Clear and high-quality underwater images are essential for marine applications, including autonomous navigation, ecological monitoring, and infrastructure inspection. However, underwater images typically suffer from severe colour distortion, low contrast, and diminished structural visibility due to wavelength-dependent attenuation, scattering, and uneven illumination conditions. Recent deep learning-based underwater image enhancement (UIE) methods primarily adopt end-to-end frameworks, directly regressing enhanced images from degraded inputs. While these approaches have achieved significant progress, they often lack explicit modeling of the degradation process, leading to limited interpretability and suboptimal recovery of fine-grained details. To address these limitations, we propose DRM-Net, an explicit residual learning framework for UIE. Rather than estimating the enhanced image directly, DRM-Net first predicts a pixel-wise Degradation Residual Map (DRM) in the perceptually uniform CIELab colour space. This map explicitly quantifies local colour, contrast, and structural degradations, thereby enabling the network to precisely reconstruct missing visual information. Furthermore, we design a lightweight Subaquatic Multi-Scale Context Fusion module, which utilizes parallel atrous convolutions with softmax-weighted feature aggregation, significantly enhancing robustness against spatially heterogeneous scattering. Trained jointly with pixel-wise DRM and VGG-based perceptual losses, DRM-Net achieves superior colour fidelity, perceptual realism, and structural detail recovery. Comprehensive experiments conducted on multiple benchmarks demonstrate that our proposed approach attains competitive quantitative results and superior qualitative visual performance compared to state-of-the-art UIE methods, while maintaining low computational overhead, making it particularly suitable for resource-constrained underwater robotic systems.

High-speed on/off valves (HSVs) are recognized for their rapid response and high reliability, and are widely employed as pilot elements in proportional valves. However, their inherent switching behavior intensifies the nonlinear characteristics of the control system, thereby limiting fluid delivery precision. In addition, the main spool of the proportional valve is subjected to significantly time-varying load forces. To address these challenges, a nonlinear flow model of the HSV is established, and the dynamics of pressure pulsation propagation are systematically analyzed. Meanwhile, a dynamic model of the proportional valve main spool is developed, explicitly incorporating the effects of time-varying load forces. Based on these models, a load-adaptive sliding mode control (LASMC) strategy is proposed to improve the motion control performance of a high-speed on/off valve–piloted proportional valve (HSVPPV). Motion control experiments conducted on the HSVPPV and its valve-controlled system validated the proposed approach. Compared to conventional PI control, LASMC reduced the main spool’s maximum displacement error from 0.696 mm to 0.312 mm (a 55.2% reduction) and standard deviation from 0.213 mm to 0.127 mm (a 40.4% reduction). At the actuator level, the cylinder’s maximum error decreased from 14.996 mm to 11.527 mm (a 23.1% reduction) and standard deviation from 11.347 mm to 10.323 mm (a 9.0% reduction). These results demonstrate that the proposed controller significantly enhances both valve positioning accuracy and overall system tracking stability.

Underwater images typically suffer from severe colour distortions, low visibility, and reduced structural clarity due to complex optical effects such as scattering and absorption, which greatly degrade their visual quality and limit the performance of downstream visual perception tasks. Existing enhancement methods often struggle to adaptively handle diverse degradation conditions and fail to leverage underwater-specific physical priors effectively. In this paper, we propose a degradation-aware conditional diffusion model to enhance underwater images adaptively and robustly. Given a degraded underwater image as input, we first predict its degradation level using a lightweight dual-stream convolutional network, generating a continuous degradation score as semantic guidance. Based on this score, we introduce a novel conditional diffusion-based restoration network with a Swin UNet backbone, enabling adaptive noise scheduling and hierarchical feature refinement. To incorporate underwater-specific physical priors, we further propose a degradation-guided adaptive feature fusion module and a hybrid loss function that combines perceptual consistency, histogram matching, and feature-level contrast. Comprehensive experiments on benchmark datasets demonstrate that our method effectively restores underwater images with superior colour fidelity, perceptual quality, and structural details. Compared with SOTA approaches, our framework achieves significant improvements in both quantitative metrics and qualitative visual assessments.

