基于分层式强化学习的移动机器人导航控制

针对未知环境下的移动机器人导航问题,本文提出了一种基于分层式强化学习的混合式控制方法。利用栅格-拓扑相结合的环境表示及地图学习方法,通过分层式强化学习在不同控制层次的扩展设计移动机器人的反应式和慎思式导航控制,实现了全局导航和局部导航控制的协调优化。实验及测试结果证明,该控制方法能实现导航任务的全局优化,避免陷入局部极小,并对未知动态环境具有较强的适应性。     

基于深度学习的发动机叶片故障检测技术

为了解决航空发动机叶片故障检测中存在的检测精度欠佳、检测效率不高的问题,提出了一种基于深度学习的目标检测方法。针对小样本数据集检测精度低、模型训练速度慢等问题,对Faster R-CNN目标检测算法进行结构优化,引入Res2Net结构,通过分割串联的策略强化残差模块的卷积学习能力,搭建了细粒级的多尺度残差模型Res2Net-50,以提升模型的特征提取能力。同时,在网络的训练过程中,采用多次余弦退火衰减法对学习率进行调整,以加快模型的训练速度,提升模型的训练质量。针对航空发动机叶片裂纹和缺损2种故障类型进行网络训练与检测试验,试验结果表明：优化后的模型识别准确率提高了0.7%,模型的平均检测精度提高了1.8%,训练时间缩短了5.56%,取得了比较好的检测效果。     

电子对抗技术发展综述

随着雷达低截获技术、通信系统抗干扰技术的发展和战场上智能化电子信息装备部署的应用,战场电磁频谱环境的复杂程度日益增加。如何提高电子战感知环境-适应环境-做出决策-采取行动(OODA)环路速度,实现电磁敏捷,是电子战应对未来复杂战场环境挑战需要解决的关键科学技术问题。通过分析现代雷达技术发展对传统电子战系统的挑战,阐述认知电子战发展的必要性,并从认知概念演变、人工智能技术发展、认知电子战关键技术等方面,对传统电子战向认知电子战发展过程中人工智能技术的应用情况进行分析与总结。     

Key Problems in Coordinated Air Combat andMulti-agent Reinforcement Learning

自从协同作战的概念提出后,各军事强国在协同空战领域均取得了重大进展,协同成为提升作战能力的倍增器。近数十年来,作为解决序列问题的现代智能方法,强化学习在各领域高速发展。然而,面对高维变量问题时,传统的单智能体强化学习往往表现不佳,多智能体强化学习算法为解决复杂多维问题提出新的可能。通过对多智能体强化学习算法原理、训练范式与协同空战的适应性进行分析,提出了协同空战与多智能体强化学习的未来发展方向,为更好地把多智能体强化学习应用于协同空战提供思路。     

基于深度强化学习的无人机栖落机动控制策略设计

无人机栖落机动飞行是一种无需跑道的降落方法,能够提升无人机在复杂环境下执行任务的适应能力。针对具有高非线性、多约束特性的无人机栖落机动过程,提出了一种基于模仿深度强化学习的控制策略设计方法。首先,建立了固定翼无人机栖落机动的纵向非线性动力学模型,并设计了无人机栖落机动的强化学习环境。其次,针对栖落机动状态动作空间大的特点,为了提高探索效率,通过模仿专家经验的方法对系统进行预训练。然后,以模仿学习得到的权重为基础,采用近端策略优化方法学习构建无人机栖落机动的神经网络控制器。最后,通过仿真验证了上述控制策略设计方法的有效性。     

期望状态序列导向的深空探测器规划修复方法

针对探测器动作持续、并行、消耗能源所导致的规划修复难求解问题,提出一种基于期望状态序列的深空探测器规划修复方法。该方法根据动作的执行完成情况,从既定规划中提取未完成执行动作的期望效果以及待执行动作的期望前提,构成一组有序的、逻辑与能源混合的期望状态集合,为规划修复提供子目标,从而将规划修复问题转化为状态转移路径搜索问题;并在搜索过程中,提出能源补给优先的规划修复策略,将逻辑修复与能源修复分离以降低问题求解难度,最终使得探测器能够自主从规划失败中恢复执行。最后,以火星环绕器为例进行数值仿真,验证了方法的有效性和合理性。     

