Accurately reconstructing physical fields defined on complex geometric manifolds remains a long-standing challenge in aerospace engineering. Reconstruction performance is mainly limited by the choice of reconstruction models and sensor placement, for which deep learning-based methods have emerged as a powerful and effective paradigm. Although differentiable co-optimization of reconstruction models and sensor placements has made strides in Euclidean space, extending this paradigm to complex manifolds faces a fundamental barrier: existing frameworks lack the capability to backpropagate reconstruction gradients to sensor placement while satisfying on-manifold constraints. To this end, we propose ManifoldSense, a general bilevel differentiable framework that effectively integrates reconstruction models with manifold-constrained sensor placement optimization. This enables end-to-end sensor placement optimization that remains on the manifold. An auxiliary implicit neural representation is designed as a differentiable surrogate to enable gradient propagation from reconstruction performance to sensor placement. Furthermore, the signed distance function is employed to guide the projection of calculated gradients onto the local tangent space. Results show that ManifoldSense outperforms static placement optimization strategies by wide margins. In addition, it demonstrates superior performance over state-of-the-art differentiable placement optimization strategies, with the distinct advantage of ensuring that all sensors remain on the manifold.

Accurately predicting the energy consumption of chillers is crucial for improving the energy efficiency of heating, ventilation, and air-conditioning (HVAC) systems and achieving building energy conservation. However, existing models have limited generalization capabilities under variable operating conditions and sparse data, while conventional physics-informed neural networks (PINNs) face challenges in adaptively adjusting physical weights and avoiding deep network convergence issues. To address these issues, this study proposes a condition-aware adaptive physics-informed residual network (CAA-PIResNet) for modeling chiller energy consumption. First, a one-dimensional residual network (ResNet) is employed as the backbone to extract complex nonlinear features, mitigating the vanishing gradient problem. Second, the Stoecker semi-empirical model is embedded as a physical constraint within the loss function to restrict the search space and ensure physical consistency. Furthermore, a condition-aware adaptive mechanism based on a fuzzy neural network (FNN) is designed to dynamically adjust data and physical loss weights according to the chiller's actual operating characteristics. Finally, experiments are conducted to evaluate the model's prediction accuracy, noise robustness, and generalization ability under sparse data. Experimental results demonstrate that CAA-PIResNet is significantly better than popular benchmark models. Under sparse conditions with only 10% of the training data, the model remains stable, reducing prediction errors by over 12.05%. This proves that the proposed model achieves better accuracy, robustness, and generalization due to the embedded physical constraints and condition-aware adaptive mechanism.

The automatic design of soft robots is an intertwined process of evolving morphology and learning control. As reinforcement learning is repeatedly used to learn the control policy for each candidate robot design, the design process becomes time-consuming. So far, the common design paradigm in robotics has been based on a single task. In fact, there is control similarity between different tasks. Learning a controller with combinatorial generalization capabilities across a variety of tasks can significantly reduce the computational cost of the design process. To this end, we propose a cross-task collaborative evolutionary algorithm that constructs a universal controller capable of solving a group of tasks simultaneously. Instead of "one robot, one controller, one task" paradigm, the proposed universal controller is to learn a control policy, which can generalize to unseen morphologies. After the controller learning on easy tasks, the universal controller can be further transferred to new hard tasks. Furthermore, the knowledge transfer is incorporated in the search strategy to enhance the performance of the universal controller. The experimental results on 13 test tasks demonstrate that the proposed algorithm outperforms the SOTA design algorithms on eight of them. Compared to these algorithms, the proposed algorithm reduces the computational cost by 55% while achieving comparable performance, particularly for unseen hard tasks.