Performing unstructured grasping tasks in cluttered or obstacle-rich food processing environments is a key challenge in robotic systems. This work presents a task-adaptive grasping approach for a dual-arm anthropomorphic robot, named Herdsman, developed for the food industry. With an articulated torso, Herdsman is able to perform human-like reaching motions for more flexible grasping operations. To recognize the target object and extract the features for grasping, a vision pipeline, including a lightweight network GDC-YOLO for real-time object detection and a U-ReSENet network for grasping detection enhancement, is designed based on convolutional neural networks. After the detection comes the grasp execution, where a task-adaptive grasping strategy that works with the articulated torso is put forward to carry out grasping tasks in unstructured environments. Comparative experiments are designed to evaluate the detection performance between the proposed network and other popular networks for object detection and grasping detection. In addition, the task-adaptive grasp strategy for the Herdsman robot is experimentally validated by grasping the objects at different heights. The results have shown that the task-adaptive grasping solution exhibits robustness against variations in the target object position, which could be a promising approach for its application in unstructured environments requiring autonomous grasping.

Anex vivoplatform that can simulate physiological vascular environments while maintaining experimental accessibility is essential for developing effective magnetically guided microswarm delivery systems and transitioning them from a laboratory concept to a practical therapeutic tool. In this article, we present a novel platform that uses decellularized spinach leaves as engineered vascular networks to manipulate magnetic microswarm. Through quantitative characterization, including structural integrity analysis, optical transparency measurements, and perfusion studies, we validate the platform’s suitability for microswarm manipulation. Integrating a customized permanent magnetic control system and machine learning-based visual tracking, we demonstrate four key maneuvers for targeted therapeutic applications within the proposed platform, including forward motion, backward motion, selective branch navigation, and localized aggregation at targeted sites. This physiologically relevant yet accessible platform establishes a crucial bridge betweenin vitromodels andin vivosystems. Specifically, it enables the systematic development, quantitative evaluation, and control strategy implementation of magnetically guided microswarm, potentially accelerating the development of magnetic microswarm-based targeted therapy systems.

In minimally invasive surgery, pneumatic soft actuators have gained significant attention due to their promising applications. Valves, which are critical components in pneumatic driving systems, typically operate in only two modes in current research. In multi-actuator systems, this limitation results in a large number of valves and a bulky driving system, which restricts the system’s portability. In this study, we report a four-mode, multi-stable, energy-efficient electromagnetic mini valve. A coil serves as the driving source. The valve has three outlets, enabling it to control up to three actuators independently. The moving components include a cylindrical magnet and a ring magnet. They are concentrically arranged and can only move along the axial direction. These magnets can realize four distinct positions based on the electromagnetic force and their mutual magnetic force. Moreover, the multi-stable design reduces energy consumption. The two cores make the magnets stable when no current is applied. The valve exhibits a shortest response time of 4.75 ms and an energy cost of 0.007 J. It can handle a maximum flow rate of 9.7 L/min at a pressure of 200 kPa, with a maximum back pressure of 100 kPa. Its energy-efficient and four-operating-mode characteristics prove its potential in pneumatic soft actuator systems. A modular design by connecting multiple mini valves reduces the number of pumps and valves, thereby lowering the overall size and weight of pneumatic driving systems.

Increasing customization has driven manufacturers to develop more flexible manufacturing systems. In these systems, different models of the same part are able to share the same production line. For parts that need multiple operations, different models are combined in some operations and separated in others. To achieve this, it is crucial to accurately send every part to its next operation site. Tag-based methods have been commonly used to sort parts, but they cannot be used when the tags may be damaged in certain operations. In these situations, vision-based methods are preferable. Traditional machine vision methods require manual feature definition and may not be suitable in complex situations. Therefore, in this article, we propose a convolutional neural networks (CNNs) based visual sorting system. To support this, a cloud-edge computing environment is developed for fast computation and continuous service maintenance and upgradation. A CNN-based element segmentation method is proposed for accurate part model classification. The prototype system shows that the proposed method can provide high classification accuracy within an acceptable time.

