面向天基监视的红外弱小飞行目标识别算法

针对当前红外弱小飞行目标特征不明显、背景干扰大等问题,提出了一种基于深度学习的红外弱小目标识别算法。检测框架以YOLOv4模型为基础,通过使用K-means++算法对训练集的候选框进行聚类处理,在初始大小的选取上放弃随机生成初始点的方式,在样本集里选取某一个样本作为初始中心使锚框（anchor）大小的选取更加合理。在模型结构中引入卷积注意力模块,使算法模型计算资源分配更合理,对红外弱小飞行目标的特征信息更加敏感。改进空间金字塔池化模块,使用平均池化可以更多保留图像的原始信息,降低天基成像中的噪点与坏点的影响。仿真实验表明采用K-means++计算Anchor大小时准确率可以达到80.13%,在加入了SPP和CBAM模块后之后在测试集上算法识别准确率达到了83.3%,经过对模型的修改有效提升了对红外弱小飞行目标识别的准确率。     

基于解析解的小范围高频率机动飞行器制导律设计

针对机动飞行器的小范围高频率侧向机动飞行问题,结合飞行器的运动学和动力学方程,通过解析方法给出了机动幅值和飞行器最大可用过载下的最大机动频率,同时利用粒子群优化方法进行了制导律最优频率的优化。基于解析规划算法给出了正弦机动制导律的解析形式以实现飞行器在侧向方向的机动导引飞行。仿真结果表明：解析算法能够精确得出制导律频率,而粒子群优化方法的精度也能很好的满足要求,误差在5%以内。正弦机动制导律在大速度下可以较好的完成任务目标;在小速度下,幅值误差会有小幅度的增加。     

基于稀疏阵列和码分信号的机载预警雷达STAP研究

基于稀疏阵列和码分正交信号,研究了机载雷达的空时自适应处理（STAP）技术,用于空中预警背景下的地面杂波抑制和运动目标探测。提出了稀疏阵列码分多相位中心孔径综合方法,采用正交编码信号实现多发多收,使综合后不同编码信号的相位中心在数量和分布情况上和满阵天线的相同,从而避免了稀疏阵列天线旁瓣较高的问题;在孔径综合的基础上,利用空时自适应处理方法完成杂波抑制,实现运动目标检测。仿真结果表明了本文方法的有效性。     

复杂结构模压模具的加工方法

分析了模压模具结构形状上的特点和其空间角度及尺寸加工上的难点,由于该模具结构形状复杂,加工精度要求高,其空间角度和尺寸在普通设备加工保证较为困难。重点阐明了利用现有的普通加工设备,采取相应的工艺方案来保证模具工件的空间角度及尺寸的加工方法。     

空间环境模拟器热沉热辐射当量吸收特性研究

文章采用蒙特卡罗法模拟射线的传输过程,分别对两种热沉形式的热辐射当量吸收系数进行了数值计算,研究了表面形式、几何高度及入射光线方向等参数对表面热辐射当量吸收系数的影响。计算表明,正六边形凹腔形式的吸收效果远好于正四棱锥形式,但这种结构形式的吸收有明显的方向性,正四棱锥形式对光线的吸收呈现各向同性。     

太空防御问题研究

空间战略对抗中的太空防御问题越来越引起人们的关注。主要介绍了美国太空防御的发展动态,分析了太空防御的技术特点。对太空防御技术的发展提出一些思考。     

KM6水平舱综合复压系统设计与仿真分析

综合复压系统是载人地面试验设备的重要分系统之一,主要用于载人试验一旦出现重大故障和事故时,对试验舱进行紧急复压,让舱体迅速恢复到正常大气压力环境。文章介绍了KM6水平舱综合复压系统的流程设计;对系统的紧急复压能力进行了数值计算和仿真分析,其结果与系统调试结果基本吻合。表明仿真模型的建立及分析正确可信,综合复压系统流程设计能够满足设计要求。     

Comparative assessment of human–Mars-mission technologies and architectures

We compare a variety of mission scenarios to assess the strengths and weaknesses of options for Mars exploration. The mission design space is modeled along two dimensions: trajectory architectures and propulsion system technologies. We examine direct, semi-direct, stop-over, semi-cycler, and cycler architectures, and we include electric propulsion, nuclear thermal rockets, methane and oxygen production on Mars, Mars water excavation, aerocapture, and reusable propulsion systems in our technology assessment. The mission sensitivity to crew size, vehicle masses, and crew travel time is also examined. Many different combinations of technologies and architectures are applied to the same Mars mission to determine which combinations provide the greatest potential reduction in the injected mass to LEO. We approximate the technology readiness level of a mission to rank development risk, but omit development cost and time calculations in our assessment. It is found that Earth–Mars semi-cyclers and cyclers require the least injected mass to LEO of any architecture and that the discovery of accessible water on Mars has the most dramatic effect on the evolution of Mars exploration.     

Autonomous navigation of formation flying spacecrafts in deep space exploration and communication by hybrid navigation utilizing neural network filter

Autonomous navigation of spacecrafts is a difficult task, however, which is a must in future deep space exploration. With multiple spacecrafts flying in space, this aim can be achieved by formation flying spacecraft (FFS) utilizing inverse time difference of arrival (ITDOA) and inverse difference Doppler (IDD) methods, which can locate the position of earth-station from one-way uplink signals in the FFS coordinate, and by way of conversion of coordinates, the position of FFS is achieved in earth-centered earth-fixed (ECEF) coordinate. The ability of neural network (NN) filter in navigation to extract position of spacecrafts from random measuring noise of signal arrival time and Doppler shift is studied with different radius of FFS and surveying parameters. The NN filter used by spacecraft group is new way of unidirectional autonomous navigation and is of high precision of hybrid navigation.     

Piloted operations at a near-Earth object (NEO)

In late 2006, NASA's Constellation Program sponsored a study to examine the feasibility of sending a piloted Orion spacecraft to a near-Earth object. NEOs are asteroids or comets that have perihelion distances less than or equal to 1.3 astronomical units, and can have orbits that cross that of the Earth. Therefore, the most suitable targets for the Orion Crew Exploration Vehicle (CEV) are those NEOs in heliocentric orbits similar to Earth's (i.e. low inclination and low eccentricity). One of the significant advantages of this type of mission is that it strengthens and validates the foundational infrastructure of the United States Space Exploration Policy and is highly complementary to NASA's planned lunar sortie and outpost missions circa 2020. A human expedition to a NEO would not only underline the broad utility of the Orion CEV and Ares launch systems, but would also be the first human expedition to an interplanetary body beyond the Earth–Moon system. These deep space operations will present unique challenges not present in lunar missions for the onboard crew, spacecraft systems, and mission control team. Executing several piloted NEO missions will enable NASA to gain crucial deep space operational experience, which will be necessary prerequisites for the eventual human missions to Mars.Our NEO team will present and discuss the following:

• new mission trajectories and concepts;

• operational command and control considerations;

• expected science, operational, resource utilization, and impact mitigation returns; and

• continued exploration momentum and future Mars exploration benefits.