Crop disease recognition is essential for effective disease control, preventing outbreaks, and minimizing farmers’ losses. Recently, advanced image processing techniques for crop disease detection, based on deep learning, have gained significant popularity. However, deploying these models in real farms remains challenging. First, these solutions often rely on remote (server)-based processing, which requires transmitting a large volume of crop images. This is often impractical in constrained connectivity environments such as rural fields. Deploying deep learning models directly on end devices is also impractical due to their high computational demands. Alternatively, lightweight models have been developed for resource-constrained edge devices and enable local processing. However, their prediction performance is often limited, particularly when handling hard and complex samples. To address these challenges, we propose iCrop+, a hybrid end-to-end system that integrates on-device AI for local processing with a powerful deep learning model for remote processing. iCrop+ selectively offloads samples via LoRa communication based on the reliability of local classification, leveraging a combination of category-based and sample-based offloading strategies. To further mitigate LoRa’s data rate limitations, the system preprocesses offloaded samples to extract and adaptively transmit only the most informative image segments, ensuring efficient data transmission without compromising accuracy. iCrop+ can operate independently or be mounted on agricultural robots or drones scouting the crop fields for remote monitoring and decision-making, such as crop disease detection. Extensive experiments on a prototype of iCrop+ demonstrate that iCrop+ outperforms two baseline approaches across multiple performance metrics, showcasing its potential for practical deployment in resource-constrained agricultural environments.

Maze-Patterned Patch Antenna Sensor for Grain Moisture Detection with Future Applications in Agricultural Robotics

This study aims to systematically map the intellectual structure, thematic evolution, and collaborative patterns of research on agricultural entrepreneurship through a bibliometric approach. Bibliographic data were collected from the Scopus and Web of Science databases, covering peer-reviewed articles published over multiple decades. Using VOSviewer, the study applied performance analysis and science mapping techniques, including keyword co-occurrence, co-authorship, institutional collaboration, and country network analyses. The results reveal that entrepreneurship constitutes the central conceptual foundation of the field, strongly interconnected with themes such as sustainability, innovation, agribusiness, and rural development. Temporal analysis indicates a clear shift from early research emphasizing institutional support, productivity, and government intervention toward more recent studies focusing on sustainability-oriented and technology-driven entrepreneurial models, including digital agriculture and agricultural robotics. Collaboration analysis shows a globally interconnected yet uneven research landscape, with scholarly output concentrated among a limited number of authors, institutions, and countries, particularly the United States, China, and India. This study contributes to the literature by providing a comprehensive overview of the evolution and current state of agricultural entrepreneurship research, offering valuable insights for scholars, policymakers, and practitioners, and identifying opportunities for future research and international collaboration in advancing sustainable and inclusive agricultural development⁠.

The working environment of fruit-picking robots is highly complex, involving numerous obstacles such as branches. Sampling-based algorithms like Rapidly Exploring Random Trees (RRTs) are faster but suffer from low success rates and poor path quality. Deep reinforcement learning (DRL) has excelled in high-degree-of-freedom (DOF) robot path planning, but typically requires substantial computational resources and long training cycles, which limits its applicability in resource-constrained and large-scale agricultural deployments. However, picking robot agents trained by DRL underperform because of the complexity and dynamics of the picking scenes. We propose a real-time policy reconstruction method based on experience retrieval to augment an agent trained by DRL. The key idea is to optimize the agent’s policy during inference rather than retraining, thereby reducing training cost, energy consumption, and data requirements, which are critical factors for sustainable agricultural robotics. We first use Soft Actor–Critic (SAC) to train the agent with simple picking tasks and less episodes. When faced with complex picking tasks, instead of retraining the agent, we reconstruct its policy by retrieving experience from similar tasks and revising action in real time, which is implemented specifically by real-time action evaluation and rejection sampling. Overall, the agent evolves into an augment agent through policy reconstruction, enabling it to perform much better in complex tasks with narrow passages and dense obstacles than the original agent. We test our method both in simulation and in the real world. Results show that the augment agent outperforms the original agent and sampling-based algorithms such as BIT* and AIT* in terms of success rate (+133.3%) and path quality (+60.4%), demonstrating its potential to support reliable, scalable, and sustainable fruit-picking automation.

