Journal of Field Robotics最新文献

This work presents an investigation of robotic technologies' effectiveness in construction activities. Sixty-four highly relevant publications were identified from the database. By systematically reviewing the publications, the secondary data that are of interest to the review theme were retrieved and further evaluated. It is found that robotic technologies for automated construction is a growing field, where the taxonomy of robot was reflected in a diversified manner in the existing studies, ranging from the muscular guy—robotic manipulator—to the dexterous ones—unmanned aerial vehicle, autonomous mobile robot, automated guided vehicle, autonomous construction machinery and quadruped robot. In addition, the existing studies have provided substantial evidence to reveal the robotic technologies' effectiveness against traditional human methods in construction scenarios, and the measures for effectiveness consisted of productivity, precision, and success rate. With the evidence, it seems that the construction sector could benefit from robotic technologies to achieve intelligent workflows. Furthermore, based on the existing knowledge foundation in the current literature, a theoretical framework for future research direction is proposed. The framework envisages the integration of large models with construction robots to address operational inefficiencies, reduce costs, and simplify management.

This article investigates the control problem of unmanned surface vessels with sensor measurement sensitivity under deception attacks, and proposes a novel self-triggered adaptive neural control scheme under the backstepping design framework. To solve the control design problem of unknown time-varying gains caused by deception attacks and measurement sensitivity in kinematic and kinetic channels, the parameter adaptive and neural network technology are involved. In addition, to decrease actuator wear caused by the high-frequency wave and sensor measurement sensitivity and reduce the computational burden caused by continuous monitoring of the triggered condition, a self-triggered mechanism is constructed in the controller–actuator channel. Finally, a self-triggered adaptive neural control solution is proposed, which can guarantee that all signals in the whole closed-loop system are bounded by theoretical analysis. The effectiveness and superiority are verified by numerical simulations.

This paper presents the design, manufacturing and testing of a tail-sitting vertically takeoff and landing fixed-wing hybrid aerial underwater vehicle (HAUV) called Nezha-SeaDart. Nezha-SeaDart can vertically take off and land from ground and water, cruise in the air with lift generated by the wings, seamlessly cross the water–air interface and operate underwater like an autonomous underwater vehicle. Nezha-SeaDart underwent a 10-day field test in China's Thousand Islands Lake of Zhejiang Province, proving its ability to perform full cross-domain missions. This research has the following contributions to the field of HAUV. (i) A working prototype of vertical takeoff and landing tail-sitting HAUV with all basic functions verified and full mission cycle capability demonstrated in a field test. (ii) An HAUV that travels fast both in the air and underwater. (iii) An HAUV capable of autonomous and seamless water exit that does not rely on a dedicated propulsion system. (iv) A method of sizing the vehicle's wing and thrust considering aerial cruises, underwater operations, and seamless water exits.

Despite various proposed control schemes for uncertain bilateral teleoperation systems under time delays, optimally restricting the system's overshoot has remained an overlooked issue in this realm. For this aim, we propose two novel control architectures based on robust L₁ theory, entitled position-based adaptive L₁ controller and transparent adaptive L₁ controller, with the former focusing on position synchronization and the latter concerning system transparency. Since developing L₁-based controllers for nonlinear telerobotic systems encompassing uncertainty and round-trip delays puts significant theoretical challenges forward, the main contribution of this paper lies in advancing L₁ theory within the field of delayed teleoperation control. To formulate the theories, the asymptotic stability of the closed-loop system for each controller is first proved utilizing the Lyapunov method, followed by transformation, along with the L₁ performance criterion, into linear matrix inequalities. Ultimately, the control gains are attained by solving a convex optimization problem. The superiority of the designed controllers over a benchmark transparent controller for teleoperators is demonstrated via simulation. Furthermore, experimental tests on a two-degrees-of-freedom nonlinear telerobotic system validate the efficient performance of the proposed controllers.

The wall-climbing robot is a growing trend for robotized intelligent manufacturing of large and complex components in shipbuilding, petrochemical, and other industries, while several challenges remain to be solved, namely, low payload-to-weight ratio, poor surface adaptability, and ineffective traversal maneuverability, especially on noncontinuous surfaces with internal corners (non-CSIC). This paper designs a high payload-to-weight ratio wheeled wall-climbing robot which can travel non-CSIC effectively with a payload capacity of up to 75 kg, and it can carry a maximum load of 141.5 kg on a vertical wall. By introducing a semi-enclosed magnetic adhesion mechanism, the robot preserves a redundant magnetic adsorption ability despite the occurrence of significant gaps between localized body components and the wall surface. In addition, by ingeniously engineering a passive adaptive module into the robot, both the surface adaptability and crossability are enhanced without increasing the gap between the body and the wall, thereby ensuring the optimization of the adsorption force. Considering the payload capacity and diversity when climbing on vertical walls, inclined walls, ceilings, and internal corner transitions, control equations for internal corner transitions and comprehensive simulations of magnetic adsorption forces are performed using FEA tools. Finally, a functional prototype was developed for rigorous experimental testing, with the results confirming that the robot successfully meets the desired functionality and performance benchmarks.

