基于改进SSD算法的航拍目标检测算法研究

研究基于深度学习技术的无人机航拍图像目标检测算法，首先介绍目标检测算法SSD（Single Shot MultiBox Detector），并对其特征提取网络进行改进，采用稠密特征提取网络替换原网络的主干特征提取网络，提高算法的特征提取能力，从而提升了算法的检测精度。针对网络实时性问题，在算法中引入分组卷积，极大地减少了网络参数量，提升了网络推理速度。为解决训练中出现的正负样本不均衡问题，利用焦点损失（Focal Loss）改进了原算法的损失函数，进一步提升了网络的收敛速度和精度。最后，通过仿真验证了改进算法在目标检测精度上的优越性。     

功能型复合材料在深空探测任务中的应用研究进展

分析了深空航天器面临的复杂空间环境以及对航天用功能型复合材料的需求,系统综述了耐高低温、抗宇宙射线、电磁屏蔽等深空探测环境用功能型复合材料的研究进展,最后,展望了功能型复合材料在深空探测领域中的潜在应用。     

基于改进式YOLOv2的红外图像目标识别方法

针对智能化导引头信息处理算法设计验证需求，本文对多类型目标、海陆空背景辐射特性数据进一步挖掘，提出引入深度学习算法应用于外场红外数据二次分析挖掘的方法。选择采集得到的包括车、船、无人机(UAVs)、客机等类型的红外数据，针对红外图像与可见光图像不同的成像特性，利用改进式YOLOv2红外目标检测与分类算法，融合多级特征，增加一个重组特征图，实现对红外弱小目标智能化快速分类研究，对提取的海量视频库中的目标红外辐射特征进行分类整理。     

Space law education in China: Status quo,shortcomings and possible ways out

Space law education has an essential role in the space capacity building of a country because of its importance for human resources development in space law. Comparisons between Canada and China, the European Union and China and the United States of America (USA) and China in space law degree education, course, teaching methodology, institute and practice and communication platform are provided in an attempt to review some issues which exist within the Chinese space law educational system and to seek a better choice for China’s space law education reform. Chinese space law education has achieved noticeable progress over the past two decades. However, it is still limited in comparison to its counterparts in Europe and Northern America. The range of the space law degree education programmes and space-law-related courses is comparatively limited. A gap remains between space law education and practice in China. Chinese space law classes are still dominated by the teacher-centred teaching methodology, although several universities have some new teaching practices. The establishment of seven space law-related research institutes plays an increasingly important role in Chinese space law education, space law and policy research and legal advisory services for the Chinese government and non-governmental sectors. However, these institutes still have a long way to go in comparison with their counterparts in Europe and Northern America. While China has established domestic research platforms for space law communication through international cooperation, it does not provide space law practice platforms for space law education. While international space law does not provide a direct legal foundation for space law education in China, Chinese educational laws can lay a profound legal foundation for Chinese space law education. To promote its national expertise and capacity in space law, based on the three-pillar model and the experiences of its counterparts in North America and Europe, China should adopt a systematic and sustainable regime for its space law education.     

Improving multi-target cooperative tracking guidance for UAV swarms using multi-agent reinforcement learning

Multi-Target Tracking Guidance(MTTG) in unknown environments has great potential values in applications for Unmanned Aerial Vehicle(UAV) swarms. Although Multi-Agent Deep Reinforcement Learning(MADRL) is a promising technique for learning cooperation, most of the existing methods cannot scale well to decentralized UAV swarms due to their computational complexity or global information requirement. This paper proposes a decentralized MADRL method using the maximum reciprocal reward to learn cooper...     

集群多机器人系统建模研究的发展与展望

凭借高效、鲁棒、应用广泛的特性,集群多机器人系统已经成为当今研究的热点课题之一,具有重要的实用价值.首先简述了自顶而下和自底而上两种多机器人系统研究思路的当前研究概况.然后从拟生物集群系统模型引入,进而引出一致性系统模型,针对低阶、高阶、异质、时延等一致性模型进行分析总结,单独阐述了多智能体强化学习系统模型的研究情况,...     

