The integration of precise landing capabilities into UAVs is crucial for enabling autonomous operations, particularly in challenging environments such as the offshore scenarios. This work proposes a heterogeneous perception system that incorporates a multimodal fiducial marker, designed to improve the accuracy and robustness of autonomous landing of UAVs in both daytime and nighttime operations. This work presents ViTAL-TAPE, a visual transformer-based model, that enhance the detection reliability of the landing target and overcomes the changes in the illumination conditions and viewpoint positions, where traditional methods fail. VITAL-TAPE is an end-to-end model that combines multimodal perceptual information, including photometric and radiometric data, to detect landing targets defined by a fiducial marker with 6 degrees-of-freedom. Extensive experiments have proved the ability of VITAL-TAPE to detect fiducial markers with an error of 0.01 m. Moreover, experiments using the RAVEN UAV, designed to endure the challenging weather conditions of offshore scenarios, demonstrated that the autonomous landing technology proposed in this work achieved an accuracy up to 0.1 m. This research also presents the first successful autonomous operation of a UAV in a commercial offshore wind farm with floating foundations installed in the Atlantic Ocean. These experiments showcased the system's accuracy, resilience and robustness, resulting in a precise landing technology that extends mission capabilities of UAVs, enabling autonomous and Beyond Visual Line of Sight offshore operations.

The use of UAV for 3D surface inspection has become an important tool in the field of large-scale structure maintenance. However, the commonly used UAV inspection path planning algorithms for 3D surface suffer from problems such as path quality dependent model accuracy, path inspection efficiency, and low inspection quality. To address these issues, this paper proposes a UAV 3D surface inspection path planning algorithm based on normal vector filtering and integrated viewpoint Evaluation. Generate a safe and effective set of viewpoints through uniform sampling and normal vector viewpoint filtering, and then use a integrated viewpoint evaluation method combined with Monte Carlo tree search to select viewpoints, thereby generating a safe, efficient, and complete UAV surface inspection path. The results of simulation and physical experiments show that the proposed method can effectively solve the path planning problem when using UAV for surface inspection of large three-dimensional structures. Meanwhile, while ensuring the quality of inspection, this method can reduce path redundancy and improve the surface inspection efficiency of UAV.

The recent era of Industry 4.0 extensively utilizes manipulators for various critical automation tasks. Handling a manipulator robot is a complex task. This research employs a 3-R (revolute) manipulator robot to achieve optimal joint trajectories for drawing different alphabets and shapes. The solution to inverse kinematics (IK) is essential; therefore, an artificial neural network (ANN) based model is used. The forward and backward reaching inverse kinematics (FABRIK) technique provides stability, though it is computationally intensive and time-consuming. To design a low-cost hybrid model, ANN combined with FABRIK is proposed to compute IK. Using the proposed model, coordinates for different alphabets and shapes within the confined workspace were calculated. We have developed a 3R (revolute) manipulator that utilizes deep learning (DL) and IK. The ANN automatically obtains specific end-effector coordinates. Due to the non-linear characteristics of IK’s mathematical model, inverse kinematics is a time-consuming procedure, making it difficult to provide a mathematical solution. Our primary goal is to reduce processing costs by using ANN and FABRIK with the IK model to solve IK problems and obtain coordinates for any form or alphabet within a given workspace. To ensure a safe region is selected, this model combines the control barrier function (CBF) with the Lyapunov function. The accuracy of the model exceeds 99.5 percent. We have calculated the mean square error (MSE) as 1.66, the root mean square error (RMSE) as 1.25, and the mean absolute error (MAE) as 0.96 for our model. The error between the model’s predicted and actual coordinates also demonstrates letter coordinates and shapes drawn using a physical 3R manipulator model. As a result, this approach can be used to accurately estimate the angles in complex 3DoF inverse kinematics models.

