This paper introduces Uncertainty Propagation Network (UPN), a novel family of neural differential equations that naturally incorporate uncertainty quantification into continuous-time modeling. Unlike existing neural ordinary differential equations (neural ODEs) that predict only state trajectories, UPN simultaneously models both state evolution and its associated uncertainty by parameterizing coupled differential equations for mean and covariance dynamics. The architecture is grounded in Gaussian moment closure approximation, which enables efficient analytical uncertainty propagation through nonlinear dynamics without requiring stochastic sampling or ensemble methods. UPN supports two operational modes: pure prediction from initial conditions, and adaptive filtering with sparse measurement updates when observations become available during the prediction horizon. The continuous-depth formulation provides principled uncertainty quantification in a single forward pass, handles irregularly-sampled observations naturally, and adapts evaluation strategy to each input’s complexity. Experimental results demonstrate UPN’s effectiveness across multiple domains: (1) four canonical non-chaotic dynamical systems achieve near-perfect 96.7% confidence interval coverage with single-point Markovian initialization; (2) chaotic Lorenz attractor modeling maintains 94.5% calibration while correctly capturing exponential uncertainty growth in a fully Markovian framework; (3) real-world CubeSat trajectory prediction achieves 89.6% error reduction through integrated measurement updates; and (4) time-series forecasting on the ETTh1 benchmark dataset demonstrates 14% improved accuracy and 6.6× faster inference compared to Neural Stochastic Differential Equations (Neural SDEs). These gains stem from UPN’s analytical distribution evolution, which provides superior computational efficiency and calibration compared to sampling-based approaches.

The vibration suppression for rigid-flexible coupling hoisting robots (RFCHRs) under compound motion has always been a focus in the engineering. Most vibration suppression control methods are overly complex in design, making them hard to effectively apply. Some methods that are relatively easy to deploy often fail to adapt well to changes in system uncertain parameters and external environmental disturbances, raising concerns about their robustness. To address this problem, a robust input shaping vibration suppression control method based on deep reinforcement learning (DRL) is proposed. Based on real-time environmental state feedback, this method utilizes the strategy gradient optimization capability of the proximal policy optimization with clipping (PPO-Clip) algorithm, enabling real-time updates of the shaper parameters according to changes in system parameters and external environmental. It can achieve faster vibration suppression for RFCHR, while also possessing strong adaptability to system parameter changes and anti-interference capability. Finally, experiments verify its effectiveness and superiority.

This article is devoted to data-driven optimal control of linear discrete systems with sensor fault via a performance triggering approach. A quadratic inequality is introduced to equivalently describe systems subject to a fault. A barrier function is provided by the input and output to treat the gain constraint. An optimal control law is built on the adaptive dynamic programming (ADP) method to improve control performance. A dynamic triggering mechanism is constructed by instantaneous data and performance index to balance triggering frequency and control performance, especially under an emergent situation. Sufficient conditions are supplied to ensure the ultimately uniform boundedness of the closed-loop systems. An illustrative example is presented to verify the validity of the proposed strategy.

On-orbit servicing (OOS) has become an essential aspect of modern space missions, encompassing satellite maintenance, orbital assembly, and debris removal. This paper presents a novel decentralized navigation algorithm designed for a swarm of servicing spacecraft to collaboratively reconstruct the shape of unknown tumbling space objects. The proposed method leverages a dynamic factor graph-based Collaborative Simultaneous Localization and Mapping (C-SLAM) framework, integrating observed and identified point cloud features across the swarm. To address the challenges associated with the target’s tumbling dynamics, the approach also incorporates a dynamic SLAM formulation, utilizing a noisy parametric model to propagate the dynamic map and construct the dynamic factor graph at the front-end. Kinematic factors are introduced to account for loop closures, enabling the swarm to recognize previously observed features as they rotate with the target, thereby enhancing mapping robustness. Additionally, the target kinematic model parameters themselves—such as its center of mass, linear velocity, and angular velocity—are estimated in real time from the reconstructed maps. The latter notably is estimated using a singular value decomposition approach that determine the best fit rotation between two consecutive sets of mapped points. These estimates are fed back into the kinematic factors and loop closure processes to refine map optimization iteratively. Simulation results demonstrate that incorporating kinematic factors, addressing loop closures, and facilitating inter-robot communication significantly enhance the swarm’s ability to track the evolving map without a central leader. This decentralized approach highlights the potential of equipping spacecraft swarms with advanced, robust, and scalable perception capabilities for the collaborative inspection and characterization of unknown tumbling space targets.

This article addresses synchronization control of switched complex network with unknown state and actuator fault. A mode-dependent proportional-integral intermediate observer is explored to estimate unknown elements with high-estimation accuracy. A hybrid controller is constructed to treat the asynchronous occurrence of impulses and switching moments. With the help of mode-dependent average dwell time and mode-dependent average impulsive interval, a mode-dependent criterion is established to guarantee the uniformly bounded synchronization performance. Two examples are simulated to deliver the effectiveness of the proposed method.

In this research, we introduce a deep reinforcement learning-based control approach to address the intricate challenge of the robotic pre-grasping phase under microgravity conditions. Leveraging reinforcement learning eliminates the necessity for manual feature design, therefore simplifying the problem and empowering the robot to learn pre-grasping policies through trial and error. Our methodology incorporates an off-policy reinforcement learning framework, employing the soft actor-critic technique to enable the gripper to proficiently approach a free-floating moving object, ensuring optimal pre-grasp success. For effective learning of the pre-grasping approach task, we developed a reward function that offers the agent clear and insightful feedback. Our case study examines a pre-grasping task where a Robotiq 3F gripper is required to navigate towards a free-floating moving target, pursue it, and subsequently position itself at the desired pre-grasp location. We assessed our approach through a series of experiments in both simulated and real-world environments. The source code, along with recordings of real-world robot grasping, is available at Fanuc_Robotiq_Grasp.

In this article, we investigate an improved dynamic guaranteed cost event-triggered-based anti-disturbance control for Takagi–Sugeno fuzzy wind-turbine systems subject to external disturbances. A guaranteed cost event-triggered paradigm with dynamic threshold and sector structure is constructed to alleviate unnecessary triggers caused by outlier measurement. An additional event condition is designed to deal with the difference of premise variable between the system and controller. A PI-type intermediate estimator is introduced to simultaneously estimate the system state and external disturbance. Subsequently, an event-triggered fuzzy controller is built to actively compensate the external disturbances. With the help of Finsler's lemma, sufficient criteria are derived in terms of linear matrix inequalities to make the wind-turbine systems asymptotically stable. Finally, the proposed method is verified by comparative studies.