新一代导弹信号模拟器的设计及应用

介绍对一种能与导弹自动综合测试（简称综测）与故障检测（简称故检）设备联机，并实现全自动模拟综合测试与故障检测运行的智能导弹信号模拟器（以下简称智能模拟器），提出了一种新一代的战术导弹信号智能模拟器设计方法。该智能模拟器经与××导弹三级维护测试设备联机运行获得成功，并已通过成果鉴定。     

多约束下的导弹螺旋机动制导控制一体化设计

针对导弹在俯冲机动突防飞行过程中攻角及落地弹道倾角受到约束的问题,基于自抗扰控制(ADRC)及反步滑模控制(SMC),提出一种多约束条件下的导弹螺旋机动制导控制一体化设计方法。首先,基于典型的螺旋机动突防弹道,同时考虑纵向平面指定落角约束,分通道进行制导控制一体化数学模型推导。然后,使用反步滑模控制进行制导控制一体化算法设计,通过设计补偿器对反步法的中间控制量进行修正实现对攻角的约束,针对系统的有界不确定性以及未知干扰,采用干扰观测器进行估计与补偿,提高系统的鲁棒性。最终使用Lyapunov理论证明了系统稳定。仿真结果表明,本文方法具有较强的鲁棒性,能够保证飞行器在满足攻角约束的条件下,按照典型螺旋机动弹道对目标进行大落角高精度打击。     

一种无人机自主避障与目标追踪方法

针对无人机自主避障与目标追踪问题，以深度Q网络(DQN)算法为基础，提出一种多经验池深度Q网络(MP DQN)算法，使无人机避障与追踪的成功率和算法的收敛性得到优化。更进一步，赋予无人机环境感知能力，并在奖励函数中设计了方向奖惩函数，提升了无人机对环境的泛化能力以及算法的整体性能。仿真结果表明，相较于DQN和双重DQN(DDQN)算法，MP DQN算法具有更快的收敛速度、更短的追踪路径和更强的环境适应性。     

事件触发机制下多导弹固定时间编队控制

针对领-从弹编队结构中从弹对领弹的固定时间协同跟踪控制问题，同时为了节省通信带宽和弹载计算资源，基于多智能体一致性理论，给出了有向拓扑下基于事件触发机制的固定时间编队控制算法。用微分几何法将导弹运动模型精确线性化，为其设计固定时间编队控制协议，保证具有较大飞行速度的导弹编队在较短的时间内收敛到稳定的队形。在此基础上，引入事件触发机制，克服有向拓扑下Laplacian矩阵的不对称性和事件触发通信引入的误差项对系统稳定性带来的影响；为从弹设计了基于自身状态误差的事件触发函数，当状态误差满足所设定的阈值时，从弹在通信网络中更新并传递自身的采样信息，有效减少了弹载资源的占用率。利用代数图论、矩阵理论和Lyapunov稳定理论证明了编队控制系统的稳定性。数值仿真结果校验了算法的有效性和稳定性分析的正确性，适用于导弹编队的总体设计。     

攻击角度约束下的分布式强化学习制导方法

为提高导弹在攻击角度约束下对目标的打击效能，提出了一种基于深度确定性策略梯度算法的分布式强化学习制导策略。为了最大限度地减小攻击角度误差，设计了一种新的奖励函数，使导弹在满足视场角约束的同时，视线角向期望值收敛。此外，为了增强强化学习模型的泛化能力，提出了一种分布式探索策略，提高了模型训练过程中对环境的探索效率。仿真结果验证了所提出的分布式强化学习制导方法能够在固定攻击角度约束下实现对目标的精准打击。与传统制导律相比，所提制导方法的攻击角度误差更小，收敛速度更快。     

A clustering based method to complete frame of discernment

When the existing information does not contain all categories, the Generalized Evidence Theory (GET) can deal with information fusion. However, the question of how to determine the number of categories through GET is still intriguing. To address this question, a modified k-means clustering, named centers initialized clustering is proposed, filling the gap of identification and complement of the frame of discernment. Based on this clustering method, the number of categories is determined. The initialized centers selected by center density keep the cluster results constant, enhancing the stability of clustering results. Besides, constructing Generalized basic Probability Assignment (GBPA) modules in a conservative way improves the reliability of the results. The mass of empty set in combined GBPAs is the indicator of the number of categories. Experiments on real and artificial data sets are conducted to show the effectiveness.     