Abstract Agricultural robots are typically required to operate for extended periods on orchard farms and perform a series of tasks. However, the demands and constraints of these tasks can vary significantly, and smooth transitions between different task states are crucial for improving the continuity and precision of autonomous operations. This paper addresses these challenges for a tracked agricultural robot by (i) deriving a unified kinematic–dynamic model of the platform, (ii) proposing a multi-task motion-planning method that enforces higher-order continuity in position, velocity, acceleration, and jerk, and (iii) designing an optimal tracking controller to follow the planned trajectories under kinodynamic and energy-efficiency considerations. The framework is evaluated through trajectory generation and closed-loop navigation experiments in orchard settings. The controller accurately follows the planned paths, achieving maximum navigation errors of 0.0758 m (lateral) and 0.0810 m (longitudinal), with corresponding root-mean-square errors of 0.0216 m and 0.0169 m. These results indicate that the proposed approach enables smooth, interruption-free transitions between adjacent task states and delivers precise navigation across all operational phases. The method provides a practical foundation for reliable, multi-task autonomous operation of tracked robots in precision agriculture.

Instance segmentation in agricultural robotics requires a balance between real-time performance and accuracy. This study proposes a lightweight pomelo image segmentation method based on the YOLOv8n-seg model integrated with the RepGhost module. A pomelo dataset consisting of 5076 samples was constructed through systematic image acquisition, annotation, and data augmentation. The RepGhost architecture was incorporated into the C2f module of the YOLOv8-seg backbone network to enhance feature reuse capabilities while reducing computational complexity. Experimental results demonstrate that the YOLOv8-seg-RepGhost model enhances efficiency without compromising accuracy: parameter count is reduced by 16.5% (from 3.41 M to 2.84 M), computational load decreases by 14.8% (from 12.8 GFLOPs to 10.9 GFLOPs), and inference time is shortened by 6.3% (to 15 ms). The model maintains excellent detection performance with bounding box mAP50 at 97.75% and mask mAP50 at 97.51%. The research achieves both high segmentation efficiency and detection accuracy, offering core support for developing visual systems in harvesting robots and providing an effective solution for deep learning-based fruit target recognition and automated harvesting applications.

India’s agricultural sector faces various challenges, including the adoption of sustainable production methods to improve yields and efficiency. Traditional farming methods demand intensive manual labor, which has become increasingly scarce due to an aging rural workforce and growing urban migration. The objective of this paper is to use automated technology that can increase yield while requiring less labor and lower cost. The use of large agricultural machines leads to soil compaction, which obstructs root growth and decreases crop yield due to the low absorption of nutrients and water. To address this issue, a Multipurpose Agricultural Robotic System (MARS) has been developed that is both lightweight and affordable, unlike other existing technologies that perform tasks sequentially. The Raspberry Pi 4B serves as the central controller, interfacing with soil and climate sensors for environmental monitoring. Field prototype tests indicated a sowing spacing error of 2.3 ± 0.4 cm, weed detection accuracy of 91%, weeding success rate of 87%, and a harvesting damage rate of only 6%, with an average field coverage speed of 18 m²/hour and battery autonomy of 3.5 hours. This robot has the potential to revolutionize agriculture by introducing a modern approach to traditional methods. The novelty of MARS lies in integrating sowing, weeding, harvesting, and soil monitoring into a single low-cost, lightweight platform, offering a distinct advantage over commercial robots. This innovation aims to support farmers by offering more sustainable and productive harvests.