Transfer learning: A new aerodynamic force identification network based on adaptive EMD and soft thresholding in hypersonic wind tunnel

The aerodynamic test in the pulse combustion wind tunnel is very important for the design, evaluation and optimization of aerodynamic characteristics of the hypersonic aircraft. The test accuracy even affects the success or failure of hypersonic aircraft development. In the aerodynamic test of pulse combustion wind tunnel, the aerodynamic signal is disturbed by the inertial force signal, which seriously affects the test accuracy of aerodynamic force. Aiming at the above problems, this paper innovatively proposes an aerodynamic intelligent identification method, that is the transfer learning network based on adaptive Empirical Modal Decomposition (EMD) and Soft Thresholding (TLN-AE&ST). Compared with the existing aerodynamic intelligent identification model based on deep learning technology, this study introduces the transfer learning idea into the aerodynamic intelligent identification model for the first time. The TLN-AE&ST effectively alleviates the problem of scarcity of training samples for intelligent models due to the high cost of wind tunnel tests, and provides a new idea for further implementation of deep learning technology in the field of wind tunnel aerodynamic testing. And this study designed residual attention block with soft threshold and dense block with adaptive EMD in TLN-AE&ST model. Residual attention block with soft threshold module can more effectively suppress the influence of instrument noise signal on model training effect. Dense block with adaptive EMD makes the deep learning model no longer a black box to a certain extent, and has certain physical significance. Finally, a series of wind tunnel tests were carried out in the Φ = 2.4 m pulse combustion wind tunnel of China Aerodynamic Research and Development Center to verify the effectiveness of TLN-AE&ST.     

基于深度强化学习的复杂地形适应机器人设计与实验

针对行星表面轻量化自主探测任务，基于仿生思想设计了一种仿海胆结构的十二足球形机器人，其具备自主改变构型以贴合复杂地形的能力，可实现无倾覆、高容错的全向运动；基于数据驱动方法，对该机器人设计了一种数据高效的无模型强化学习运动策略，可实现无先验知识的从0到1步态训练以及步态的实物样机快速部署。通过在平面地形和非结构化地形中对其进行仿真实验，验证了经过训练的机器人具备自主运动、适应非结构地形等能力；通过与常用基准策略进行对比，证实了本文提出的运动策略具有训练高效、鲁棒性好的优势；最后通过开发原理样机，开展实物实验验证了仿真环境中所生成的步态在真实物理环境中的动力学可行性。     

Reinforcement learning based dynamic distributed routing scheme for mega LEO satellite networks

Recently, mega Low Earth Orbit (LEO) Satellite Network (LSN) systems have gained more and more attention due to low latency, broadband communications and global coverage for ground users. One of the primary challenges for LSN systems with inter-satellite links is the routing strategy calculation and maintenance, due to LSN constellation scale and dynamic network topology feature. In order to seek an efficient routing strategy, a Q-learning-based dynamic distributed Routing scheme for LSNs (QRLSN) is proposed in this paper. To achieve low end-to-end delay and low network traffic overhead load in LSNs, QRLSN adopts a multi-objective optimization method to find the optimal next hop for forwarding data packets. Experimental results demonstrate that the proposed scheme can effectively discover the initial routing strategy and provide long-term Quality of Service (QoS) optimization during the routing maintenance process. In addition, comparison results demonstrate that QRLSN is superior to the virtual-topology-based shortest path routing algorithm.     

A policy iteration method for improving robot assembly trajectory efficiency

Bolt assembly by robots is a vital and difficult task for replacing astronauts in extra-vehicular activities (EVA), but the trajectory efficiency still needs to be improved during the wrench insertion into hex hole of bolt. In this paper, a policy iteration method based on reinforcement learning (RL) is proposed, by which the problem of trajectory efficiency improvement is constructed as an issue of RL-based objective optimization. Firstly, the projection relation between raw data and state-action space is established, and then a policy iteration initialization method is designed based on the projection to provide the initialization policy for iteration. Policy iteration based on the protective policy is applied to continuously evaluating and optimizing the action-value function of all state-action pairs till the convergence is obtained. To verify the feasibility and effectiveness of the proposed method, a noncontact demonstration experiment with human supervision is performed. Experimental results show that the initialization policy and the generated policy can be obtained by the policy iteration method in a limited number of demonstrations. A comparison between the experiments with two different assembly tolerances shows that the convergent generated policy possesses higher trajectory efficiency than the conservative one. In addition, this method can ensure safety during the training process and improve utilization efficiency of demonstration data.