The current mainstream approaches for plant organ counting are based on convolutional neural networks (CNNs), which have a solid local feature extraction capability. However, CNNs inherently have difficulties for robust global feature extraction due to limited receptive fields. Visual transformer (ViT) provides a new opportunity to complement CNNs' capability, and it can easily model global context. In this context, we propose a deep learning network based on a convolution-free ViT backbone (tea chrysanthemum-visual transformer [TC-ViT]) to achieve the accurate and real-time counting of TCs at their early flowering stage under unstructured environments. First, all cropped fixed-size original image patches are linearly projected into a one-dimensional vector sequence and fed into a progressive multiscale ViT backbone to capture multiple scaled feature sequences. Subsequently, the obtained feature sequences are reshaped into two-dimensional image features and using a multiscale perceptual field module as a regression head to detect the overall scale and density variance. The resulting model was tested on 400 field images in the collected TC test data set, showing that the proposed TC-ViT achieved the mean absolute error and mean square error of 12.32 and 15.06, with the inference speed of 27.36 FPS (512 × 512 image size) under the NVIDIA Tesla V100 GPU environment. It is also shown that light variation had the greatest effect on TC counting, whereas blurring had the least effect. This proposed method enables accurate counting for high-density and occlusion objects in field environments and this perception system could be deployed in a robotic platform for selective harvesting and flower phenotyping.

This paper presents a comprehensive study of dynamics identification-driven diving control for unmanned underwater vehicles (UUVs). Initially, a diving dynamics model of UUVs is established, serving as the foundation for the development of subsequent algorithms. A noise-reduction least squares (NRLS) algorithm is then derived for parameter identification, demonstrating convergence under measurement noise from a probabilistic perspective. A notable feature of this algorithm is its skill in correcting raw data, thereby improving parameter identification accuracy. Based on the identified model, an improved fast terminal sliding mode control (FTSMC) algorithm is introduced for diving control, consistently ensuring rapid convergence under 16 scenarios. Importantly, the proposed diving control algorithm effectively mitigates chattering by incorporating a dedicated filter, adaptively adjusting the switching gain, and substituting saturation function for sign function. Through experimental validation, the NRLS algorithm's advantage over the traditional least squares method becomes evident, with depth errors consistently below 3.5 cm. This indicates that the identified model closely aligns with the actual model, showcasing a commendable fit. Additionally, when compared to the traditional sliding mode controller and the proportional-integral-derivative algorithm, the FTSMC algorithm has superior performance, as indicated by a mean absolute percentage error consistently below 4%.

This paper investigates the differentiable dynamic modeling of mobile manipulators to facilitate efficient motion planning and physical design of actuators, where the actuator design is parameterized by physically meaningful motor geometry parameters. The proposed differentiable modeling comprises two major components. First, the dynamic model of the mobile manipulator is derived, which differs from the state-of-the-art in two aspects: (1) the model parameters, including magnetic flux, link mass, inertia, and center-of-mass, are represented as analytical functions of actuator design parameters; (2) the dynamic coupling between the base and the manipulator is captured. Second, the state and control constraints, such as maximum angular velocity and torque capacity, are established as analytical functions of actuator design parameters. This paper further showcases two typical use cases of the proposed differentiable modeling work: integrated locomotion and manipulation planning; simultaneous actuator design and motion planning. Numerical experiments demonstrate the effectiveness of differentiable modeling. That is, for motion planning, it can effectively reduce computation time as well as result in shorter task completion time and lower energy consumption, compared with an established sequential motion planning approach. Furthermore, actuator design and motion planning can be jointly optimized toward higher performance.

The task of collecting and transporting luggage trolleys in airports, characterized by its complexity within dynamic public environments, presents both an ongoing challenge and a promising opportunity for automated service robots. Previous research has primarily developed on universal platforms with robot arms or focused on handling a single trolley, creating a gap in providing cost-effective and efficient solutions for practical scenarios. In this paper, we propose a low-cost mobile manipulation robot incorporated with an autonomy framework for the collection and transportation of multiple trolleys that can significantly enhance operational efficiency. The method involves a novel design of the mechanical system and a vision-based control strategy. We design a lightweight manipulator and the docking mechanism, optimized for the sequential stacking and transportation of trolleys. On the basis of the Control Lyapunov Function and Control Barrier Function, we propose a vision-based controller with online Quadratic Programming, which improves the docking accuracy. The practical application of our system is demonstrated in real-world scenarios, where it successfully executes the multiple-trolley collection task.

A machine learning-based approach was utilized to develop a device for groundnut bud necrosis virus (GBNV) disease severity detection and estimation in tomato plants (Solanum lycopersicum L.). The study involved inoculating tomato plants with GBNV, monitoring changes in morphological and spectral characteristics, evaluating machine learning algorithms (decision tree [DT] classifier) for analysis and classification of disease severity, and developing and validating a device for disease detection and severity estimation. Spectral data analysis revealed distinct patterns in reflectance, with notable peaks observed in the 680 and 760 nm bands, while reflectance remained low and constant beyond 900 nm. Machine learning techniques, specifically a DT model, were employed to classify disease severity based on spectral data with high accuracy (95.01% training accuracy and 93.65% testing accuracy). The model identified the near-infrared band as highly correlated (correlation coefficient of 0.82) with disease severity. Furthermore, a compact handheld device integrating a spectral sensor, organic light-emitting diode display, and Raspberry Pi 3B was developed for real-time disease severity estimation. The device demonstrated robust performance, accurately predicting disease severity at different growth stages, even in the absence of visible symptoms. Additionally, disease severity percentages obtained via reverse transcription polymerase chain reaction were used to validate the accuracy of the device's estimations. Its responsive nature, with estimated response times ranging from milliseconds to seconds, facilitates timely interventions in agricultural settings. Overall, this interdisciplinary approach, combining spectral analysis, machine learning, and device development, presents a promising solution for efficient disease monitoring and management in agriculture, contributing to enhanced crop health and food security.