Agricultural robot plant automatic detection integrating visual navigation and phenotype recognition

HighlightsWhat are the main findings?A lightweight YOLOv8n model integrated with C2f-FasterBlock and SE attention achieves high apple detection accuracy (mAP = 0.885) and real-time performance (83 FPS) with 37% fewer parameters and a compact 4.3 MB sizeAn end-to-end active perception framework based on ResNet50 and multi-modal fusion enables the robotic arm to autonomously navigate to optimal viewpoints, significantly reducing occlusion and improving recognition success.What are the implications of the main findings?The proposed co-design of efficient perception and active sensing offers a practical solution for reliable fruit detection in cluttered orchard environments, addressing a key bottleneck in agricultural automation.The system’s direct mapping from visual input to motion planning demonstrates a scalable paradigm for closed-loop robotic harvesting, paving the way for deployment in real-world field conditions.Addressing the issue of fruit recognition and localization failures in harvesting robots due to severe occlusion by branches and leaves in complex orchard environments, this paper proposes an occlusion avoidance method that combines a lightweight YOLOv8n model, developed by Ultralytics in the United States, with active perception. Firstly, to meet the stringent real-time requirements of the active perception system, a lightweight YOLOv8n model was developed. This model reduces computational redundancy by incorporating the C2f-FasterBlock module and enhances key feature representation by integrating the SE attention mechanism, significantly improving inference speed while maintaining high detection accuracy. Secondly, an end-to-end active perception model based on ResNet50 and multi-modal fusion was designed. This model can intelligently predict the optimal movement direction for the robotic arm based on the current observation image, actively avoiding occlusions to obtain a more complete field of view. The model was trained using a matrix dataset constructed through the robot’s dynamic exploration in real-world scenarios, achieving a direct mapping from visual perception to motion planning. Experimental results demonstrate that the proposed lightweight YOLOv8n model achieves a mAP of 0.885 in apple detection tasks, a frame rate of 83 FPS, a parameter count reduced to 1,983,068, and a model weight file size reduced to 4.3 MB, significantly outperforming the baseline model. In active perception experiments, the proposed method effectively guided the robotic arm to quickly find observation positions with minimal occlusion, substantially improving the success rate of target recognition and the overall operational efficiency of the system. The current research outcomes provide preliminary technical validation and a feasible exploratory pathway for developing agricultural harvesting robot systems suitable for real-world complex environments. It should be noted that the validation of this study was primarily conducted in controlled environments. Subsequent work still requires large-scale testing in diverse real-world orchard scenarios, as well as further system optimization and performance evaluation in more realistic application settings, which include natural lighting variations, complex weather conditions, and actual occlusion patterns.

To address the requirements for automation and intelligence of agricultural robots, this paper develops a 14degree-of-freedom dynamic model for wheeled agricultural robots. The model aims to provide a dynamic modeling foundation under the framework of modern control theory for the automation and intelligence of wheeled agricultural robots. It incorporates the Ackermann steering mechanism, MacPherson independent suspension system, tire model, and deformable soil model based on Bekker's formula. The vertical tire pressure is calculated using the deformable soil model via Bekker's formula, while tire forces are predicted by combining the tire slip angle and slip ratio with the Magic Formula Tire Model. By analyzing the force transmission effect of the suspension system, integrating the center-of-mass coupling effect analysis and the robot body model equations, the precise prediction of the attitude and motion trajectory of the wheeled agricultural robot is achieved. A co-simulation experiment using MATLAB and CarSim under the double lane change (DLC) condition is designed for validation. Experimental results demonstrate that the proposed model exhibits high consistency with the CarSim simulation results. The mean absolute errors (MAE) are 0.327° for steering wheel angle, 0.677°/s for yaw rate, 0.691° for body roll angle, and 0.944 m/s² for lateral acceleration. All errors are less than 1.5, meeting the requirements of dynamic simulation. This model can effectively predict the body attitude of wheeled agricultural robots and lay a foundation for the subsequent development of optimal control algorithms for agricultural robots.