This paper studies the coverage path planning and path following problems for an underwater biomimetic robotic fish to finish the deep-sea net cage water quality monitoring. Firstly, with a focus on minimizing total path length, repetition rate, and turning occurrences to enhance path coverage efficiency and reduce energy consumption, we propose a novel coverage path planning strategy. This strategy incorporates DQN, a reward function, and a rewrite strategy inspired by RRT*. Secondly, a high-performance path-following method, which takes into account robot performance, is designed to cope with adverse conditions in net cages. Finally, the simulation and field experiments demonstrate significant improvements over existing methods and the effectiveness in practical applications, showcasing its applicability in aquaculture management. The proposed algorithm offers valuable insights into optimizing coverage path planning for underwater robots in practical scenarios.

To deal with the issue of poor visibility caused by water turbidity during the operation of underwater robotics, we propose an attenuation prior formation model-guided enhancement algorithm for turbid underwater images. Specifically, we establish an imaging model suitable for turbid water by studying the influence of water turbidity on light attenuation and transmission. For this model, we first propose a scoring formula that takes into account multiple prior knowledge to estimate the global background light with the help of hierarchical searching technique. Then, we make full use of the advantages of different scale neighborhoods in image restoration, and propose an adaptive multi-scale weighted fusion transmission estimation method to balanc e brightness and contrast. In addition, to correct the color of the images with a natural appearance, a variation of white balance is introduced as post-processing. Extensive experiments on two image datasets show that our algorithm achieves better results than state-of-the-art methods.

Miniature personal aerial vehicles (PAVs) with vertical take-off and landing (VTOL) capabilities offer numerous advantages over existing vehicles, particularly in terms of high maneuverability, manned flight capability, and load-carrying capacity required during rescue missions. However, the detailed research work was reported infrequently. In this paper, a miniature PAV capable of piloted operation in both standing and sitting postures is introduced. This PAV, weighing 45kg and measuring 45cm*87cm*154cm, is designed for easy transport and minimal spatial footprint. It incorporates five vertically arranged micro turbojet engines that enable VTOL capabilities and can carry loads exceeding 100kg while maintaining high maneuverability. Notably, a two-degree-of-freedom nozzle mechanism attached to the engines allows for precise thrust direction adjustments. Building upon this propulsion system and the physical model of the PAV, a cascade proportional-integral-derivative (PID) controller is specifically designed for the PAV to regulate its position and attitude. Additionally, a feed-forward-based proportional-derivative (PD) controller is implemented to enhance the engine’s thrust response. The PAV prototype underwent rigorous testing in various outdoor conditions, ranging from temperatures of -7℃ to 42℃ and wind speeds of 0 to 7.2m/s. The test results substantiate the feasibility of the proposed PAV’s working principles, demonstrating its adaptability to environmental fluctuations. While the focus of this paper lies on the miniature PAV system, its implications extend to broader applications within advanced air mobility research.

Transparent liquid volume estimation is crucial for robot manipulation tasks, such as pouring. However, estimating the volume of transparent liquids is a challenging problem. Most existing methods primarily focus on data collection in the real world, and the sensors are fixed to the robot body for liquid volume estimation. These approaches limit both the timeliness of the research process and the flexibility of perception. In this paper, we present SimLiquid20k, a high-fidelity synthetic dataset for liquid volume estimation, and propose a YOLO-based multi-modal network trained on fully synthetic data for estimating the volume of transparent liquids. Extensive experiments demonstrate that our method can effectively transfer from simulation to the real world. In scenarios involving changes in background, viewpoint, and container variations, our approach achieves an average error of 5% in real-world volume estimation. In addition, our work conducts two application experiments integrate with ChatGPT, showcasing the potential of our method in service robotics. The accompanying video and supplementary materials are available at https://simliquid.github.io/.

