Robust visual Simultaneous Localization and Mapping (vSLAM) is essential for autonomous navigation, yet a limited field of view (FoV) leads to feature loss during fast rotations, degrading tracking and localization performance. Traditional parallel stereo vSLAM offers limited FoV due to high camera overlap, resembling monocular systems. Existing FoV extension methods, based on wide-angle lenses or multi-camera setups, suffer from distortion or high system complexity. Inspired by navigation-oriented species, which use divergent eye placements to expand their FoV, we propose a divergent stereo vSLAM system that uses two outward-facing cameras with partial overlap, thereby extending the overall FoV. Traditional stereo systems use the left camera for primary vision and the right for depth estimation, limiting the use of the divergent configuration. To overcome this, we propose divergent stereo modeling, treating the cameras as virtual cameras to improve depth estimation and wide-angle perception. Although multi-camera systems can adopt a divergent configuration, existing methods overlook inter-camera correlation. We address this with a cross-camera map point association and a dual-map bundle adjustment (BA) with fixed extrinsics to maintain consistency and reduce redundancy. Experimental results on simulated and real-world datasets show that our framework improves localization accuracy (ATE) by 65.8% over parallel stereo and achieves superior rotational robustness. All code will be open source.

Existing LiDAR SLAM methods typically rely on parameterizable explicit local models-planes, quadric surfaces, or Gaussian distributions-to formulate geometric constraints for state estimation. In unstructured environments, however, structures like tree trunks, weeds and uneven terrains are difficult to be precisely represented by simple parametric models, leading to fewer reliable constraints and introducing modeling errors. To address this issue, we propose a novel implicit point-to-voxel observation model that directly predicts residuals and their uncertainty from local voxel geometries without explicit fitting. The resulting measurements are incorporated into an iterated error-state Kalman filter (IESKF) for pose update. To support efficient online retrieval and update of local geometric context, we further introduce an implicit voxel map representation that partitions the map into voxels and maintains an online-updated implicit representation for each voxel. Experiments show that the proposed method provides more stable effective constraints and improves localization accuracy in unstructured scenes, while maintaining real-time performance on a CPU platform.

When exploring complex unknown environments, unmanned aerial vehicles (UAVs) often experience reduced efficiency and robustness due to unevenly distributed occlusions. This paper proposes an efficient hybrid autonomous exploration algorithm that adapts to environmental complexity, enabling effective frontier detection and viewpoint sampling to minimize overall exploration time. We introduce a frontier detection method based on a limited field of view (FOV), along with an unique ID-based frontier management mechanism, which ensures detection completeness while significantly reducing computational and memory overhead. Furthermore, an adaptive sampling strategy incorporating environmental complexity is introduced. By adaptively switching sampling modes and relaxing obstacle-free sphere generation constraints, the method improves both sampling efficiency and visibility evaluation performance. For path planning, a hierarchical planner based on a topological graph is constructed. It jointly optimizes global coverage paths and local frontier information to generate smooth and time-optimal trajectories. Both simulation and real-world experiments validate the advantages of the proposed approach in terms of exploration efficiency, computational overhead, and coverage rate.

To address high communication overhead and limited robustness in large-scale multi-UAV task allocation under dynamic communication conditions, this work proposes a communication-aware distributed task allocation framework that integrates communication-topology construction with robustness-enhanced distributed allocation. The core objective is to achieve low-overhead and robust distributed task allocation that remains reliable under packet loss, transmission delay, and topology switching. For communication topology construction, a distance-based fitness coefficient is incorporated into the Newman-Watts small-world model to prioritize long-range links that effectively reduce network diameter and average path length while preserving connectivity. For task allocation, an adaptive consensus mechanism is developed by integrating extended bidding information, a locking mechanism, and lightweight robustness enhancement strategies. These designs mitigate inconsistencies caused by packet loss, transmission delays, and topology switching, thereby improving allocation stability under the considered communication disturbances. Experimental results demonstrate that, under the tested conditions, the proposed method maintains a 100% task allocation rate under packet loss of up to 60%, persistent communication delays, and controlled topology switching. In large-scale scenarios, it reduces communication data volume to 3.3%–35.7% of that required by representative distributed task allocation baselines while preserving comparable task completion efficiency. Overall, the proposed framework provides an efficient, scalable, and robust solution for communication-constrained multi-UAV coordination.

