基于DDPG-PID的机器人动态目标跟踪与避障控制研究

针对机器人在动态复杂环境下的操作问题,研究机械臂跟踪运动目标及避障的运动控制方法。采用传统PID控制方法与DDPG强化学习算法相结合的方式,利用PID控制使得机械臂的工作平面快速接近目标物并与之重合,再使用DDPG算法让机械臂在平面内自主学习追踪目标物投影同时避开障碍物投影,最终在三维空间中实现跟踪与避障。该方法将机械臂作为一个智能体,智能体通过感知目标物和障碍物的距离偏差来自主学习避障跟踪策略。将本方法用于多自由度机械臂路径规划与避障任务中,对比单纯使用DDPG算法将机械臂作为智能体在空间中进行决策控制,仿真结果显示本文所提出的方法收敛效果和控制性能更好,机械臂能在训练后实现目标物的稳定跟踪和障碍物的躲避。     

改进PointNetLK的点云智能配准与位姿图优化方法

针对空间在轨服务任务中的非合作目标相对位姿测量问题,提出一种目标可测部位点云的智能配准方法。首先,通过Straight Through滤波算法对半物理仿真平台采集得到的点云进行目标提取,以消除背景数据等杂乱信息;其次,改进PointNetLK神经网络点云配准算法,将提取后的点云数据作为输入,从而获得初步配准结果,解决非合作目标先验信息缺失导致的无法配准问题;最后,建立基于位姿图的优化模型,以降低配准误差,提高配准精度。实验结果表明,与传统迭代最近点(ICP)算法相比,配准综合误差从6.3598降低到1.7291,精度提高约 72.81% 单次耗时从33.16 s降低到4.2 s,效率提升约87.33%,与当前SM ICP等其他算法相比,也具有一定的优势。     

一种星-箭连接动态界面力的深度学习反演方法

针对于星-箭连接动态界面力无法通过力传感器直接测量,且典型时域动载反演方法难以准确计算界面力的时域变化等难点,提出了基于长短时记忆(LSTM)神经网络的星-箭界面力深度学习反演方法。首先通过卫星地面测试试验得到数据依据,以卫星主体结构的加速度测量数据为输入层,以星-箭界面力测量数据为输出层,利用LSTM神经网络建立输入和输出间的反演映射关系模型,实现卫星在发射过程中较高精度的界面力反演。进而,设计并开展了某典型卫星结构的正弦扫频和随机振动实验,测试LSTM界面力反演方法的可行性。结果分析可知,所提出的基于LSTM深度学习反演方法能够精确地获得动态界面力时程数据,两项性能指标均优于目前典型的载荷反演方法。     

航天器轨道追逃博弈多阶段强化学习训练方法

针对航天器轨道追逃博弈问题,提出一种多阶段学习训练赋能方法,使得追踪星在终端时刻抵近逃逸星的特定区域,而逃逸星需要通过轨道机动规避追踪星。首先,构建两星的训练策略集,基于逻辑规则设计追踪星和逃逸星的机动策略,通过实时预测对方的终端位置,设计己方的期望位置和脉冲策略,显式给出追逃策略的解析表达式,用于训练赋能;其次,为提升航天器的训练赋能效率及应对未知环境的博弈能力,提出一种基于强化学习技术多模式、分阶段的学习训练方法,先使追踪星和逃逸星分别应对上述逻辑规则引导下的逃逸星和追踪星,完成预训练;再次,开展二次训练,两星都采用邻近策略优化(PPO)策略进行追逃博弈,在博弈中不断调整网络权值,提升决策能力;最后,在仿真环境中验证提出的训练方法的有效性,经过二次训练后,追踪星和逃逸星可有效应对不同策略驱动下的对手,提升追逃成功率。     

情感因素对语言学习的影响

情感因素对语言学习具有重大的影响。在学生智商水平相当的情况下,情感差异会导致学生学习成绩的高低之分。情感差异是由家庭、环境、情趣、学习方法等综合因素造成的。积极的情感会对学生的学习起到促进作用；反之,消极的情感则会阻碍学生的学习。教师在教学中应善于引导学生,促进学生积极情感的产生,减少消极情感的作用,以提高教学效果。     

