The heartbeat is both a physiological signal and a universal expression of emotion. This work presents a pneumatically actuated soft robotic heart for interactive art. A fully 3D-printed, shielded soft capacitive bending sensor is integrated onto the heart's surface, providing stable, real-time displacement feedback without affecting its motion. A lightweight multilayer perceptron (MLP) model estimates deformation from the sensor signal and enables closed loop control. Emotional states are selected via a graphical interface or inferred from ambient light to modulate target heart rates (0.5-3 Hz) for displacement control. Physiologically inspired trajectories derived from left ventricular volume data (Wiggers diagram) are reproduced at 0.5 Hz and 1 Hz. A state-switched proportional-derivative (PD) controller drives the heart to track these targets with good accuracy and robustness. In public exhibitions, the heart was paired with the poetic narrative "To My Love. " The work shows how soft actuation, soft sensing, machine learning, and feedback control can be combined with physiological modeling and artistic expression in a single soft robotic platform.

Specular highlights in images can obscure or completely destroy color and texture details, posing significant challenges to various visual tasks. Traditional highlight removal methods often struggle with complex surfaces or require strict shooting conditions. Deep learning-based methods excel at handling intricate surfaces but still suffer from texture distortion in highlight regions. This paper proposes a novel highlight removal method based on the observation that specular highlights increase brightness and decrease saturation. We generate pseudo-SV (saturation-value) modulated image bases that effectively constructing a discrete color space that closely approximates the brightness, saturation, and hues of highlight-free pixels. We propose a dual-network architecture that jointly trains a highlight detection sub-network and a highlight removal sub-network. By integrating image bases and highlight positional priors, the removal network effectively learns to restore texture and color details through a weighted fusion process. Our results indicate that our approach effectively generates high-quality, highlight-free images, surpassing existing methods in PSNR and SSIM scores. The source code can be accessed at: https://github.com/XufangPANG/Highlight-Removal-based-on-Pesudo-image-bases-fusion .

Robust humanoid stair climbing remains challenging due to geometric discontinuities, sensitivity to step height variations, and perception uncertainty in real-world environments. Existing learning-based locomotion policies often rely on implicit terrain representations or blind proprioceptive feedback, limiting their ability to generalize across varying stair geometries and to anticipate required gait adjustments. This paper proposes an explicit stair geometry conditioning framework for robust humanoid stair climbing. Instead of encoding terrain as high-dimensional latent features, we extract a compact set of interpretable geometric parameters, including step height, step depth, and current yaw angle relative to the robot heading. These explicit stair parameters directly condition a Proximal Policy Optimization (PPO)-based locomotion policy, enabling proactive modulation of swing-foot clearance and stride characteristics according to stair structure. Simulation experiments demonstrate improved generalization across unseen stair heights beyond the training distribution. Real-world experiments on the Unitree G1 humanoid validate reliable indoor and outdoor stair traversal. In challenging outdoor scenarios, the robot successfully ascends 33 consecutive steps without failure, demonstrating robustness and practical deployability.

This paper presents a novel cable-climbing mechanism: the Collaborative Climbing Robot Squad (CCRobot-S), a variant of Reconfigurable Cable-Driven Parallel Robots (R-CDPR), specifically designed for the inspection and maintenance of stay cables. The forward kinematics of the CCRobot-S robotic system, however, is inherently mathematically intractable. This research proposes a novel cascaded strategy with Embodied Artificial Intelligence (EAI) to effectively tackle the forward kinematics problem. In this proposed strategy, a lightweight deep learning-based model integrated with numerical method optimization supplants traditional methods, providing feedback on the poses of the flying platform to the control loop of the CCRobot-S robotic system. It provides an approximate solution as initial values through a deep neural network by learning from physical or simulated interactive experiences of CCRobot-S, and then transfers the suitable initial values with kinematic constraints or physical constraints that are near the real solution to the numerical method. This process achieves a stable and robust solution for the forward kinematics of CCRobot-S. This article includes the foundational kinematic analysis of CCRobot-S, the formulation of the CCRobot-S model, a comprehensive introduction and analysis of the cascaded strategy, including the dataset preparation, the training configuration, the solution inference, and the numerical method optimization. Comprehensive evaluations and experiments were undertaken to examine the proposed strategy. The results reveal and confirm that the deep-learning neural network implemented in the CCRobot-S robotic system is effective. Additionally, the proposed cascaded strategy achieves higher prediction accuracy than the standalone neural network approach under the condition of real-time execution (position error reduced from mm to mm in the X direction, from mm to mm in the Y direction, and from to in the orientation). The cascaded strategy also guarantees convergence in 100 of test cases (50/50) and demonstrates enhanced stability and robustness (1:1 mapping from the joint space to the task space)relative to the conventional Newton-Raphson algorithm's numerical method. These attributes are crucial and necessary for the CCRobot-S system to be effectively deployed in real-world applications.