ObjectiveTo improve the waterproof reliability, locomotion efficiency and elbow maneuverability of robots for shipboard water-filled pipeline inspection, a novel pipeline inspection robot with high waterproof reliability, low-power propulsion and excellent elbow passability is proposed. MethodA multi-objective optimization framework is established for the shipboard pipeline inspection robot. The envelope dimensions of the robot are optimized with comprehensive consideration of hydrodynamic drag, internal space requirements and layout constraints of sensors and actuators. The Pareto-optimal solutions are solved by the NSGA-II algorithm. A hydrodynamic model of the robot is built, and CFD simulations are performed under typical inflow conditions to analyze the velocity field, vortex structures and pressure coefficient distributions. The influences of the robot configuration on drag and flow separation characteristics are evaluated accordingly. Finally, waterproof tests in water-filled pipelines, 90° elbow passability tests and visual inspection experiments in actual water-supply pipelines are implemented to verify the proposed design. ResultsNumerical simulation results show that the overall drag coefficient of the proposed configuration is 0.45 under typical inflow conditions. Experiments prove that the robot can operate stably in water-filled pipelines and realize smooth passage through a 90° elbow. It is suitable for pipelines with diameters of 100 mm to 400 mm, and its maximum moving speed reaches 0.5 m/s. In the test of actual water-supply pipelines, the robot can transmit real-time first-person-view images, which supports the visual identification of pipeline inner wall features including joints, wear marks and potential cracks. ConclusionThe proposed three-spherical waterproof robot configuration can effectively enhance the propulsion efficiency and turning stability in water-filled pipelines with elbows, while ensuring reliable sensor integration and sealing performance. This design provides a valuable reference for the structural design and engineering application of condition monitoring devices for shipboard pipelines.

To overcome the limitations of traditional homogenization methods that neglect micromorphic effects and data-driven approaches that are sensitive to data quality, this paper proposes a mechanism-data dual-driven hybrid decoupled residual architecture network (HydraNet) for the mechanical analysis of architected lattice structures. The key novelty is the decoupled residual architecture that explicitly embeds micromorphic effects into the learning framework, enabling both physical consistency and efficient solution of high-order governing equations. First, based on a two-phase nonlocal homogenization theory, the governing equations for nonlocal homogenized plates are established with intrinsic parameters to capture micromorphic effects. To avoid the direct evaluation of high-order derivatives, a mixed-variable formulation is adopted to transform the original high-order system into coupled second-order equations, while a sparse dataset generated from computational homogenization is incorporated to guide learning and enable inverse identification of the intrinsic parameters. Results demonstrate that the proposed mechanism-data dual-driven HydraNet achieves high predictive accuracy, with spatially averaged full-field relative errors consistently below 5% across various lattice topologies, which are significantly lower than those obtained using classical homogenization methods. Compared with traditional finite element method (FEM) simulations, the proposed framework accelerates online prediction by approximately 3-4 orders of magnitude, while requiring substantially less training data than purely data-driven surrogate models owing to its embedded physical mechanisms. By integrating nonclassical mechanical mechanisms with FEM data, the mechanism-data dual-driven HydraNet provides a novel paradigm for predicting the mechanical performance of architected lattice structures and offers robust technical support for the design and investigation of such structures.

To address the challenge of achieving both low drag and high maneuverability in complex water-filled pipeline environments, this study proposes a novel dual-spherical pipeline robot with integrated leak detection and mapping capabilities. A multi-objective optimization framework was established to simultaneously improve hydrodynamic performance, motion stability, and internal spatial layout, while adopting a streamlined shell design to achieve both low-drag and sensor integration requirements. Based on a task-driven configuration optimization method, an energy-efficient propeller arrangement was derived under the constraint of maintaining maneuvering performance. The robot employs a helical differential propulsion system and integrates multiple sensors, including a vision module, an inertial navigation unit, and a pressure sensor, to enable leak detection and mapping. Its fully sealed spherical housing ensures stable operation in water-filled pipelines. Based on the proposed configuration, an experimental platform incorporating four representative pipeline environments was constructed, and a series of inspection, mapping, and environmental adaptability tests were conducted. The results demonstrate that the robot can achieve agile turning and stable locomotion in water-filled pipelines, showing strong potential for practical engineering applications.

