神经网络修正的速度约束辅助车载SINS定位算法

对于车载全球导航卫星系统(GNSS)/捷联惯性导航系统(SINS)组合导航系统,针对GNSS失效而SINS单独工作时仅使用速度约束辅助SINS其纵向位置误差逐渐发散的问题,提出一种神经网络修正的速度约束辅助车载SINS定位算法。通过径向基函数(RBF)神经网络预测SINS纵向位置误差修正系数,以提高SINS单独工作时的定位精度;此外,提出一种限定记忆指数加权实时估计量测噪声的自适应滤波算法。在人为设置GNSS失效以及真实隧道场景下进行车载试验,结果表明本文算法能够在不停车情况下在线修正SINS纵向位置误差,相比于速度约束与卡尔曼滤波相结合的常规算法,有效地提高了GNSS失效时的车载SINS定位精度。     

Geometric reasoning under uncertainty for map-based localization

Map-based navigation in outdoor terrain lacking man-made structures or other highly distinctive landmarks can produce severe localization problems. This paper presents an approach to navigation which implements high level geometric reasoning and matching strategies based on those used by skilled human navigators. This approach, which is demonstrated on a real example involving imagery of mountainous terrain obtained with a video camera and USGS map data, is designed to avoid many of the pitfalls occurring when an attempt is made to navigate by modeling the environment mathematically. It exploits feature attributes which cannot be easily expressed quantitatively but are central to the successful human navigation process.     

Localization,Exploration, and Navigation Based on Qualitative Angle Information

Abstract

Three of the major problems in building autonomous mobile robots are localization, exploration, and navigation. This paper investigates how well different qualitative methods based on angle information, most of them originally invented for representation of spatial knowledge, are suited for addressing these problems. It combines results from discrete and computational geometry with methods from qualitative spatial reasoning, gaining some new insights on the complexity of robot navigation. It turns out that essentially only with panoramas (special roundviews) the qualitative localization problem can be solved in a satisfactory manner. The exploration problem (qualitative map building), remains difficult for all considered approaches.     

一种弹道修正弹SINS任意滚转角快速粗对准方法

针对弹道修正弹药出炮口后滚转角处于随机状态,捷联惯导系统(SINS)失准角过大时卡尔曼滤波收敛困难的问题,提出在卫星拒止环境下利用神经网络快速估计初始滚转角的改进方法。在炮口处布设少量导航信标,建立反向传播(BP)神经网络拟合初始滚转角与观测量间的非线性映射模型。针对信标辅助下姿态弱可观的问题,引入惯导测量参数作为输入,提高网络估计精度。采用主成分分析法进行特征提取,简化网络结构。仿真结果表明,与基于非线性卡尔曼滤波的对准方法相比,本算法可实现任意滚转角下的快速粗对准;对射角、初始俯仰角误差未在训练范围内以及存在布设误差等场景也进行了测试,与未优化的BP网络相比,对准精度更高,鲁棒性更好。     

航天器自主控制与智能信息处理技术

首先从系统的体系结构出发 ,将自主控制系统划分为分层递阶型、包容型、混合型三种基本形式和多智能代理结构的复杂形式 ,然后分别讨论对一般轨道飞行航天器、多个航天器所组成的空间网络、星球表面探测系统和空间机器人等几类不同对象进行自主控制的特点 ,最后论述自主控制对智能信息处理技术的基本需求并提出有关的关键技术     

高轨非合作目标多视线分布式相对导航方法

针对仅采用单视线难于实现高轨非合作目标的远距离自主相对导航问题,提出基于双空间机器人的多视线分布式相对导航方法。首先,建立基于多视线测量的相对导航模型,并给出空间机器人与非合作目标的相对动力学模型;然后,为降低集中式滤波相对导航的维数,设计分布式滤波的相对导航方法,采用扩展卡尔曼滤波算法实现相对运动估计;最后,建立数学仿真模型,对多视线分布式滤波相对导航算法及可观测性进行仿真校验,仿真结果表明该方法是稳定有效的。     

基于YOLO智能网络的红外弱小多目标检测技术

复杂背景下的红外弱小多目标检测是红外目标检测的难点,现有算法的稳健性难以满足实际弹载应用需求。针对弹载环境下的红外场景图像开展了基于典型深度学习网络模型的目标检测应用研究,提出一种基于YOLO(you only look once)网络的智能目标检测方法,通过高维特征学习表征和推理实现红外弱小多目标检测。采用传统的模板匹配算法和YOLO深度学习算法进行识别性能对比分析,验证了YOLO网络在红外弱小多目标检测方面的良好性能。实验结果表明:YOLO算法的检测概率可达92.2%,平均检测精度为0.844,与传统的模板匹配方法相比,YOLO具有明显的优势。     