This study introduces a novel climbing strategy, reconfigurable parallel-type cable-driven climbing designed for long-span, large-scale bridge stay cable robotic applications, which has the potential to revolutionize the stay cable inspection and maintenance practice. The proposed methodology features the development of a Collaborative Climbing Robot Squad (CCRobot-S), which builds upon the design principles of the previous CCRobot series. In this study, CCRobot-S implements a parallel-type cable-driven manipulation design, allowing for reconfigurable kinematic morphology by its movable anchor bases and realizing the capacity of crossing over the stay cables for its flying platform. The collaborative robot squad design liberates the dimensions and scales of the robot's reachable workspace and moves the part of the robotic system that indeed needs to be moved, enhancing the working efficiency and climbing agility. This strategy also utilizes controllable adhesion instead of friction to interact with the bridge cable surface for the flying platform, realizing force multiplication for forceful manipulation. Toward bringing high efficiency and heavy-duty capacity, we propose the applicable climbing frameworks (zero-downtime climbing gait for cable inspection and spider-like climbing gait for cable maintenance) and the optimization frameworks (optimal anchor configuration for the movable anchor bases and optimal grasp arrangement for the flying gripper). This article includes the exploration of the design and climbing gaits of CCRobotS, the formulation of the CCRobot-S model, a comprehensive analysis of its workspace, and its climbing strategy and optimization. Extensive experiments have assessed the proposed climbing strategy's effectiveness and showcased CCRobot-S' capabilities.

The development of embodied AI (EAI) critically depends on efficient data acquisition, yet faces persistent challenges including high costs, limited training scenarios, and lack of standardized datasets. We present AIRSPEED, an open source data production platform designed to address these bottlenecks through three core innovations. First, AIRSPEED achieves hardware–software decoupling via unified robot and simulation interfaces, enabling seamless integration with diverse data collection devices and simulation platforms. Second, it supports comprehensive data production methods spanning teleoperation and teaching approaches, as well as synthetic data generation through data synthesis and virtual teleoperation. Third, AIRSPEED automates pyramid-structured dataset construction compatible with both HDF5 and LeRobot formats, significantly reducing manual overhead. Experimental validation demonstrates substantial efficiency gains, achieving up to 35.6× acceleration in dataset construction and 6.0× overall speedup compared to manual workflows. With end-to-end latency as low as 3 ms and compression throughput exceeding 296 MB/s, AIRSPEED establishes a scalable foundation for EAI data production. AISPEED is open sourced on this website: URL .

Soft arms, characterized by their compliance and adaptability, have gained significant attention in applications ranging from industrial automation to biomedical fields. Modeling these systems presents unique challenges due to their high degrees of freedom, nonlinear behavior, and complex material properties. This review provides a comprehensive overview of three primary modeling approaches: numerical methods, analytical techniques, and data-driven models. Numerical methods, including finite element analysis and multi-body dynamics, offer precise but computationally expensive solutions for simulating soft arm behaviors. Analytical models, rooted in continuum mechanics and simplified assumptions, provide insights into the fundamental principles while balancing computational efficiency. Data-driven approaches, leveraging machine learning and artificial intelligence, open new avenues for adaptive and real-time modeling by bypassing explicit physical formulations. The strengths, limitations, and application scenarios of each approach are systematically analyzed, and future directions for integrating these methodologies are discussed. This review aims to guide researchers in selecting and developing effective modeling strategies for advancing the field of soft robotic arm design and control.