Several room disinfection robots using Ultraviolet C (UV-C) light have emerged recently, especially with the COVID-19 outbreak. This systematic review aims to identify the current development status of Ultraviolet Germicidal Irradiation (UVGI) disinfection robots, their limitations, and the technologies they use. An automated search was performed on the Scopus, ACM Digital Library, IEEE Xplore, and SpringerLink platforms for papers up to July 2023, followed by a snowballing search. Were found 96 studies, of which the majority were not concerned with the dose of UVGI applied or did not implement any technique to deliver the appropriate dose; the positioning of the lamps was carried out subjectively; and most of the works did not implement any technique to prevent accidents with UVGI. From the analysis of these studies, it was possible to propose a novel classification for the different types of robots based on their technological and readiness levels. The data shows that despite recent advances, the development of disinfection robots is still early, with many advances to be made.

： The secure gripping of soft grippers remains a difficult challenge to address. To solve this challenge, the Single-axis deformation (SD) soft grippers were designed that were inspired by the blooming behavior of petunias. Four SD finger structures were designed by analyzing the anatomical structure of petunia petals. The stress concentration resulting from the deformation of the fingers with different structures was analyzed using numerical methods under the same deformation load conditions. In order to prove the SD soft grippers have a improving ability of secure gripping, The dexterity, gripping capacity, and radial load-bearing capacity of the SD soft grippers were tested through experiments. The experimental results shown that the opening and closing range of the SD soft gripper is 18mm-144mm, the maximum gripping load is 39.4N, and the radial bending angle is only 2.76 when the radial load is 24N.

The ability to perceive environments supports an important foundation for our self-developed robotic rat to improve kinematic performance and application potential. However, the most existing visual perception of quadruped robots suffers from poor perception accuracy in real-world dynamic environments. To mitigate the problem of erroneous data correlation, which is the main cause of low accuracy, the work presents an approach that combines leg odometry (LO) and IMU measurements with VSLAM to provide robust localization capabilities for small-scale quadruped robots in challenging scenarios by estimating the depth map and removing moving objects in dynamic environments. The method contains a depth estimation network with higher accuracy by combining the attention mechanism in the Transformer with the RAFT-Stereo depth estimation algorithm. Besides, the method combines target identification and segmentation with 3D projection of feature points to remove moving objects in dynamic environments. In addition, LO and IMU data are fused in the modified framework of ORB-SLAM3 to achieve highly accurate localization. The proposed approach is robust against erroneous data correlation due to moving objects and wobbles of quadruped robots. Evaluation results on multiple stages demonstrate that the system performs competitively in dynamic environments, outperforming existing visual perception methods in both public benchmarks and our costumed small-scale robotic rat.

This paper proposes a novel framework for large-scale scene reconstruction based on 3D Gaussian splatting (3DGS) and aims to address the rendering deficiency and scalability challenges faced by existing embodied AI tasks. For tackling the scalability issue, we split the large scene into multiple cells, and the candidate point-cloud and camera views of each cell are correlated through a visibility-based camera selection and a progressive point-cloud extension. To reinforce the rendering quality, three highlighted improvements are made in comparison with vanilla 3DGS, which are a strategy of the ray-Gaussian intersection and the novel Gaussians density control for learning efficiency, an appearance decoupling module based on ConvKAN network to solve uneven lighting conditions in large-scale scenes, and a refined final loss with the color loss, the depth distortion loss, and the normal consistency loss. Finally, the seamless stitching procedure is executed to merge the individual Gaussian radiance field for novel view synthesis across different cells. Evaluation of Mill19, Urban3D, and MatrixCity datasets shows that our method consistently generates more high-fidelity rendering results than state-of-the-art methods of large-scale scene reconstruction. We further validate the generalizability of the proposed approach by rendering on self-collected video clips recorded by a commercial drone.