Robotics are becoming prevalent in our everyday lives. Agriculture is the world's most significant industry, with a tremendous technological demand. It is presently appropriate to adopt robotics applications in farming since the global food chain is under strain from factors such as population expansion, climate change, population drift from rural to urban areas, and ageing populations. Robotics are seen as a means of escaping the unsettling reality. The robots are very sophisticated; they know how much water a specific plant needs. The same situation holds with fertiliser. Every plant will get just the proper quantity of fertiliser to keep it healthy. They may move into fields fast and resemble corn plants. This paper is about the need for robotics in agriculture. Several available cooperative robots, their tasks, and the associated challenges and prospects for the agriculture domain are discussed. Finally, it identified and discussed significant applications of Robotics for Agriculture. Robots can be used to monitor every plant in a field, whether big or small. This can assist in spotting any faults or concerns and provide their report immediately to the farmers. Farmers may quickly determine what types of problems exist in their fields in this manner without having to inspect them physically. These robots are remarkable for their accuracy and fineness. Agricultural robots are specialised technological devices that may help farmers with various tasks. They may be designed to develop and adapt to meet the requirements of different activities, and they can assess, consider, and perform multiple tasks. There are some limitations to implementing this technology in agriculture, such as safety, maintenance, environmental factors, high initial cost, training requirements, and increased unemployment. In the future, robots will determine the optimum planting locations, the ideal harvesting times, and the best paths for crisscrossing the farms.

Unmanned Aerial Vehicles (UAVs) have become integral to modern applications, including smart agricultural robotics, where reliability is essential to ensure safe and efficient operation. It is commonly recognized that traditional fault diagnosis approaches usually rely on vibration and noise measurements acquired via onboard sensors or similar methods, which typically require continuous data acquisition and non-negligible onboard computational resources. This study presents a portable Laser Doppler Vibrometer (LDV)-based system designed for noncontact, offboard, and high-sensitivity measurement of UAV vibration signatures. The LDV measurements are analyzed using a Deep Extreme Learning-based Neural Network (DeepELM-DNN) capable of identifying both propeller fault type and severity from a single 1 s measurement. Experimental validation on a commercial quadcopter using 50 datasets across multiple induced fault types and severity levels demonstrates a classification accuracy of 97.9%. Compared to conventional onboard sensor-based approaches, the proposed framework shows strong potential for reduced computational effort while maintaining high diagnostic accuracy, owing to its short measurement duration and closed-form learning structure. The proposed LDV setup and DeepELM-DNN framework enable noncontact fault inspection while minimizing or eliminating the need for additional onboard sensing hardware. This approach offers a practical and scalable diagnostic solution for large UAV fleets and next-generation smart agricultural and industrial aerial robotics.

In the context of modern agricultural transformation, the integration of robotic systems into plant care is emerging as a vital solution to address challenges such as labour shortages, increased production demands, and the need for sustainable farming practices. This research focuses on the mechanical design and fabrication of a compact, modular robotic platform specifically tailored for agricultural plant care applications. The robot is designed to operate in greenhouses or open fields and is equipped with a four-wheel differential drive system, a chain transmission mechanism, and a load-distributing aluminium top plate to support essential components such as a water tank. Finite Element Analysis (FEA) was conducted to validate the structural reliability of the chassis and loadbearing elements, showing low stress and strain well below material limits, thereby ensuring operational stability and safety. A prototype was manufactured using accessible materials and methods, demonstrating the feasibility of the proposed design in terms of assembly, mobility, and structural integrity. This study contributes a mechanically robust and scalable foundation for future integration with sensors and control systems, advancing the development of smart, automated agricultural robotics.

Autonomous navigation is a core enabler of smart agriculture, where path planning and trajectory tracking control play essential roles in achieving efficient and precise operations. Path planning determines operational efficiency and coverage completeness, while trajectory tracking directly affects task accuracy and system robustness. This paper presents a systematic review of agricultural robot navigation research published between 2020 and 2025, based on literature retrieved from major databases including Web of Science and EI Compendex (ultimately including 95 papers). Research advances in global planning (coverage and point-to-point), local planning (obstacle avoidance and replanning), multi-robot cooperative planning, and classical, advanced, and learning-based trajectory tracking control methods are comprehensively summarized. Particular attention is given to their application and limitations in typical agricultural scenarios such as open-fields, orchards, greenhouses, and hilly slopes. Despite notable progress, key challenges remain, including limited algorithm comparability, weak cross-scenario generalization, and insufficient long-term validation. To address these issues, a scenario-driven “scenario–constraint–performance” adaptive framework is proposed to systematically align navigation methods with environmental and operational conditions, providing practical guidance for developing scalable and engineering-ready agricultural robot navigation systems.

