Journal of Field Robotics最新文献

This paper presents a dual Fuzzy Logic Controller system for robot navigation in dynamic and unknown environments. The proposed system integrates two core Fuzzy Logic Controllers:one for goal-seeking and another for obstacle avoidance. To ensure smooth transitions between both behaviors, a third fuzzy fusion block is introduced, inferring a continuous weighting parameter K based on real-time sensor data. Additionally, the obstacle avoidance behavior is reinforced with a dynamic direction prioritization mechanism that selects the safest and most efficient turn based on obstacle layout. To overcome local minima and isolated regions, a subgoal generation module is embedded, enabling the robot to dynamically reposition itself toward free space without predefined waypoints. The overall architecture is presented through a structured introduction comprising background, literature survey, and a clear statement of contributions. Extensive validation is carried out through both simulation in ROS2-Gazebo and real-world experiments using a Pioneer 3-DX platform. Comparative analyses demonstrate that the proposed improvements significantly enhance navigation efficiency, path smoothness, and adaptability in cluttered or evolving environments, while preserving low computational cost and interpretability.

The development of selective harvesting robotics in agriculture and horticulture has gained a lot of attention and momentum in recent years, but commercial products are still lacking. This paper develops a sweet pepper harvesting robot and focuses on sweet pepper peduncle detection, the design of the end-effector, and the motion planning of the manipulator to improve the harvesting success rate. To address the low recognition accuracy of sweet pepper peduncles, a detection algorithm designed explicitly for sweet pepper peduncles is proposed by storing the observation position in advance, which involves dividing the recognition process into two steps. The recognition success rate of picking points is 94.4%. To address the problem of low efficiency and low success rate of harvesting robots, a two-stage motion planning (TSMP) method is proposed to divide the whole picking process into offline and online planning. Offline planning adopts a database to plan trajectories to pre-stored observation points to ensure that the manipulator identification of fruit peduncles at close range is with solutions and the trajectory is reliable, while online planning proposes an adaptive end-effector grasping pose control algorithm and analyzes and calculates the best grasping pose for the manipulator based on the sweet pepper peduncles pose. A novel sweet pepper end-effector is designed and integrated into a fully automatic harvesting robot system. Finally, the proposed method is experimentally verified on the developed sweet pepper harvesting robot. The field experiment results demonstrate that the robot can continuously harvest sweet peppers with a harvesting success rate of 77.16%. The average picking time is about 15 s with a fruit recovery device. Supplementary video is available at https://www.youtube.com/watch?v=vnlRTkjPD2U.

To address the high labor intensity and low efficiency associated with traditional manual lotus seedpod harvesting, this study proposes the design of an under-actuated, three-finger linkage grasping mechanism that mimics the grasping motion of the human hand. The design is based on the physical properties of the lotus seedpod, ensuring a compact structure and ease of control. Kinematic and static models of the grasping mechanism were developed, and the range of structural parameters was determined using custom-developed software. Additionally, genetic algorithms were employed to optimize these parameters. The Denavit-Hartenberg (D-H) parameter method was used to analyze the working space of the grasping mechanism, and virtual motion simulations confirmed the rationality and accuracy of the mechanism design. A prototype of the lotus seedpod harvesting grasping mechanism was developed, and a harvesting test platform was constructed to conduct simulated harvesting experiments in a laboratory environment. The experimental results indicate that the grasping mechanism requires an average of 5.34 s to harvest a lotus seedpod, with an enveloping success rate of 95% and a cutting success rate of 80%. These results demonstrate the promising application potential of the proposed grasping mechanism in lotus seedpod harvesting. This study provides a theoretical foundation and practical roadmap for the mechanization and automation of harvesting lotus seedpods and similar fruits and vegetables.

Serial robotic manipulators find applications across diverse sectors such as industries, military, medical, space exploration, and underwater operations, where precision, efficiency, and safety are paramount. However, achieving effective control performance necessitates the effective resolution of the inverse kinematics (IK) problem inherent to manipulators. This challenge involves determining the requisite joint configurations to attain a desired endpoint position and orientation within the manipulator's workspace. The complexity arises due to the nonlinear equations and geometric relationships between cartesian space and joint space. Although various methods have been employed to address the IK problem in serial manipulators, there is a noticeable scarcity of publications comprehensively reviewing these techniques. Therefore, this paper undertakes an extensive review, thoroughly examining diverse IK methods applied in serial manipulators. This investigation of IK compiles literature spanning the last decade (2014–2023). Articles were collected from several web-based repositories such as WoS, IEEE Xplore, ACM Digital Library, Google Scholar, and ScienceDirect, and a preferred reporting items for systematic reviews and meta-analyses (PRISMA) protocol was adopted to acquire appropriate literature. The protocol involves examining the literature methodologies, major findings, and qualitative analysis. Thus, nearly 321 articles were screened and utilized in this review. Further, the IK methods are segregated into model-driven methods (analytical, geometric, algebraic, numerical, and other model-driven methods) and model-free/data-driven methods (metaheuristic (particle swarm optimization, genetic algorithm, and other metaheuristic methods), machine learning (artificial neural network, neuro-fuzzy, and reinforcement learning), and other data-driven methods). This paper offers a detailed exploration of limitations, challenges, performance comparisons, and future research directions of IK techniques. By conducting this study, we aim to assist prospective stakeholders in expanding their understanding of the remarkable potential of IK methods for serial manipulators.