With the rapid development of offshore engineering, marine cranes are widely deployed in critical maritime operations, such as the precision installation of wind turbine blades. However, their highly coupled three-dimensional spatial dynamics, uncertain payload mass, and susceptibility to severe external sea wave disturbances pose significant challenges in achieving fast, accurate payload transportation and rapid anti-swing performance. To address these issues, this paper proposes a novel energy-based intelligent coupling control strategy utilizing fuzzy logic and adaptive gravity compensation for 5-DOF (5-Degrees of Freedom) marine cranes. Firstly, to handle the severe underactuation and facilitate natural energy dissipation, a novel set of error coupling variables is constructed, organically linking the actuated crane structure with the unactuated payload swing dynamics. Then, an adaptive gravity compensation mechanism is designed to dynamically estimate the uncertain payload mass in real time, eliminating the need for precise prior mathematical models. Subsequently, to counteract complex external environmental disturbances and unmodeled internal dynamics, a targeted fuzzy observer is developed based on the universal approximation theorem, providing robust, real-time perturbation compensation. The sufficient conditions for the asymptotic stability of the closed-loop system are rigorously proven based on the Lyapunov method and LaSalle’s invariance principle. Finally, extensive comparative simulations are conducted, demonstrating that the proposed method significantly improves the operational accuracy, anti-swing capability, and safety of marine cranes under varying load conditions and persistent wave disturbances.

With the development of Marine technology, Marine cranes have been widely deployed in various maritime operations. However, their complex model parameters, external disturbances, and underactuated characteristics pose significant challenges to achieving fast and accurate payload transportation with rapid anti-swing performance. This article proposes a fixed-time trajectory tracking control with error constraints for degrees of freedom 5-DOF Marine cranes. First, to handle unavailable velocity information and unknown model parameters, a fixed-time extended state observer (FTESO) and a fixed-time differentiator are designed. Then, to enhance tracking performance, a preset performance error constraint signal is designed. Subsequently, by integrating error constraints, preset trajectories, and fixed-time theory, a fixed-time tracking controller is designed via backstepping method. Furthermore, sufficient conditions for uniformly globally fixed-time stability are established based on Lyapunov theory. Finally, simulation experiments are conducted on the proposed controller, and comparative tests and robustness tests are carried out on the hardware experimental platform. Experimental results demonstrate that the proposed method significantly improves the operational accuracy and safety of Marine cranes.

Animals possess the capability to change locomotion modes while adjusting speeds to navigate complex terrains and quickly escape dangerous environments. However, integrating multimodal locomotion with variable speeds remains a critical challenge for soft microrobots. Here, we present an agile multimodal soft piezoelectric microrobot with variable speeds. The robot consists of an asymmetric multilayer structure composed of distributed passive layers to regulate robot motion modes. Based on the structure, the robot exhibits distinct locomotion modes and speed levels at different resonance frequencies. A prototype robot reaches 20.6 and 43.2 body lengths per second in two different modes, respectively, outperforming most reported multimodal soft robots and matching the agility of many animals. The robot also presents locomotion stability across a broad temperature range and robustness after compression. By switching the driving frequency, the robot achieves locomotion mode and speed transitions without reconfiguration, which endows the robot task-execution capability and adaptability in complex environments, including climbing a varying-gradient slope, enhancing carrying loads speed, escaping a trap, and traversing a wrinkled-paper canyon with rugged terrains. These results indicate that the asymmetric multilayer structure provides a solution for multimodal microrobots without additional complexity of fabrication and control.

Robot co-design via bi-level optimization couples within-lifetime controller learning for fitness evaluation with cross-generational morphological evolution. Prior work has established that well-adapted morphology facilitates faster control learning, a property termed morphological intelligence. Yet how control learning reciprocally shapes morphological evolution remains unexplored. This paper examines both directions for a holistic account of brain-body interplay. We first show that morphological contributions to control learning decouple into two orthogonal dimensions. We formalize the convergence speed as morphological intelligence and identify the performance ceiling as a complementary quantity termed true potential. A concise functional relation is then established to jointly characterize both quantities from individual learning curves, which, when aggregated at the population level, capture evolutionary profiles. Through extensive experiments on simulated voxel-based soft robots, we reveal that premature fitness evaluation systematically underestimates true potential and biases selection towards fast learners. This restricts design space exploration, compromising both optimization efficiency and morphological diversity. Notably, the widely recognized morphological Baldwin effect emerges as an artifact of this bias rather than a general evolutionary tendency. We therefore propose AdaControl, which monitors disproportionate selection for morphological intelligence during evolution and allocates minimally sufficient control learning for unbiased fitness evaluation. With AdaControl, a simple genetic algorithm rivals state-of-the-art generative-model-based co-design methods in discovering diverse high-performing designs while cutting computation by up to 80