Aiming at the challenges in planning for the synchronized execution of multiple geometrically heterogeneous tasks by swarms in uncertain environments, this work proposes a distributed task planning method. This method is intended to overcome the inefficiencies in traditional methods caused by fragmented task modeling, high sensitivity to environmental uncertainty, and the risk of collaborative deadlocks. First, a Conditional Value-at-Risk (CVaR) model is introduced to dynamically estimate task execution time under uncertainty. Second, multi-geometry tasks (area, line, point) are uniformly transformed into equivalent point-task sets via grid-based decomposition and key-point discretization. Finally, a cooperative task dependency graph is embedded into a distributed auction consensus mechanism with cycle detection to prevent deadlocks, and an adaptive cooperative obstacle avoidance strategy based on task progress is integrated to ensure synchronized and collision-free motion. Simulation results demonstrate our method's superiority over benchmarks in completion time, success rate, and efficiency. Its practicality is validated through real-world UAV tests, while scalability experiments confirm its robustness in large-scale scenarios. This work provides a solid theoretical foundation and technical support for efficient coordination of robot swarms in complex and dynamic environments.

Ion wind technology holds potential for the miniaturization of flying microrobots, generating thrust without mechanical moving parts, enabling simplified and light designs free from high-frequency actuators. Nevertheless, two fundamental challenges have limited its practical implementation: insufficient payload capacity and inadequate controllability, particularly in multi-degrees-of-freedom flight. Here, we demonstrated an inertial measurement unit-based closed-loop controlled flight of a light (36.7 mg) ion-propelled microrobot, achieving a thrust-to-weight ratio of 5:1 and 1-h long-endurance tethered hovering without mechanical actuators. Experimental validation indicates that the control strategy enhances stability, with a decrease of 83.11% and 89.21% in the root mean square error of pitch and roll angle, respectively. Leveraging origami-inspired design and cost-effective metal-polymer composites, our manufacturing approach enables the rapid assembly of complex microrobots at a disposable cost, overcoming a key barrier to practical swarm deployment. The microrobot's high load capability allows it to carry a high-fidelity image sensor and a fiber Bragg grating sensor while maintaining sufficient maneuverability to complete predefined tasks such as environmental surveillance and material identification. This work introduces an approach to autonomous microrobotic swarm flight, suggesting potential applications in confined-space surveillance, disaster rescue, and hazardous environment exploration.

To address the challenge of simultaneously satisfying task capability requirements and minimizing energy consumption in thruster-configuration optimization, this study proposes a hierarchical optimization framework. First, a task-induced required capability boundary is constructed using a hydrodynamic model, CFD simulations, and prescribed trajectory and velocity requirements to screen feasible thruster configurations. Second, within the feasible set, a mission-level evaluation pipeline integrating thrust allocation and closed-loop control is established to evaluate energy consumption and tracking errors over the task. Finally, an outer-loop meta-heuristic global search optimizes thruster orientation parameters to obtain the minimum-energy configuration. To validate the proposed method, twelve physical configurations (TC1-TC12) were fabricated and tested in a water tank using three representative trajectories (circular, S-shaped, and L-shaped). Experimental results show that the optimal configuration TC1 achieves the lowest energy consumption for all trajectories and the total metric, with a 10.4% reduction compared to the best among remaining configurations. In addition, TC1 exhibits the smallest tracking errors in the primary motion directions while maintaining low yaw error, indicating improved planar force synthesis and attitude stability. These results demonstrate that the proposed framework can identify the optimal thruster configuration within the feasible set, providing a practical reference for amphibious unmanned platform design.

Sparse optical flow provides stable inter-frame correspondence, playing a key role in Visual Odometry (VO) and Visual-Inertial Odometry (VIO). Classical optimization-based methods, such as Lucas-Kanade (LK), perform well under small displacements but are sensitive to large motions and illumination changes. Modern regression-based learning methods, while more robust in complex scenes, are often computationally heavy and lack explicit geometric consistency, making them less suitable for efficient VO/VIO front-ends. To bridge this gap, we propose a hybrid neuro-symbolic framework that combines the strengths of both paradigms. Our method uses a Convolutional Neural Network (CNN) to extract robust feature representations, which is fed into a differentiable LK optimizer to estimate optical flow in an end-to-end trainable manner. Through implicit differentiation, gradients are propagated across the iterative solver, enabling joint optimization of feature extraction and flow estimation. The resulting system integrates seamlessly into existing VO/VIO pipelines and runs in real-time on embedded platforms. Experiments show that our method outperforms conventional optimization-based flow in challenging conditions such as dynamic lighting and low texture, while also achieving higher accuracy and lower latency than purely regression-based alternatives. When deployed in a VIO system, our method demonstrates significant performance improvement, achieving an average error reduction of 42% on challenging datasets while enhancing tracking stability. The code is publicly available.