双臂自由飞行空间机器人捕捉目标实验研究

本文主要介绍了在地面上模拟的空间微重力环境下双臂自由飞行空间机器人捕捉目标的实验。首先介绍了双臂自由飞行空间机器人地面实验平台，该实验平台由机器人模型、视觉系统、无线通讯系统、网络系统、操作台、运动规划系统、气浮系统七个子模块构成。其次简单介绍了双臂自由飞行空间机器人捕捉目标策略，最后介绍了基于该实验平台的几种实验。     

Ad hoc wireless sensor networks for exploration of Solar-system bodies

In this work, we evaluate the exploration of the Solar system by ad hoc wireless sensor networks (WSN), i.e., networks where all nodes (either moving or stationary) can both provide and relay data. The two aspects of self-organization and localization are the major challenges to achieve a reliable network for a variety of missions. We point out the diversity of environmental and operational constrains that WSN used for space exploration would face.We evaluate two groups of scenarios consisting in static or moving sensing nodes that can be either located on the ground or in the atmosphere of a Solar-system object. These scenarios enable collecting data simultaneously over a large surface or volume.We consider physical and chemical sensing of the atmosphere, surface and soil using such networks. Emerging technologies such as nodes localization techniques are reviewed. Finally, we compare the specific requirements of WSN for space exploration with those of WSN designed for terrestrial applications.     

空间黏附足式爬行机器人的稳定性判据及蠕动步态

面向空间在轨服务任务中的黏附足式爬行机器人应用需求,提出一种通过足端和腹部黏附实现爬行的机器人构型,分析了空间黏附足式爬行机器人的稳定性原理,推导了一种以机器人黏附和脱附力矩平衡为稳定条件的空间黏附足式爬行机器人稳定性判据。在此基础上,分析得出空间黏附足式爬行机器人3+1步态的不稳定性,并规划了一种适用于黏附足式爬行机器人在空间微重力环境下的稳定行走步态,即蠕动步态。最后,通过仿真验证了所提出的空间黏附足式爬行机器人稳定性判据及所规划蠕动步态的有效性。     

跨传感器异步迁移学习的室内单目无人机避障

针对强化学习策略由仿真环境向实际迁移困难的问题,以提高无人机采用无深度信息单目视觉时的行人规避能力为目标,提出一种基于异步深度神经网络结构的跨传感器迁移学习方法。首先,在仿真环境中仅使用虚拟单线激光雷达作为传感器,通过基于确定性策略梯度(DDPG)的深度强化学习方法,训练得到一个稳定的初级避障策略。其次,用单目摄像头和激光雷达同步采集现实环境中的视觉和深度数据集并逐帧绑定,使用上述初级避障策略对现实数据集进行自动标注,进而训练得到无需激光雷达数据的单目视觉避障策略,实现从虚拟激光雷达到现实单目视觉的跨传感器迁移学习。最后,引入YOLO v3-tiny网络与Resnet18网络组成异步深度神经网络结构,有效提高了存在行人场景下的避障性能。     

基于深度学习算法的极化合成孔径雷达通用分类器设计

地物分类是PolSAR(极化合成孔径雷达)的重要应用方向。传统算法需要基于特定数据人工选取特征和设计分类器,而深度学习算法能够自行从海量数据中提取层次化特征。在深度学习算法总结的基础上,结合深度学习和PolSAR大数据,提出了一种高效率、高精度的通用分类器设计方法。使用人工标记的数据训练CNN(深度卷积网络),自动化地进行特征学习和提取,并实现高精度的地物自动分类。在具有不同分辨率的机载和星载PolSAR数据上对通用分类器进行测试,都能快速、准确地分类。研究成果可快速将PolSAR数据转译为更直观的地物分类结果,对海量数据,特别是GF-3卫星PolSAR图像的利用有一定的辅助价值。     