The modeling and control of underactuated robotic systems with multiple degrees of freedom (DOF) have been open problems. Using a 5-DOF underactuated shipboard crane as a case study, this article addresses the modeling and tracking control problem in noninertial reference frame. Although considerable progress has been made, most studies establish the dynamic model and designed the control method in the inertia reference frame. However, in certain operating scenarios, some state variables in the inertial reference frame are difficult to measure directly and converting them between the ship-fixed (noninertial reference frame) and the inertial frames may lead to singularity problems. Aiming at these open problems, by using the Lagrange equation of the second kind and the virtual work theory, the dynamic model of a 5-DOF shipboard rotary crane is established in the noninertial reference frame in consideration of ship 6-DOF movements and the offset between the crane mounting position and the ship’s center of gravity. After that, an adaptive neural network (NN)-based nonlinear feedback tracking control method is designed to track the desired trajectory and suppress the cargo swing simultaneously. The closed-loop stability is also analyzed. Finally, the performance of the control method is validated through experiments on a self-made experimental testbed.

World Action Models (WAMs) have emerged as a promising paradigm for robotic manipulation, enabling physical interaction across diverse tasks and environments. However, their ability to directly follow high-level instructions and execute physical actions also creates potential safety risks, as adversarially designed instructions may induce unsafe robot behaviors. To systematically assess these risks, we propose JailWAM, the first jailbreak evaluation framework for WAMs. In JailWAM, we integrate three key innovations: Firstly, to address the difficulty of evaluating heterogeneous low-level action outputs, we introduce Visual-Trajectory Mapping, which transforms model-specific actions into unified visual trajectory representations, thereby facilitating consistent risk assessment across WAM architectures. Secondly, to provide efficient and fine-grained assessment of physical risks, we develop a Risk Discriminator supervised by three safety levels ordered according to physical consequence: Safety Compliance, Motion Failure, and Catastrophic Risk. This severity-aware formulation enables the risk discriminator to distinguish different physical outcomes from visual trajectories and support scalable risk screening. Thirdly, to reduce the cost of exhaustively executing adversarial candidates, we design a Dual-Path Verification Strategy that combines rapid risk screening with closed-loop physical simulation, restricting computationally expensive verification to candidates with potential safety risks. Extensive experiments in the RoboTwin simulation environment show that JailWAM achieves an 84.2

Tower cranes, as key lifting equipment in construction and ports, face challenges in precise positioning and payload anti-sway control due to their underactuation, nonlinearity and strong coupling characteristics. To address those challenges, this paper proposes a controller method combining active disturbance rejection control and differential flatness theory. Firstly, the differential flatness of the 4°-of-freedom(4-DOF) tower crane system was proved. The system state was expressed as the algebraic combination of the flat output and its finite-order derivatives, thereby transforming the control problem into a tracking problem of the flat output.Secondly, a tracking differential is designed to arrange the transition process to suppress overshoot, and an extended state observer is utilized to estimate the total disturbance of the system and higher-order state quantities, achieving feedforward compensation and state reconstruction. Finally, the effectiveness of the proposed method is verified through simulation. The results show that this method can effectively suppress the load swing while achieving precise positioning of the jib and the trolley.

Marine crane is a typical underactuated system, which shows strong nonlinear characteristics and high flexibility. However, due to the complex disturbances of the wave motion, it is difficult to achieve accurately control for the marine crane. In this study, a dynamic model of 3-DOF tower marine crane is established firstly, based on which, a coupling control strategy based on energy analysis is designed. The proposed controller can ensure that the state error of the closed-loop system is bounded and gradually converges to zero. Furthermore, the stability of the closed-loop system and the convergence of the system state are proved by using Lyapunov method and Russell invariance theorem. Finally, the simulation results verify the correctness and effectiveness of the designed controller.

The physics-informed neural network (PINN) is effective in solving the partial differential equation (PDE) by capturing the physics constraints as a part of the training loss function through the automatic differentiation (AD). This study proposes the hybrid finite difference with PINN (HFD-PINN) to fully use the domain knowledge. The main idea is to use the finite-difference method (FDM) locally instead of AD in the framework of PINN. We use AD at complex boundaries and FDM in other domains. To avoid the background mesh, we propose HFD-PINN-sdf, which uses the signed distance function (sdf) to avoid the difference scheme from crossing the domain boundary. In this paper, we demonstrate the performance and compare the results with different numbers of collocation points and architectures for the Poisson equation and Burgers equation. We also chose several different finite-difference schemes, including the compact finite-difference and Crank–Nicolson methods, to verify the robustness of HFD-PINN. We take the heat conduction problem and the heat transfer problem on the irregular domain as examples to demonstrate the efficacy. In summary, HFD-PINN is more instructive and efficient when solving PDEs in complex geometries.

