the high reliable Carrier Landing Approach Decision Aid and is the fundamental guarantee of carrier landing safety. Since the existing flight path prediction approaches are unable to combined characterization the complicated factors, such as carrier velocity, deck motion, carrier airwake, natural environment and pilot skill, the prediction accuracy can’t meet the safety margin of carrier landing. It is the constraint of carrier landing development and applying. This research aiming at the complicated factors of carrier landing, will propose a novel path prediction approach which is in lazy learning pattern. The novel path prediction approach will select the reference paths dynamically by real-time matching the current paths with database paths, so the complicated factors of carrier landing will be implied in the reference paths, after that the variable structure prediction model with the characteristic input structure will be set up basing on the reference paths, then the goal of improving the on fly short-term path prediction accuracy of carrier landing will achieve.

着舰指挥官是着舰安全性的保障者，但人工指挥着舰存在环境适应性差、易受主观影响等问题。着舰辅助决策系统可替代着舰指挥官，具有自主性高、环境适应性强等优点，是着舰安全性保障的未来发展趋势。着舰短期航迹自主预测是着舰辅助决策系统的核心，高精度短期航迹预测是着舰辅助决策系统高可靠性的基础，是着舰安全性的根本保证。然而，现有飞行航迹预测方法缺乏对航母速度、甲板运动、舰尾流场、自然环境、飞行员技能等复杂着舰影响因素的综合表征能力，航迹预测的准确性无法满足着舰安全裕度要求，制约了着舰辅助决策系统的发展和应用。本项目针对着舰复杂因素的影响，拟提出一种懒惰航迹预测方法，通过舰载机当前航迹与数据库航迹的多因素实时匹配动态选取参考航迹，将复杂因素的影响蕴含于参考航迹中，进而基于参考航迹在线建立带有特征输入结构的变结构预测模型，达成提高舰载机着舰航迹在线预测精度的目的，为舰载机着舰安全性的自主保障提供理论基础。

预测舰载机的未来着舰航迹，进而预先修正着舰航迹，是避免撞舰事故、保障着舰安全性最直接、有效的手段。目前着舰安全保障主要由着舰指挥官（LSO）预测未来2秒的着舰航迹，并以此为依据给出航迹修正指令。然而人工着舰航迹预测对LSO的技能要求极高，受能见度影响大、且预测结果因人而异。建立舰载机着舰航迹在线预测模型，利用舰载雷达的测量数据实现着舰航迹的高精度在线预测是替代人工航迹预测解决上述问题的有效途径，同时着舰航迹预测也是着舰安全保障的自动化、自主化的发展方向。根据着舰航迹预测的高精度需求，如何体现着舰航迹的不确定动态特性是研究的难点，也是提高预测精度的关键。现有基于机器学习的飞行航迹预测方法都采用离线学习算法，即利用历史数据建立统计意义模型，这种模型只反应航迹的普遍规律，无法体现预测航迹的局部动态特性，相比之下，在线预测模型利用多个局部最优预测模型逼近全局最优预测模型，具有预测精度高、泛化能力稳定的优点。据此，针对着舰航迹的显著动态特性，我们提出一种基于航迹匹配的着舰航迹在线预测模型（OTPMT），该模型通过航迹匹配在线获得训练航迹集，利用少量训练航迹在线建立航迹预测模型。其中为了缩短算法的复杂度和建模时间，我们将航迹匹配蕴含于历史着舰航迹的存取过程中，提出一种基于数据容器的航迹快速匹配方法。为了使在线模型同时体现航迹的全局普遍特特性和局部动态特性我们，将批量学习算法和RBF结构参数在线调整方法相结合，提出基于小样本批量学习的LM在线预测模型。仿真实验表明ON-TP-MP能够满足指挥决策对航迹预测的需求。对比分析了OTPMT和离线预测模型的性能，得到OTPMT在预测精度上和动态环境适应能力均优于离线预测模型模型的结论。

内容获取失败，请点击重试