To meet the growing demand for automated blueberry harvesting in smart agriculture, this study proposes an improved lightweight detection network, termed M-YOLOv11n, for fast and accurate blueberry maturity detection in complex natural environments. The proposed model enhances feature representation through an improved lightweight multi-scale design, enabling more effective extraction of fruit features under complex orchard conditions. In addition, attention-based feature refinement is incorporated to emphasize discriminative ripeness-related cues while suppressing background interference. These design choices improve robustness to scale variation and occlusion, addressing the limitations of conventional lightweight detectors in detecting small and partially occluded fruits. By incorporating MsBlock and the attention mechanism, M-YOLOv11n achieves improved detection accuracy without significantly increasing computational cost. Experimental results demonstrate that the proposed model attains 97.0% mAP50 on the validation set and maintains robust performance under challenging conditions such as occlusion and varying illumination, achieving 96.5% mAP50. With an inference speed of 176.6 FPS, the model satisfies both accuracy and real-time requirements for blueberry maturity detection. Compared with YOLOv11n, M-YOLOv11n increases the parameter count only marginally from 2.60 M to 2.61 M, while maintaining high inference efficiency. These results indicate that the proposed method is suitable for real-time deployment on embedded vision systems in smart agricultural harvesting robots and supports early yield estimation in complex field environments.

Collaborative robotics in agriculture is designed to automate labor-intensive processes. In contrast to traditional autonomous systems, collaborative multi-agent robotic systems require active interaction between robots and human operators. This interaction creates the need for new methods for coordination, adaptation, and safety assurance in uncertain and dynamically changing environments. ( Research purpose ) The study aims to develop both theoretical and practical approaches to modeling the behavior and control of collaborative multi-agent robotic systems. The primary objective is to ensure efficient task allocation, coordinated agent behavior, and safe human-robot interaction during fruit harvesting operations. ( Materials and methods ) To achieve these objectives, the study employed methods from game theory, machine learning, and risk-aware control. A mathematical model was developed to describe the interactions among agents, incorporating the probabilistic nature of the environment and the involvement of a human operator. The proposed solutions were validated through a combination of numerical simulations and experimental data collected from a testbed replicating real-world agricultural scenarios. ( Results and discussion ) Algorithms were developed to enable coordination, adaptation, and dynamic task redistribution within the collaborative multi-agent robotic system. These algorithms demonstrated robustness against sensor inaccuracies, communication delays, and external disturbances typical of agricultural settings. Special attention was given to the system’s ability to adapt to human operator inputs, including task prioritization and context-sensitive interaction strategies. Simulation results showed enhanced system performance, characterized by more balanced task distribution among robots, reduced conflict during joint operations, and minimized idle time. Safety metrics also improved, including a reduction in collision risks and fewer incorrect responses to the presence of human operators in the work area. ( Conclusions ) The developed models and algorithms provide a foundation for the design of intelligent collaborative multi-agent robotic systems capable of adaptive and safe interaction in agricultural production. Their application can enhance the efficiency of automated harvesting processes while reducing reliance on manual labor.

ABSTRACT To improve the auxiliary role of agricultural robots in agricultural farming, based on the screw theory, this paper studies the description of the position and posture coordinates of agricultural robots in two-dimensional and three-dimensional spaces. This paper then establishes the coordinate system of the manipulator in accordance with the regulations for establishing the coordinate system of agricultural robots. It uses mathematical formulas to derive and calculate the solution process of forward kinematics and inverse kinematics in detail, providing sample data for machine learning inverse kinematics algorithms. In addition, according to the control needs, this paper simplified the structure of the crawler mobile platform in the coordinate system. It established a kinematic model, which serves as the basis for the path-planning algorithm. Finally, this paper combines experimental research to verify the agricultural assisted agricultural robots developed in this paper. The experimental research results confirm the feasibility of this method.