This paper proposes a fusion network based on spatiotemporal BEV feature enhancement, aiming to strengthen the representation capability of spatiotemporal BEV information in global contexts. The network employs a spatial BEV feature enhancement module to deeply integrate individual BEV features with the initial global BEV features obtained from the stacking and aggregation of all BEV features, thereby generating guided BEV features with enhanced representational capabilities. These features are then re-stacked and aggregated to form a refined global BEV feature. Additionally, a temporal BEV feature enhancement module is introduced to fuse historical and current BEV features, expanding the receptive field to address the association challenges posed by rapidly moving targets. Experimental results demonstrated that this enhanced architecture achieves state-of-the-art performance on the Wildtrack and MultiviewX datasets, with MODA reaching 92.3% and 96.5%, and IDF1 scores reaching 95.0% and 84.8%, respectively. The results validated the effectiveness of our method from multiple perspectives.

Inertial parameter identification is crucial for accurate robot control, but existing methods for fixed-base manipulators are insufficient for floating-base systems. To address this, we propose the Decomposed Inertia Identification (DII) framework, which utilizes inertia transfer theory and the Recursive Parameter Null Space Algorithm (RPNA) to decompose base parameters into fixed-base and residual subsets. This approach reduces optimization complexity and enables symbolic parameter identification. Inspired by animal spinning behaviors, we use spinning trajectories to excite leg dynamics, overcoming high-DoFs challenges. The Covariance Matrix Adaptation Evolution Strategy (CMA-ES) optimizes parameters under physical consistency constraints. The method was validated on a Unitree Go1 quadruped robot, achieving 98.1% parameter convergence within 200 iterations during spinning locomotion (0.5–2 rad/s yaw velocity). Updating inertial parameters reduced tracking errors by 63% in body posture control and improved straight-line locomotion accuracy by 89% under payload variations on leg. The DII framework bridges fixed- and floating-base systems, enabling the application of mature fixed-base methodologies to floating-base robots and advancing self-model identification for real-world applications.

Autonomous UAV inspections in substations are often affected by navigation errors and environmental disturbances, leading to deviations from predefined viewpoints and inaccurate image acquisition. To overcome this, this paper proposes a new framework for estimating the relative pose of the viewing camera (LiDAR) mounted on a UAV with respect to target substation equipment, providing geometric guidance for accurate alignment. A standardized equipment library is constructed, containing 3D models and annotated interest point coordinates of typical substation devices. The framework consumes high-resolution scene point clouds captured by a UAV-mounted laser scanner. After ground removal, denoising, and clustering-based segmentation, equipment point clouds are extracted and classified using PointNet. GeoTransformer is then applied to register the segmented point cloud with the corresponding template, yielding the 6-DoF relative pose. Experiments in complex substation environments show strong robustness and accuracy, with an average translation error of 0.086 m and a rotation error of 1.26 ^∘ , providing a reliable basis for UAV pose self-adaptation in inspection tasks.

Automatic 3D shape analysis is heavily influenced by the pose of input 3D models, as the continuous nature of pose space introduces complexities that usually exceed the encoding capacities of standard deep learning frameworks. To tackle this challenge, we present Upright-Net+, an enhancement of our previous model, Upright-Net, specifically developed for estimating upright orientation in 3D point clouds. Our approach is grounded in the design principle that ”form ever follows function,” treating the natural base of an object as a functional structure that stabilizes it in its typical pose, influenced by physical laws and geometric properties. We reformulate the continuous orientation problem into a discrete classification task, focusing on learning the points that constitute the natural base of a 3D model. The upright orientation is determined by aligning the normal orientation of this base towards the mass center. To mitigate over-smoothing in the global feature embeddings from stacked graph convolutional layers, we introduce a Global Positional Encoding Module using Relative Distance Histogram Statistics Embedding (GPE-RDHS), which reduces structural ambiguity and enhances orientation estimation. We also enhanced a weighted residual loss term to penalize false positive predictions, enhancing overall model performance. Our method demonstrates exceptional performance in upright orientation estimation and reveals that the learned orientation-aware features significantly benefit downstream tasks, particularly in classification.