Fabrication of Diamond-like Carbon Films by Ion Assisted Middle Frequency Unbalanced Magnetron Sputtering

Diamond-like carbon （DLC） films are deposited by the Hall ion source assisted by the mid-frequency unbalanced magnetron sputtering technique. The effects of the substrate voltage bias, the substrate temperature, the Hall discharging current and the argon/nitrogen ratio on the DLC film＇s performance were studied. The experimental results show that the film＇s surface roughness, the hardness and the Young＇s modulus increase firstly and then decrease with the bias voltage incrementally increases. Also when the substrate temperature rises, the surface roughness of the film varies slightly, but its hardness and Young＇s modulus firstly increase followed by a sharp decrease when the temperature surpassing 120 ℃. With the Hall discharging current incrementally rising, the hardness and Young＇s modulus of the film decrease and the surface roughness of the film on 316L stainless steel firstly decreased and then remains constant.     

元强化学习综述

元强化学习是指自动从一组相关任务中学习强化学习所需归纳偏置的相关理论和方法,对于提高强化学习算法在困难场景下的样本效率和泛化能力具有重要用途。本文提出一种新的元强化学习框架,指出设计和分析一个元强化学习算法需要同时考虑学习经验（相关任务）、归纳偏置及学习目标3个独立因素及这3个因素之间的依赖关系。在此基础上对该领域的研究现状进行了分析和总结,特别对近年来元强化学习若干文献进行了分析和归类,并详细阐述了几种代表性算法的原理及各自特点。本文还对元强化学习常用的实验环境和性能评价方法进行了介绍,对该领域的不足和未来的发展方向进行了讨论和分析。     

基于 LSTM的飞控系统状态监控

监测飞控系统状态参数是保证无人机飞行安全的重要手段。针对无人机飞控系统的组成特点和飞行控制律,设计并构建了基于长短期记忆网络(Long Short Term Memory Network,LSTM)的飞控系统状态监控模型。首先,利用无人机历史飞参数据训练模型,建立输入飞参数据与状态参数的回归映射关系;然后,利用训练好的网络模型,实时预测飞控系统的状态参数,通过对比实测值与预测值之间的差异,实现飞控系统的状态监控。选取无人机飞参数据进行实验,基于 LSTM的算法比反向传播神经网络(BPNN)、支持向量机(SVM)预测精度高,MSE平均值分别低 0.01和 0.26,MAE平均值分别低 0.05和 0.12。结果表明,所提出的方法能够有效监控飞控系统,为无人机飞行管理决策提供数据支持。     

Stall flutter prediction based on multi-layer GRU neural network

The modeling of dynamic stall aerodynamics is essential to stall flutter, due to the flow separation in a large-amplitude pitching oscillation process. A newly neural network based Reduced Order Model (ROM) framework for predicting the aerodynamic forces of an airfoil undergoing large-amplitude pitching oscillation at various velocities is presented in this work. First, the dynamic stall aerodynamics is calculated by solving RANS equations and the transitional SST-γ model. Afterwards, the stall flutter bifurcation behavior is calculated by the above CFD solver coupled with structural dynamic equation. The critical flutter speed and limit-cycle oscillation amplitudes are consistent with those obtained by experiments. A newly multi-layer Gated Recurrent Unit (GRU) neural network based ROM is constructed to accelerate the calculation of aerodynamic forces. The training and validation process are carried out upon the unsteady aerodynamic data obtained by the proposed CFD method. The well-trained ROM is then coupled with the structure equation at a specific velocity, the Limit-Cycle Oscillation (LCO) of stall flutter under this flow condition is predicted precisely and more quickly. In order to predict both the critical flutter velocity and LCO amplitudes after bifurcation at different velocities, a new ROM with GRU neural network considering the variation of flow velocities is developed. The stall flutter results predicted by ROM agree well with the CFD ones at different velocities. Finally, a brief sensitivity analysis of two structural parameters of ROM is carried out. It infers the potential of the presented modeling method to depict the nonlinearity of dynamic stall and stall flutter phenomenon.