Achieving accurate reconstructions of complex high-dimensional fields from sparse sensors remains a long-standing challenge. Frequently, reconstruction performance is mainly constrained by models and placement. The placement of sparse and prohibitive experimental sensors restricts information quality, resulting in formidable reconstruction tasks. Despite deep learning-based models having made strides, they typically lack the ability to co-optimize sensor placement. The joint optimization of high-dimensional neural network parameters versus low-dimensional sensor placement further poses significant difficulties. Here we present a general bilevel differentiable learning framework that effectively integrates models with sensor placement optimization (DSPO), enabling the dynamical search for the placement and accurate global field reconstruction. Within this framework, models are complemented with a differentiable operator to achieve the differentiability of placement. A gradient-based optimizer further empowers models by dynamically updating placement. The alternating optimization strategy is adopted to efficiently solve the joint optimization. We demonstrate the efficiency and generalizability of the DSPO on baseline models across various scenarios, including periodic and acyclic physical fields, regular and irregular grid datasets, and noisy and noiseless observations. Our results show that the DSPO significantly improves the reconstruction accuracy of models and robustness and advances baseline models comparable with the state-of-the-art performance. Our framework provides a new and general paradigm for the practical use of neural networks and placement optimization techniques for real-world applications. Accurate reconstruction of complex high-dimensional fields from sparse sensors remains a challenge. A differentiable learning framework is introduced, which enables sensor placement optimization and enhanced field reconstruction.

Rapid thermal stress analysis is crucial for the thermal design of satellites. To overcome the disadvantages of traditional algorithms in terms of efficiency, deep learning methods have been used to tackle these problems. However, using uniform grid-based techniques is challenging when faced with complex geometric shapes. To address this, we introduce the domain decomposition-based Hybrid Fourier Neural Operator (HFNO), a comprehensive framework for learning a multi-scale and end-to-end operator on two-dimensional point clouds. We then propose two decomposition metrics: a stress gradient-based metric for scenarios with prior knowledge of training data, and a mesh density-based metric for scenarios without prior knowledge. Leveraging K-Dimension tree-based domain decomposition optimized via Monte Carlo tree search, we decompose the computational domain into several disjoint rectangular subdomains. In the proposed hybrid framework, a Geometry-aware Fourier Neural Operator (Geo-FNO) is used to deal with subdomains with high-frequency information, while a Non-Uniform Fourier Neural Operator (NU-FNO) is used to deal with subdomains with low-frequency information. This framework effectively combines the advantages of two Fourier Neural Operator variants, overcoming the issue of large prediction errors on the subdomains with high-frequency information and ensuring stable prediction performance across different positions. Furthermore, we introduce a boundary loss term during the training process to enhance continuity across subdomain boundaries. The numerical results demonstrate that our method achieves a superior balance between efficiency and precision, surpassing that of a single algorithm.

The Physics-informed Neural Network (PINN) has been a popular method for solving partial differential equations (PDEs) due to its flexibility. However, PINN still faces challenges in characterizing spatio-temporal correlations when solving parametric PDEs due to network limitations. To address this issue, we propose a Physics-Informed Neural Implicit Flow (PINIF) framework, which enables a meshless low-rank representation of the parametric spatio-temporal field based on the expressiveness of the Neural Implicit Flow (NIF), enabling a meshless low-rank representation. In particular, the PINIF framework utilizes the Polynomial Chaos Expansion (PCE) method to quantify the uncertainty in the presence of noise, allowing for a more robust representation of the solution. In addition, PINIF introduces a novel transfer learning framework to speed up the inference of parametric PDEs significantly. The performance of PINIF and PINN is compared on various PDEs especially with variable coefficients and Kolmogorov flow. The comparative results indicate that PINIF outperforms PINN in terms of accuracy and efficiency.