The main cable is the primary load-bearing component of a suspension bridge, continuously exposed to harsh environmental conditions, such as wind and rain, throughout the year. These adverse conditions contribute to varying degrees of degradation and damage to the main cable, necessitating regular inspections to prevent catastrophic failures. Traditional manual inspection methods not only suffer from low efficiency but also pose significant safety risks to personnel. To address these challenges and ensure the safe and effective inspection of suspension bridge main cables, this study introduces a novel cooperative climbing robot, designated as Main Cable Robot Version II (CCRobot-M-II), inspired by the locomotion of the inchworm. The robot employs an alternating opening and closing mechanism of four gripper sets, mimicking the inchworm’s movement to achieve efficient crawling along the suspension bridge handrails. This paper provides a comprehensive analysis of the structural design, key components, and motion mechanisms of CCRobot-M-II. A detailed force analysis of the robot’s crawling process is also presented, followed by the design of the control system and the development of an efficient motion control algorithm. Laboratory experiments demonstrate that the robot achieves a positional error of 0–0.64

Humanoid robots equipped with advanced dexterity and real-time motion mapping capabilities have significant potential in daily and industrial tasks. However, achieving accurate and real-time motion tracking, especially for complex hand and wrist movements, remains a key technical hurdle. This paper addresses the challenge of transferring full-body human motion to humanoid robots for applications requiring precise and dynamic motion replication. We propose a system to capture full-body human motion using an optical motion capture system and map it to the H1 humanoid robot equipped with a wrist, dexterous hand, and fully articulated fingers. The system ensures accurate and real-time motion mapping for comprehensive whole-body tracking. Experiment results demonstrate that the H1 robot can swiftly and precisely imitate human full-body motions, showcasing its potential for applications requiring human-like flexibility and efficiency.

The magnetic flux leakage (MFL) method is widely acknowledged as a highly effective non-destructive evaluation (NDE) technique for detecting local damage in ferromagnetic structures such as steel wire ropes. In this study, a multi-channel MFL sensor module was developed, incorporating a purpose-designed Hall sensor array and magnetic yokes specifically shaped for steel cables. To validate the proposed damage detection method, artificial damages of varying degrees were inflicted on wire rope specimens through experimental testing. The MFL sensor module facilitated the scanning of the damaged specimens and measurement of the corresponding MFL signals. In order to improve the signal-to-noise ratio, a comprehensive set of signal processing steps, including channel equalization and normalization, was implemented. Subsequently, the detected MFL distribution surrounding wire rope defects was transformed into MFL images. These images were then analyzed and processed utilizing an object detection method, specifically employing the YOLOv9 network, which enables accurate identification and localization of defects. Furthermore, a quantitative defect detection method based on image size was introduced, which is effective for quantifying defects using the dimensions of the anchor frame. The experimental results demonstrated the effectiveness of the proposed approach in detecting and quantifying defects in steel cables, which combines deep learning-based analysis of MFL images with the non-destructive inspection of steel cables.

ABSTRACT The significance of climbing robotic systems for cable maintenance is escalating in both academic research and real‐world applications. As these systems are poised for real‐world deployment, it is imperative to develop security and rescue mechanisms that ensure robots' intrinsic safety and robustness in dealing with uncertainty factors. This study presents a novel cable climbing robot designed with a climbing platform, a robotic manipulator integrated with specialized maintenance tools, and a gripper to withstand dynamic loads and impacts from maintenance operations. In addition, we propose the variable‐damping safe‐landing mechanism, the rescue mechanism, and the fusible gripper mechanism to counteract substantial disturbances in worst‐case scenarios. Extensive experiments have been conducted to evaluate the proposed robot and its security and rescue mechanisms. The cable climbing robot has a heavy‐duty capacity of 45 kg and an obstacle‐negotiation ability of 10 mm. It also demonstrated its capabilities in various maintenance tasks, such as cable inspection, grinding, or repair. The variable‐damping safe‐landing mechanism was tested, showing the maximum falling speed can decrease from 1 to 0.1 m/s to promise safety, and the falling time can increase from about 5 to 45 s. Meanwhile, the rescue mechanism successfully retrieved the trapped robot. The results demonstrate the capabilities of the cable climbing robot and the feasibility of using the security and rescue mechanisms for the climbing robotic system, which have implications that the cable climbing robot with security and rescue mechanisms is more reliable and can be deployed in the real world with greater confidence.

This paper presents a demolition system, which is composed of a spherical inspection robot and a demolition robot. For the former, it is designed to get information about the demolition scene. The demolition robot with three-fold arm was developed to perform demolition tasks. The arm can be folded if necessary to pass through narrow space, which improves its flexibility. Besides that, to protect the arm while working, a modified oriented bounding box method was proposed to detect the potential collision between the robot and the obstacles. This method not only offers the advantages of simplicity and feasibility, but also takes into consideration the computational requirements. Besides that, an integrated simulation system is established based on Simulink, which demonstrates the relative motion intuitively. The simulation and experimental results provide evidence of the effectiveness of collision detection.