Artificial lighting has become a common auxiliary equipment for underwater visual tasks. It not only effectively extends visibility into regions where natural light is insufficient or entirely absent, but it also inevitably introduces nonuniform illumination. In underwater environments, visual perception is inherently affected by scattering-induced haze and by wavelength-dependent attenuation, which result in color distortion. The interaction between nonuniform artificial illumination and these inherent degradations further exacerbates the complexity of underwater image formation, thereby significantly increasing the difficulty of underwater image enhancement (UIE). However, most existing UIE methods fail to explicitly account for this coupling effect. To this end, we propose a novel illumination component for the Retinex model, consisting of two parts: the color part and the light part. Based on the new illumination component, we further introduce a nonuniform illumination UIE (NIUIE) method. First, an illumination correction method is proposed to obtain the light part by combining the gamma correction with dual illumination maps. Then, the color part is derived from a color correct method based on nonlinear channel compensation. Finally, the initial illumination component is optimized with a weighted regularization term, which is efficiently solved using the alternating direction method of multipliers (ADMMs). Comprehensive comparisons on five challenging underwater datasets demonstrate that NIUIE outperforms several state-of-the-art UIE algorithms qualitatively and quantitatively. Furthermore, the effectiveness of NIUIE is also validated on other visual tasks, including nonuniform illumination enhancement for aerial images and feature matching in underwater images. Experimental results show that the proposed method effectively addresses the problem of nonuniform illumination and common underwater issues simultaneously, improving details in both high-light and low-light regions.

Underwater and in-air environments exhibit distinct imaging characteristics, which should be carefully considered and effectively exploited for accurate depth estimation. In this work, we analyze the effectiveness of wavelength-dependent attenuation for underwater depth estimation and show that it is helpful but insufficient to perform depth estimation independently. Therefore, we propose a fast underwater monocular depth estimation network that incorporates underwater light absorption difference (ULAD) as supplementary information. Compared with methods that rely solely on RGB input, the proposed approach provides more accurate depth predictions. In our network, RGB and ULAD features are extracted by MobileNetV4 and fused using FusionMamba, followed by decoding and refinement with a micro Vision Transformer. The network is trained on the USOD10K dataset and evaluated on both its test set and the FLSea dataset. Experimental results demonstrate that our method achieves more accurate depth estimation and higher efficiency compared with other lightweight networks. Furthermore, Compared with existing state-of-the-art fast underwater depth estimation methods, our network further reduces the number of parameters by 10% and improves inference speed by 43%.

To address the issue of trajectory tracking under dynamic wave disturbances, this study proposes a nonlinear model predictive control method based on disturbance observer (DO-NMPC). The disturbance observer performs online observation of the generalized disturbance forces encountered by the unmanned platform during its motion, estimating the loads caused by wave disturbances in real-time. This information is used for active compensation in the model predictive control algorithm, enabling the control system to actively suppress wave disturbances related to unmodeled dynamics that cannot be measured. Simulation tests were conducted to compare the performance of the proposed DO-NMPC control strategy with the LQR control strategy in tracking straight-line and rotational trajectories on the water surface under wave disturbances. The simulation results validate the robustness and advantages of the proposed control strategy.

Simultaneous Localization and Mapping (SLAM) datasets are essential for evaluating SLAM algorithms, since selecting an accurate and reliable method is critical to the autonomous operation of unmanned systems. However, existing SLAM datasets focus primarily on scene diversity and lack a systematic evaluation of differing motion patterns-such as variations in speed and abrupt maneuvers. In real-world applications (e.g., field exploration or emergency response), unmanned ground vehicles (UGVs) often encounter these complex motions, which can lead to SLAM failures. To address this shortcoming, we have constructed the M2PT Dataset: a Multi-Motion-Pattern dataset for UGVs operating across varied Terrains. This dataset comprises LiDAR, inertial measurement unit (IMU), and wheel-odometry data recorded in three environments under four distinct motion patterns (low speed, high speed, sharp turns, and collisions), and provides high-precision GPS trajectory ground truth alongside map ground truth acquired by scanning devices. We compare the performance of several SLAM algorithms on this dataset to highlight its challenges and demonstrate its significance. The dataset can be accessed at https://drive.google.com/drive/folders/ 1_yRVQLY7cjDKHoFnMSsYqRr6ojSF6FGy.