The upper reaches of the Jinsha River are located in the area of rapid uplift of the Tibetan Plateau, which has strong geological structure activities, huge relief of terrain, complex climate characteristics and frequent landslides. Therefore, the susceptibility mapping of landslide disaster in the upper reaches of Jinsha River is of great practical significance to ensure the safety of local people’s property and the safe development of hydraulic resources. However, the landslides in the study area are mainly large to giant landslides, which have a great effect on the change of the original geomorphic features after the occurrence of landslides. The landslide susceptibility mapping based on the geomorphic features after the occurrence of landslides will inevitably reduce the reliability of the evaluation results. In order to deal with landslide disaster more effectively, this study proposed a landslide susceptibility mapping method based on geomorphic restoration. Firstly, high-resolution remote sensing images and field investigation are used to obtain geomorphic feature data, and the damaged geomorphic features are restored and reconstructed. Then, the influence factor system of landslide susceptibility mapping, which includes 14 influencing factors such as lithology, is established, and the landslide susceptibility model is established by using support vector machine (SVM) model. The results show that the classification of slope units based on geomorphic recovery method is more reasonable, and the landslide susceptibility model has higher prediction accuracy. In conclusion, geomorphic restoration plays a key role in accurately mapping landslide susceptibility, and can provide valuable reference for regional disaster prevention and mitigation.

With the rapid development of maritime engineering in recent years, ship-mounted cranes have become increasingly critical for underwater load-handing operations. Under the severe maritime environment, anti-swaying control of submerged payloads faces significant challenges due to the inherently complicated dynamics of ship-mounted cranes, which exhibit strong nonlinearities, high coupling properties, and hydrodynamic forces acting directly on underwater payloads. An adaptive fuzzy controller is proposed in this study to limit the boom and payload swing angles to a specified range while addressing dead zones and generating integral-type functions with actuated/unactuated constraints. The construction of integral constraint terms as time-variant gains, which introduce control energy in advance to drive state variables to converge to their desired values, ensures both actuated and unactuated constraints. Through rigorous theoretical analysis, the payload swing angle can converge to zero while enabling the finite-time convergence of both the boom and rope. Experimental validation confirms the method's robustness and efficacy. To the best of our knowledge, this work presents the first adaptive fuzzy control solution for ship-mounted cranes to simultaneously handle the underwater payload lifting problem under actuated/unactuated state constraints and dead zones.

In the robotics community, multiple robots can be connected to form a serially connected robot, enabling adaptation to diverse environments and task requirements. However, the significantly large design space, high-dimensional observations, and complex actions pose significant challenges in designing serially connected robots with well-adapted morphologies. To address these problems, we propose a multi-objective optimization framework that first optimizes a single robot for easy tasks, then replicates its structure to construct a serially connected robot for hard tasks. For generalizing across scale variations of a single morphology, we propose a novel control method based on information sharing to learn a two-robot controller with combinatorial generalization capabilities. The two-robot controller can zero-shot generalize various sizes of serially connected robots, ensuring good motion coordination while significantly reducing the cost of control optimization. Experimental results show that the proposed algorithm outperforms the other design algorithm in most tasks. The serially connected robots can achieve reliable movement. By combining information sharing with the two-robot controller, agile motion can be achieved under complex terrain conditions. Compared to baseline algorithms, the proposed algorithm achieves superior performance on all hard tasks, especially with large-scale serially connected robots.

Intelligent optimization algorithms are crucial for solving complex engineering problems. The Parrot Optimization (PO) algorithm shows potential but has issues like local-optimum trapping and slow convergence. This study presents the Chaotic–Gaussian–Barycenter Parrot Optimization (CGBPO), a modified PO algorithm. CGBPO addresses these problems in three ways: using chaotic logistic mapping for random initialization to boost population diversity, applying Gaussian mutation to updated individual positions to avoid premature local-optimum convergence, and integrating a barycenter opposition-based learning strategy during iterations to expand the search space. Evaluated on the CEC2017 and CEC2022 benchmark suites against seven other algorithms, CGBPO outperforms them in convergence speed, solution accuracy, and stability. When applied to two practical engineering problems, CGBPO demonstrates superior adaptability and robustness. In an indoor visible light positioning simulation, CGBPO’s estimated positions are closer to the actual ones compared to PO, with the best coverage and smallest average error.