基于深度学习的空间站舱内服务机器人视觉跟踪

为提升舱内跟随服务机器人的任务辅助能力,解决机器人对航天员的视觉跟踪问题,提出了一种基于深度学习和概率模型的人体视觉跟踪算法。利用深度卷积神经网络实现了对穿着多样、姿态任意人体的稳定检测。结合人体检测结果,设计了运动预测概率模型,实现了对指定人员准确、连续的跟踪。算法对包含大多数航天员活动的多个数据集进行了验证。实验结果表明:提出的跟踪算法实现了对穿着多样、姿态任意人体的稳定跟踪,并有效避免了由于穿着相似、遮挡可能造成的误跟踪问题。该算法为空间站舱内跟随服务机器人对航天员的视觉跟踪提供了有效的解决方法。算法基于融合的RGB-D图像,工程上易于构建和实现,也可拓展到其他跟随服务机器人视觉跟踪任务中。     

基于偏振光传感器的导航系统实验测试

导航传感器的研制具有重要的研究意义及实用价值,根据沙蚁(Cataglyphis)的导航定位机理研制了一种新型的偏振光导航传感器,介绍了传感器的结构原理及功能模型,设计了基于偏振光传感器的导航控制系统,利用现有的移动机器人平台对导航系统的性能进行了实验测试。测试结果表明,该传感器无误差累积,在移动载体自主导航系统中有广泛的应用前景。     

采用双目视觉的非合作空间目标相对导航与惯性参数辨识方法

针对面向在轨服务的非合作空间目标测量感知问题,提出了一种基于双目视觉的相对导航与惯性参数辨识方法。利用目标表面的特征点建立几何坐标系,并分别设计了姿态测量和相对导航滤波器,实现了目标姿态、角速度、轨道,质心位置与惯量比的高精度估计。在此基础上,通过黏附卫星与非合作目标形成组合体,利用相对导航算法获得的质心位置和惯量比在黏附前后的变化,实现了目标质量和转动惯量的辨识。数值仿真试验证明了算法的有效性。     

When gestures show us the way: Co-thought gestures selectively facilitate navigation and spatial memory

How does gesturing during route learning relate to subsequent spatial performance? We examined the relationship between gestures produced spontaneously while studying route directions and spatial representations of the navigated environment. Participants studied route directions, then navigated those routes from memory in a virtual environment, and finally had their memory of the environment assessed. We found that, for navigators with low spatial perspective-taking performance on the Spatial Orientation Test, more gesturing from a survey perspective predicted more accurate memory following navigation. Thus, co-thought gestures accompanying route learning relate to performance selectively, depending on the gesturers’ spatial ability and the perspective of their gestures. Survey gestures may help some individuals visualize an overall route that they can retain in memory.     

Representing a stable environment by egocentric updating and invariant representations

To represent a stable environment despite the experience of changes during self movements, one can either develop an invariant allocentric representation, or update the egocentric representations as one moves. Using a disorientation paradigm, three sets of studies investigated these mechanisms in human navigation and scene recognition. Accuracy in the configuration of multiple object localization is impaired by disorientation, an effect not due to artifacts such as memory deterioration over time, intervening physical activities, uncertainty in self position and orientation, etc., suggesting one can locate objects primarily by updating their egocentric positions as she or he moves. Disorientation also impaired the judgment of changes to a scene after viewer movements, suggesting a similar egocentric updating process. On the contrary, representation of the shape of the surroundings is invariant and persists through disorientation. The coexistence of multiple mechanisms may increase the flexibility and robustness of the system.     

皮卫星星载计算机存储模块的容错结构设计

空间单粒子翻转现象对航天飞行器中电子设备特别是静态存储器的可靠性具有重大影响,而 星载计算机的可靠运行在很大程度上依赖于系统中存储模块的可靠设计。就皮卫星星载计算 机存储模块的可靠性及提高其可靠性应采取的技术进行了论述,提出了使用汉明纠错码作为 星载计算机存储模块的容错控制策略,并简要描述了存储模块的硬件结构。该设计具有可靠 性高、兼容性强等特点,使皮卫星满足体积小、重量轻、功耗低等工程应用要求。     

伴飞卫星自主相对导航衰减记忆滤波算法

为实现伴飞卫星的自主相对导航,基于C-W方程提出了一种在伴飞卫星轨道系中建立简化的相对运动方程构成星上轨道递推、根据导航敏感器输出进行量测更新的导航滤波器,并利用衰减记忆滤波算法抑制系统模型误差对相对导航结果的影响。仿真结果表明,该滤波算法可实现伴飞星的自主相对导航。它不仅较C-W方程更简单,而且适于模型不准确时的相对导航,工程实现易。