弱引力下小行星动量驱动机器人的可控跳跃行为

以小行星表面着陆探测为背景,提出一种动量驱动机器人(MoRo)以满足弱引力复杂环境下的探测需求。该机器人利用弱引力环境下的摩擦和碰撞特性,通过主动辨识环境参数,规划和控制动量轮以产生期望的驱动力矩,完成可控性跳跃及腾空后的稳定拍照等任务。首先,基于MoRo的动量轮刹车机构特性,分析了MoRo在弱引力环境下的跳跃机理并对其跳跃方式进行了规划;接着考虑动量轮驱动机构三闭环伺服系统的非线性特性,基于Herze碰撞模型和Karnopp摩擦模型建立了MoRo在小行星表面的跳跃行为动力学模型;其次,使用机器学习算法建立环境参数和MoRo运动的函数关系,并基于环境参数规划动量轮转速实现跳跃距离和腾空高度的可控。最后,通过数值仿真校验了MoRo跳跃规划方法和控制方法的可行性。     

基于隶属度和LMK-ELM的航空电子部件诊断方法

为提高航空电子部件模块级故障诊断精度，提出一种新的面向"软聚类"的局部多核学习（LMKL）-超限学习机（ELM）离线诊断方法。通过引入模糊C均值聚类对样本空间进行模糊划分，挖掘聚类内部多样性的同时，实现了对过学习的抑制；将模糊划分产生的隶属度信息融入LMKL-ELM的优化过程，运用基于初始-对偶混合优化问题的三步优化策略克服了局部核权重二次非凸的问题，在l₁-范数与l₂-范数约束下分别给出了相应的更新方法。将所提方法应用于某型机前端接收机，结果表明：与4种流行的多核诊断方法相比，该方法可有效避免漏警、抑制虚警，在l₁-范数和l₂-范数约束下，其诊断精度比其他方法的平均值分别提升了4.09%和5.13%。     

Gravity waves generated by thunderstorms: South-East Asia tropical region study

Gravity waves are recognized as an integral part of earth’s atmosphere which are mainly responsible for energy and momentum distribution among different layers and regions in the atmosphere. Various sources present in land, ocean, and atmosphere such as mountains, convection, jets and fronts etc. are responsible for gravity waves generation. Thunderstorms (deep convection) are one of the major sources of gravity waves in the tropical region, capable of generating waves with a wide range of frequencies and scales and significantly affecting the existing waves. Previous numerical studies have characterised the wave properties that are generated from thunderstorms, but there are no statistically quantified studies. In this paper, we have modelled the relationship between the latent heat generated inside a thunderstorm and the gravity wave properties at the geo-collocated points. Gravity waves are identified over Singapore radiosonde station (with data available until 30?km altitude with 12?h temporal resolution) in the stratosphere using wavelet studies. Based on the GROGRAT ray tracing methods to identify the thunderstorm locations, and RAMS cloud-resolving models simulations to obtain the latent heating of the thunderstorm, a regression analysis is performed using 200 cases of gravity waves. Furthermore, cloud-top momentum flux analysis is performed for various cases latent heat. This study is expected to provide more quantified and concrete information on the coupling between the thunderstorm and gravity wave which includes the variance in these relationships due to wave frequency spectrum and generation mechanisms.     

Advanced ensemble modeling method for space object state prediction accounting for uncertainty in atmospheric density

For objects in the low Earth orbit region, uncertainty in atmospheric density estimation is an important source of orbit prediction error, which is critical for space traffic management activities such as the satellite conjunction analysis. This paper investigates the evolution of orbit error distribution in the presence of atmospheric density uncertainties, which are modeled using probabilistic machine learning techniques. The recently proposed “HASDM-ML,” “CHAMP-ML,” and “MSIS-UQ” machine learning models for density estimation (Licata and Mehta, 2022b; Licata et al., 2022b) are used in this work. The investigation is convoluted because of the spatial and temporal correlation of the atmospheric density values. We develop several Monte Carlo methods, each capturing a different spatiotemporal density correlation, to study the effects of density uncertainty on orbit uncertainty propagation. However, Monte Carlo analysis is computationally expensive, so a faster method based on the Kalman filtering technique for orbit uncertainty propagation is also explored. It is difficult to translate the uncertainty in atmospheric density to the uncertainty in orbital states under a standard extended Kalman filter or unscented Kalman filter framework. This work uses the so-called “consider covariance sigma point (CCSP)” filter that can account for the density uncertainties during orbit propagation. As a test-bed for validation purposes, a comparison between CCSP and Monte Carlo methods of orbit uncertainty propagation is carried out. Finally, using the HASDM-ML, CHAMP-ML, and MSIS-UQ density models, we propose an ensemble approach for orbit uncertainty quantification for four different space weather conditions.