Marine intervention tasks for retrieving large and heavy objects from vessels at sea is significantly important for maritime security. However, it remains a challenging task due to the lack of effective retrieval methods. Harsh open-sea conditions, unreliable GNSS due to either interruptions or deliberate spoofing by the vessels and the limitations of single robots in localizing and manipulating these objects pose significant challenges for the operations. To address these challenges, this article proposes an autonomous heterogeneous multi-robot system comprising two UAVs and a 6-DOF robot arm mounted on a USV to perform this complex task in a GNSS-denied marine environment. GNSS limitations for UAV localization are addressed using UWB beacons specially installed on the USV. Our approach involves the robot arm initially estimating the object’s position. Then the UAVs fly over the target vessel and search for the object using this information. Once object is located, a custom-designed tethered bar mechanism attached to the UAVs drags the object towards the workspace of the robot arm mounted in the USV. This work also introduces a decentralized framework to facilitate effective information sharing and collaboration among the robots. Field trials in maritime conditions demonstrate that by combining these elements, our system possesses significant potential for real-world maritime security and rescue applications. Videos for the field trails can be found at https://youtu.be/a_JzZf7QqWM

Quadrupedal robots are highly regarded for their superior locomotion capabilities and terrain adaptability, making them competent in a wide range of applications. In autonomous navigation tasks, they are required to track upper-level trajectories to reach designated locations with flexible obstacle avoidance. This is typically achieved by a planner which generates a reference velocity and a controller which accurately tracks the velocity commands. In this article, we propose a learning-based controller for quadrupedal robots that achieves accurate and robust velocity tracking. To bridge the gap between simulation and reality, an analytical actuator model is proposed to simulate physical actuator dynamics. We then train a control policy in simulation using Constrained Reinforcement Learning, where symmetry and smoothness constraints are incorporated into reinforcement learning. The symmetry constraint promotes coordinated locomotion and consistent velocity tracking performance, while the smoothness constraint reduces jerky actions and generates stable velocity performance. The control policy is zero-shot deployed on the Unitree AlienGo. It demonstrates a tracking error of less than 0.084 m/s over the entire velocity range and robust locomotion on natural terrain. We also test our controller by integrating it to a pedestrian tracking framework and prove its capability of trajectory following and long-term reliablitiy.

Quick and effective steps are essential to save patients during life-saving surgery. Literature evaluations show hospitals using robotic chest compression manipulators to improve resuscitation. Emergency cardiac arrest patients are resuscitated via chest compressions. Just squeezing the chest is enough for CPR. Even qualified professionals may struggle to conduct chest compressions at the CPR recommended depth and rate. Automatic chest compressions did not outperform manual compressions in most studies and meta-analyses. The combination of robotics and automation improves patient survival. Gentle actuators and sensors can compression and discusses their potential drawbacks. The device will be handled for portability. By observing a human mannequin, the pressure will be adjusted. The goal is to perform as many high-quality compressions per minute up to 50 mm depth. Additional vibrational testing of air muscles gives better device balancing. The device will be suitable for any venue, not just hospitals. The pressure should be gentle, therefore we'll evaluate it. We will conduct experiments comparing mannequin and human arm readings compress the body. This study analyses soft sensing-actuation systems for interface to achieve the desired result. A solenoid valve regulates airflow while a microprocessor controls compression. Three pneumatically soft-actuated air muscle work together to compress 80-120 times per minute at 260 newton. The solenoid coil's deactivation duration is precisely controlled to 50 milliseconds to create compression depth equivalent to one-third of the mannequin's chest breadth. Research is needed to assure structural strength and test Tab software or a biological counterpart for safety features.