With the paradigm shift from hospital-centric healthcare to home-centric healthcare in Healthcare 4.0, healthcare robotics has become one of the fastest growing fields of robotics. The combination of robot capabilities with human intelligence, for example, telerobotics for home care, is gradually showing promising potentials. In this paper, the Home-TeleBot system, a generalized IoT-enabled telerobotic architecture designed to support home-centric healthcare system, is proposed. In particular, the implementation of it is realized by integrating human-motion-capture subsystem with robot-control subsystem. The dual-arm cooperative robot, YuMi, imitates human motion captured by a set of wearable inertial motion capture devices to complete tasks. The proposed approach using workspace mapping and path planning of robot manipulators, facilitates telerobot to execute tasks in a natural and human-like way. Based on the constant of proportionality calculated by comparing the human original workspace with the robot original workspace, the workspace mapping is achieved by making assumptions of the distance between end-effectors (human hands, robot's grippers) and shoulders. Additionally, robot manipulators' path is planned by setting virtual obstacles to constrain robot motion, which aims to improve the performance of robot's human-like motion. As a specific example of application, we apply the proposed architecture to a fetching task based on dual-arm motion capture and mapping for telerobotics in home care.

Evolution of smart sensing technologies provide an increasingly utilization for IoT-enabled healthcare. In the context of the aging population, the demand for elderly-assistant robots is increasing. At the same time, more and more attention has been paid to the more intuitive way of remote human-robot interaction. In this article, we present the design, implementation, and evaluation of a telerobotic-assisted healthcare system with the ability to achieve the remote elderly assistant and healthcare application. In this work, a remote operation interface using wearable inertial motion capture suit is proposed to control the YuMi robot remotely. The motion capture subsystem and the robot control subsystem are all based on robot operation system (ROS) to carry out the distributed design and integration. The robot arm is controlled by the position and orientation data of the operator's hand acquired by the motion capture suit. Furthermore, the robot's gripper is controlled by the finger bending signal acquired by a data glove. The achievement and performance of the introduced system was verified by experiments.

For the primary mirror of a large-scale telescope, an electro-hydraulic position control system (EHPCS) is used in the primary mirror support system. The EHPCS helps the telescope improve imaging quality and requires a micron-level position control capability with a high convergence rate, high tracking accuracy, and stability over a wide mirror cell rotation region. In addition, the EHPCS parameters vary across different working conditions, thus rendering the system nonlinear. In this paper, we propose a robust closed-loop design for the position control system in a primary hydraulic support system. The control system is synthesized based on quantitative feedback theory. The parameter bounds are defined by system modeling and identified using the frequency response method. The proposed controller design achieves robust stability and a reference tracking performance by loop shaping in the frequency domain. Experiment results are included from the test rig for the primary mirror support system, showing the effectiveness of the proposed control design.

This paper presents an efficient method to adaptively determine the locations for taking measurements for automated calibration of electromagnetic systems using reconstructed magnetic fields. This coupled measurement- computation method solves the Laplace's equation with measured boundary conditions. Along with the formulation of two selection criteria (chord-height and data-spacing), an adaptive scanning algorithm has been developed, which bases four local measurements to determine the next measurement point. This adaptive method, which relaxes the assumption of approximately known structure, has been illustrated (with experimental verification) with three practical applications; electromagnetic velocity probe, electromagnetic flow-meter (EMF) and spherical motor. Comparisons against published data demonstrate that the adaptive measurement algorithm greatly reduces the number of measurements and shortens the scanning route length of all three applications without sacrificing the accuracy of the computed results. Dry calibration results of an EMF were also experimentally verified against test data obtained from a standard flow-rig confirming that a relative error of 0.2% can be achieved. This finding makes the cost-effective dry calibration a practical alternative to the conventional flow rig calibration. As demonstrated on a spherical motor, the flexibility to include a least-square curve fit offers a practical means to filter measurement noise.