基于飞行时间相机的非合作慢旋卫星稠密重建研究

针对非合作慢旋卫星的模型重建问题，提出基于飞行时间(time-of-flight, TOF)相机和同时定位与制图(simultaneous localization and mapping, SLAM)的稠密重建方法。基于预先检测与自适应阈值方法提高旋转提取与描述(oriented fast and rotated brief, ORB)的特征尺度适应性。利用运动度量方法选取关键帧。利用子模型拼接方法加快重建效率。利用仿真环境制作非合作慢旋卫星的数据集。仿真实验结果表明:该方法能够实现长时间稳定地工作，可在3 min内重建出卫星模型的稠密点云，点云密度大于5 000，重建误差小于5 cm。利用机械臂、卫星模型及光学暗室搭建半物理实验系统，表明算法的精度及抗噪声能力基本满足非合作目标感知的任务的需求。     

Periodic acoustic source tracking using propagation delayed measurements

There exist a large class of acoustic sources which have an underlying periodic phenomenon. Unlike the well-studied Bearings-Only Tracking(BOT) of an aperiodic acoustic source,this paper considers the problem of tracking a periodic acoustic source. For periodic acoustic tracking, the signal emission time is known. However, the true measurement reception time is unknown because it is corrupted by noise due to propagation delay. We augment the sensor’s signal reception time onto bearing measuremen...     

拦截大气层内机动目标的深度强化学习制导律

针对大气层内高速机动目标的拦截问题，提出了一种基于双延迟深度确定性策略梯度(TD3)算法的深度强化学习制导律，它直接将交战状态信息映射为拦截弹的指令加速度，是一种端到端、无模型的制导策略。首先，将攻防双方的交战运动学模型描述为适用于深度强化学习算法的马尔科夫决策过程，之后通过合理地设计算法训练所需的交战场景、动作空间、状态空间和网络结构，并引入奖励函数整形和状态随机初始化，构建了完整的深度强化学习制导算法。仿真结果表明：与比例导引和增强比例导引两种方案相比，深度强化学习制导策略在脱靶量更小的同时能够降低对中制导精度的要求；具有良好的鲁棒性和泛化能力，并且计算负担较小,具备在弹载计算机上运行的条件。     

Measurement-driven Gauss-Hermite particle filter with soft spatiotemporal constraints for multi-optical theodolites target tracking

Multi-Optical Theodolite Tracking systems (MOTTs) can stealthily extract the target’s status information from bearings only through non-contact measurement. The constrained MOTTs are partially compatible, yet many existing research works and results are based on the known model, ignoring its discrimination with the target maneuvering behavior pattern. To compensate for these mismatches, this paper develops a Measurement-driven Gauss-Hermite Particle Filter (MGHPF), which elegantly fuses the spatiotemporal constraints and its soft form to perform MOTT missions. Specifically, the target dynamic model and tracking algorithm are based on the target behavior pattern with the adaptive turn rate, fully exploiting the spatial epipolar geometry characteristics for each intersection measurement by a minimax strategy. Then, the center of the feasible area is approximated via the analytic coordinate transformation, and the latent samples are updated via the deterministic Gauss-Hermite integral method with the target’s predictive turn rate. Simultaneously, the effects of truncation correction and compensation feedback from the current measurement and historical estimation data are adaptively incorporated into the PF’s importance distribution to cover the mixture likelihood. Besides, an effective causality-invariant updating rule is provided to estimate the parameters of these soft spatiotemporal constrained MOTTs with convergence guarantees. Simulated and measured results show good agreement; compared with the state-of-the-art Multi-Model Rao-Blackwell Particle Filter (MMRBPF), the proposed MGHPF improves the filtering accuracy by 7.4%-34.7% and significantly reduces the computational load.