Although industrial anomaly detection (AD) technology has made significant progress in recent years, generating realistic anomalies and learning priors of normal remain challenging tasks. In this study, we propose an end-to-end industrial anomaly detection method called FractalAD. Training samples are obtained by synthesizing fractal images and patches from normal samples. This fractal anomaly generation method is designed to sample the full morphology of anomalies. Moreover, we designed a backbone knowledge distillation structure to extract prior knowledge contained in normal samples. The differences between a teacher and a student model are converted into anomaly attention using a cosine similarity attention module. The proposed method enables an end-to-end semantic segmentation network to be used for anomaly detection without adding any trainable parameters to the backbone and segmentation head, and has obvious advantages over other methods in training and inference speed.. The results of ablation studies confirmed the effectiveness of fractal anomaly generation and backbone knowledge distillation. The results of performance experiments showed that FractalAD achieved competitive results on the MVTec AD dataset and MVTec 3D-AD dataset compared with other state-of-the-art anomaly detection methods.

The concentric cable-driven manipulator (CCDM) has the characteristics of high dexterity, light weight, and safe movement, making them widely used in confined spaces. However, there are difficulties in the fine control with proper stiffness of CCDMs due to their flexible structures and various configurations. This article proposes an analytical variable-stiffness method for the fine control of CCDMs. First, the stiffness model is established by taking into account key factors, including the middle elastic backbone, cable tensions, configurations, and external loads. Then, the stiffness mesh is generated based on the stiffness model, which visually represents changing trends of its stiffness. Simultaneously, the stiffness of CCDMs can be accurately adjusted by optimizing their configurations and cable tensions. Therefore, the fine control with high or low stiffness of CCDMs can be realized in practical applications. Finally, experiments are conducted to verify the analytical variable-stiffness method of CCDMs. Results indicate that the average error of the stiffness model is 5.42%. It also confirms the effectiveness of the proposed method for achieving analytical variable-stiffness control of CCDMs. Furthermore, the proposed method is also applicable to cable-driven manipulators with similar structures.

Soft actuators are gaining increasing popularity in various fields, including marine engineering and biomedical engineering. However, due to their nonlinear properties and significant material deformation, dynamic modeling of soft actuators for motion behavior is quite a challenge, especially for underwater environments. This article ingeniously combines Euler-Bernoulli theory and Cosserat rod theory to propose an efficient and accurate dynamic model for solving the underwater motion behavior of fiber-reinforced soft actuators. Under the assumption of neglecting external forces, we first utilize Euler-Bernoulli theory to establish a mathematical model describing the relationship between water pressure and the bending deformation of the soft actuator. This model is then employed as a boundary condition when solving the dynamic model. Based on this, we use Cosserat rods theory to depict the dynamic behavior of the soft actuator's motion in an underwater environment. In particular, we analyze the impact of its own gravity, buoyancy, and water damping on the motion of the soft actuator. To validate the proposed dynamic model, we fabricated a novel fiber-reinforced soft actuator and a water-driven control system. Subsequently, we conducted a series of model validation experiments. Experimental results show the model maximum error rate is below 13%, thereby confirming the effectiveness of the model, which can predict the motion behavior of the soft actuator under different water-driven pressures.

Generative Adversarial Networks (GANs), as a form of generative self-supervised learning, have garnered significant attention in anomaly detection. However, the generator's capacity for representation learning is constrained due to its excessive focus on pixel-level details, which hinders its ability to effectively learn abstract semantic representations from label prediction pretext tasks compared to the discriminator. To enhance the generator's representation learning capabilities, we introduce a self-supervised learning framework that integrates generative and discriminative approaches. Our proposed discriminative-generative representation learning method not only rivals the performance of discriminative methods but also offers a significant speed advantage. When applied to one-class anomaly detection tasks, our method surpasses several state-of-the-art models on various benchmark datasets, improving upon the top-performing GAN-based baseline by 6% on CIFAR-10 and 2% on MVTAD. Furthermore, ablation studies reveal that absolute positional information negatively impacts the representational learning ability of generative methods in geometric transformation tasks, offering a valuable guideline for the utilization of positional information.