IntroductionThree-dimensional (3D) point clouds acquired by LiDAR are fundamental for applications such as autonomous navigation, mobile robotics, infrastructure inspection, and cultural-heritage documentation. However, environmental disturbances and sensor limitations often yield incomplete or noisy point clouds, degrading downstream performance. This study addresses robust, high-fidelity point cloud completion under such practical conditions.MethodsWe propose an unsupervised deep learning framework, Multi-Resolution Completion Net (MRC-Net), which builds on ShapeInversion by integrating a Generative Adversarial Network (GAN) inversion strategy with multi-resolution principles. The architecture comprises an encoder for feature extraction, a generator for completion, and a discriminator to assess geometric integrity and detail. Two key designs enable strong performance without supervision: (i) a multi-resolution degradation mechanism that guides reconstruction across coarse-to-fine scales, and (ii) a multi-scale discriminator that captures both global structure and local details.ResultsExtensive experiments on multiple datasets demonstrate that MRC-Net achieves accuracy comparable to leading supervised approaches. On virtual datasets (e.g., CRN), MRC-Net attains an average Chamfer Distance (CD) of 8.0 and an F1 score of 91.3. On a custom dataset targeting agricultural scenarios, the model preserves object integrity across varying complexity: for regular cartons, it achieves CD 3.3 and F1 97.3; for structurally complex simulated plants, it maintains overall shape while delivering average CD 8.6 and F1 88.1.DiscussionThese results indicate that MRC-Net advances unsupervised point cloud completion by balancing global shape consistency with fine-grained detail. The method provides a reliable data foundation for downstream tasks—including autonomous navigation, high-precision 3D modeling, and agricultural robotics—thereby contributing to improved data quality in precision-agriculture and related domains.

In this study, the importance of utilizing UGVs and UAVs within the agricultural ecosystem is highlighted, emphasizing their potential to reduce environmental impact, conserve resources, and ensure food security. The challenges and future prospects of this technology in the pursuit of a more sustainable and productive agriculture sector are also examined. Ultimately, the integration of UGVs and UAVs into Industry-Agriculture 5.0 represents a shift towards a data-driven and environmentally conscious approach to farming, promising a brighter and more sustainable future for agriculture. Prototype agricultural robot with sensing capabilities developed, tested. Mechanical, electronic, and software components integrated. Design created using Autodesk Inventor and SolidWorks. Electronic circuits designed in Proteus. Software development in Matlab and Visual C++. The chassis is constructed from aluminum and steel. Robot tested at Aydın Adnan Menderes University and Manisa Viticulture Research Institute. Operates on electrical power, 8-hour working capacity, 49-minute recharge time. Individual motors for each wheel, differential drive method, 34.85 horsepower.

An approach to address current challenges in agriculture caused by climate change, the increasing global population and the loss of biodiversity is precision farming, for which agricultural robotics is a key enabler. The Field Robot Event (FRE) 2023 has challenged student teams to develop and improve autonomous agricultural robots. This paper presents the improvements to our field robot “Carbonite,” which is developed at the Schülerforschungszentrum (SFZ) Südwürttemberg. Our lightweight and compact robot design, supported by our advanced and efficient navigation algorithm, enabled our robot to quickly move through fields. Additionally, we introduced our newly developed system for targeted and precise application of water, fertilizer and herbicides, based on an intelligent gap detection algorithm to avoid wasting resources. Also, we trained an object recognition AI model based on the You Only Look Once (YOLO) models, allowing the robot to appropriately respond based on the type of obstacle. Carbonite managed to secure the first place in both the navigation task and the plant treatment task, benefiting from the lightweight design and the resulting high robot driving speed, enabling us to win the overall FRE 2023 contest.