With the rapid advancement of intelligent technologies, the application of robots in agriculture has expanded significantly. Path planning, a critical technology for the autonomous navigation of agricultural robots, has emerged as a key research direction. This paper classifies path-planning algorithms into four categories: traditional classical algorithms, modern intelligent bionic algorithms, sampling-based planning algorithms, and machine learning algorithms. It systematically examines the concepts and characteristics of each algorithm type, evaluates their suitability across various agricultural environments, compares their convergence speeds and computational efficiencies, and discusses potential improvement strategies. The analysis reveals that traditional classical algorithms offer high precision and stability in structured farmland environments but lack dynamic adaptability. Modern intelligent bionic algorithms enhance path robustness in complex terrains through group collaboration and global optimization mechanisms, yet they face challenges with slow convergence and parameter sensitivity; sampling-based planning algorithms excel in obstacle avoidance within unstructured, dynamic scenarios, but the quality of the generated paths depends heavily on the sampling strategy; machine learning algorithms enable environment-adaptive decision-making through data-driven approaches, though they require substantial labeled data and significant computing resources. Further comparisons suggest that path-planning algorithms' future development trend will involve integrating multiple algorithms' strengths and leveraging advanced technologies such as artificial intelligence, cloud computing, and edge computing to improve adaptability, real-time performance, and intelligent decision-making capabilities in complex agricultural environments. This paper provides theoretical support and practical guidance for research on path planning for agricultural robots and offers new insights for accelerating the development of modern agriculture.

In this paper, a non-singular terminal sliding mode controller based on the adaptive technique is proposed to realize high-precision control of a spatial three-degree-of-freedom robotic arm under strong disturbances. Firstly, to ensure that the trajectory tracking error can converge to zero in finite time and to avoid the singularity problem in the control law, a control law containing an inverse tangent function is chosen. Secondly, the chattering phenomenon is eliminated by introducing the boundary layer technique. In addition, the adaptive technique is introduced to greatly improve the disturbance rejection capability of the controller while retaining the advantages of the original sliding mode surface, and the drift problem of the adaptive law is solved by introducing the dead-zone correction term. Based on the Lyapunov theory, the finite-time convergence of the system is proved, and a simulation platform and a physical experimental setup are built to physically verify the controller. Taking the IAE of the tracking process in the first joint as an example, the proposed controller improves tracking accuracy by 73%, 66%, and 34% compared to non-singular terminal sliding mode controller (NTSMC), inverse tangent-based non-singular terminal sliding mode controller (ATNTSMC), and adaptive robust non-singular fast terminal sliding mode controller (ARNFTSMC), respectively. Additionally, the proposed methodology achieves the fastest convergence rate, with improvements of 55%, 45%, and 29% over NTSMC, ATNTSMC, and ARNFTSMC, respectively. These results demonstrate the significant potential of the proposed methodology in enhancing the robustness, accuracy, and applicability of robotic systems.

Research in the field of navigable area perception technology in complex field environments holds considerable theoretical and practical value, with its notable applications including battlefield support and emergency rescue for autonomous intelligent agents. Field environments are characterized by their unstructured nature and complexity. Current research mainly focuses on fine-grained segmentation techniques in structured environments such as urban roads. However, perception technology for highly complex, field-based emergency environments remains underutilized. In this study, we perform a formal description and modeling simulation of challenging field conditions. We propose a Bidirectional Multimodal information Fusion Segmentation (BMF-Seg) model for navigable area point cloud segmentation, which employs a Texture Enhancement Method based on Gabor filters and self-attention mechanisms to enhance the visual texture features of potentially navigable areas. To further investigate, we develop a Bidirectional Fusion learning Network (BFNet) to perform fusion analysis of ground material features from RGB visual images and terrain features from 3D LiDAR point clouds. This network enables effective segmentation of navigable areas in difficult natural environments. Some experiments on a research platform demonstrate that the BMF-Seg model achieves mIoU 5% higher than commonly used point cloud segmentation models. The experiments demonstrate that the model is suitable for the point cloud segmentation task of navigable areas in complex field environments, facilitating effective perception in complex environments for autonomous intelligent agents.