High-speed ground robots moving on unstructured terrains generate intense high-frequency vibrations, leading to LiDAR scan distortions in Lidar-inertial odometry (LIO). Accurate and efficient undistortion is extremely challenging due to (1) rapid and non-smooth state changes during intense vibrations and (2) unpredictable IMU noise coupled with a limited IMU sampling frequency. To address this issue, this paper introduces post-undistortion uncertainty. First, we model the undistortion errors caused by linear and angular vibrations and assign post-undistortion uncertainty to each point. We then leverage this uncertainty to guide point-to-map matching, compute uncertainty-aware residuals, and update the odometry states using an iterated Kalman filter. We conduct vibration-platform and mobile-platform experiments on multiple public datasets as well as our own recordings, demonstrating that our method achieves better performance than other methods when LiDAR undergoes intense vibration.

To address the issue of high drag in the planetary wheelset area during the underwater operation of amphibious robots, this study investigates the underwater drag characteristics of planetary wheelsets and introduces a novel foldable streamlined fairing. Initially, the drag characteristics of planetary wheelsets at various fixed angles are systematically analyzed to identify the minimum-drag state. The underlying drag reduction mechanisms are also examined. Based on the identified minimum-drag state, a spherical shear-fork foldable fairing is designed. This fairing remains folded during terrestrial operations, ensuring that it does not interfere with the robot's obstacle-crossing performance. During underwater operations, the fairing unfolds into a specific streamline shape, thereby improving the flow field structure around the planetary wheelset to achieve drag reduction. Furthermore, the surface of the fairing is designed with continuous transverse V-shaped grooves, which reduce frictional drag effectively. This study employs both numerical simulations, based on the Lattice Boltzmann Method (LBM), and towed drag experimental methods to validate the proposed design. Simulation results reveal that the planetary wheelset achieves the lowest drag when fixed at a 0 degrees position. The optimal drag reduction occurs when the width and height of the transverse V-shaped grooves are 0.22 mm, resulting in a drag reduction rate of 23.93%. Experimental results further confirm the effectiveness of the design, demonstrating a maximum drag reduction rate of 25.74%. This study is the first to investigate the underwater drag reduction of planetary wheelsets and proposes novel methods for the drag reduction of amphibious robots.

In order to address the high complexity and low efficiency of amphibious propulsion systems, this paper proposes a novel variable wheel-propeller integrated mechanism for amphibious robots. By adjusting the blade pitch angle, it enables multiple motion modes, including rapid and stable movement on flat ground, obstacle crossing, and omnidirectional movement on water surface. This study establishes a kinematic model for the propeller blades and conducts multi-objective optimization of the structural parameters by considering both the land obstacle-crossing performance and underwater propulsion performance. Based on the optimized structural parameters, a virtual simulation prototype is constructed. Simulation results indicate that when water surface movement, with a driving torque of 3N.m, robot achieves a maximum linear velocity of 1.25m/s and a maximum angular self-rotation velocity of 3.5rad/s. Moreover, varying the blade pitch angle can alter the thrust direction, enabling omnidirectional mobility on water surface. During land movement, with a rotation speed of 60rpm, the highest obstacle-crossing height is 184mm. This wheel-propeller integrated mechanism exhibits robust comprehensive motion performance and environmental adaptability, with convenient motion modes switching.

The sparse optical flow method is a fundamental task in computer vision. However, its reliance on the assumption of constant environmental brightness constrains its applicability in high dynamic range (HDR) scenes. In this study, we propose a novel approach aimed at transcending image color information by learning a feature map that is robust to illumination changes. This feature map is subsequently structured into a feature pyramid and integrated into sparse Lucas-Kanade (LK) optical flow. By adopting this hybrid optical flow method, we circumvent the limitation imposed by the brightness constant assumption. Specifically, we utilize a lightweight network to extract both the feature map and keypoints from the image. Given the challenge of obtaining reliable keypoints for the shallow network, we employ an additional deep network to support the training process. Both networks are trained using unsupervised methods. The proposed lightweight network achieves a remarkable speed of 190 fps on the onboard CPU. To validate our approach, we conduct comparisons of repeatability and matching performance with conventional optical flow methods under dynamic illumination conditions. Furthermore, we demonstrate the effectiveness of our method by integrating it into VINS-Mono, resulting in a significantly reduced translation error of 93% on a public HDR dataset.