The nonlinear control problem of quadrotor UAVs which perform cooperative transportation of payloads is treated with the use of nonlinear optimal and multi-loop flatness-based control methods. The load is suspended with a link from a cart which is turn in connected through cables with two quadrotors. The aim is to compute the flight path and the control inputs of the quadrotors that will allow to lift the load and move it to any desirable final position. First, the dynamic model of the cable-suspended load is obtained through Euler-Lagrange analysis. Despite underactuation the associated nonlinear optimal control problem is solved, thus allowing to compute the lift forces of the cables that enable the load to move on the vertical plane until it reaches the targeted position. These forces are also applied with opposite sign to the quadrotors’ side through joints at the other end of the cables. Thus, the dynamic model of the quadrotors is updated by including in it additional drag forces which are due to the tension of the cables. The flight paths for the two quadrotors that enables to bring the suspended load to its final position are also computed. Next, for each quadrotor the nonlinear control and path following problem is solved, taking into account the cable-induced drag forces effects. To this end, a flatness-based control approach which is implemented in successive loops is applied to each quadrotor. The state-space model of each quadrotor UAV is separated into subsystems, which are connected between them in cascading loops. Each one of these subsystems can be viewed independently as a differentially flat system and control about it can be performed with inversion of its dynamics as in the case of input-output linearized flat systems. The state variables of the second subsystem become virtual control inputs for the first subsystem. In turn, exogenous control inputs are applied to the second subsystem. The whole control method is implemented in two successive loops and its global stability properties are also proven through Lyapunov stability analysis. The whole procedure is repeated at each sampling instance, that is (i) solution of the nonlinear optimal control problem for the transportation of the payload (ii) computation of the drag forces which are exerted on the UAVs due to lifting the load, (iii) solution of the multi-loop flatness-based control problem for the individual UAVs. This control method allows each quadrotor to follow precisely the defined flight path and finally achieves to bring the load to the targeted position.

not-yet-known not-yet-known not-yet-known unknown Public datasets play a crucial role in advancing autonomous robotics research. The rapid evolution of sensors and applications continually drives the need for new datasets. For instance, the shift from mechanical LiDAR (Light Detection and Ranging) sensors on autonomous vehicles to (hybrid) solid-state LiDAR technologies like MEMS (Micro-electromechanical systems) LiDAR has brought about enhanced durability and reduced costs. However, datasets supporting research on these sensors are scarce. This paper presents the multi-modular HK-MEMS dataset, incorporating data from LiDARs, a camera, GNSS, and Inertial Navigation Systems. Notably, it is the first dataset to offer automotive-grade MEMS LiDAR data for research in Simultaneous Localization and Mapping (SLAM). Compared with existing datasets, our data emphasize extreme environments like degenerate urban tunnels and dynamic scenarios, aiming to enhance the robustness of SLAM systems. The data are collected on various platforms including a handheld device, a mobile robot, and notably, buses with real driving behaviors. We collect 187 minutes and 75.4 kilometers of data. State-of-the-art SLAM methods are evaluated on this benchmark. The result highlights the challenges in extreme environments and underscores the ongoing need to enhance the robustness of SLAM systems. This dataset serves as a valuable platform for exploring the potential and limitations of MEMS LiDAR, and a challenge to enhance the robustness of SLAM in urban navigation scenarios. The data is available at https://github.com/RuanJY/HK_MEMS_Dataset.

A document by Xiran Li. Click on the document to view its contents.

Planetary exploration is rapidly gaining importance within the space research community. Autonomous locomotion of rovers requires consideration of several mobility aspects to ensure safety, including avoiding hazardous areas that can cause the robot to become immobilized in soft soil or damaged in sharp terrains. Furthermore, when executing autonomous guidance, selecting an appropriate path to follow is crucial to reduce energy consumption and improve the overall distance traveled by the rover. This directly impacts the rover’s performance and the possible scientific outcome of the mission. This paper addresses the optimization of the autonomous locomotion of Mars rovers by acting on the guidance and control layers. Firstly, an enhanced velocity-based traction controller is proposed, permitting omnidirectional motion while simultaneously addressing slip and kinematic incompatibilities. The controller acts directly at the wheel command level to further improve traction and tracking performances, reducing position and heading errors. The performance metrics evaluated within the traction controller are then used to dynamically update the cost map of the environment. Finally, a higher-level path planner is integrated considering kino-dynamic constraints, continuously providing new paths according to the map updates. The proposed framework has been validated through simulation and real-world experiments on the MaRTA rover of ESA’s Planetary Robotics Laboratory. The results demonstrated that the proposed controller achieves better traction and tracking performance, further improved by the dynamic cost map updates.