[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"item-111":3,"related-111":63},{"id":4,"title":5,"url":6,"summary":7,"summary_zh":8,"content":9,"source_name":10,"source_url":8,"published_at":11,"category":12,"cover_url":8,"hotness":13,"is_selected":14,"score":15,"score_detail":16,"sources":18,"tags":20,"primary_topic":24,"secondary_topics":30,"classification":31,"search_phrases":56,"slug":59,"view_count":60,"doi":61,"paper":8,"created_at":62},111,"Global-Local Integrated Path Planning Strategy for Agricultural Robots in Complex Hilly Orchard Environments","https:\u002F\u002Fwww.frontiersin.org\u002Fjournals\u002Fsustainable-food-systems\u002Farticles\u002F10.3389\u002Ffsufs.2026.1866736\u002Ffull","Li Tian, Xingjia Pan提出一种面向复杂丘陵果园环境的农业机器人全局-局部一体化路径规划策略,有效解决农业机器人在复杂三维地形下的导航与障碍物避让问题。",null,"About us\nAll journals\nAll articles\nSubmit manuscript\nSubmit data\nSearch\nLogin\nFrontiers in Sustainable Food Systems\nSections\nArticles\nResearch Topics\nEditorial board\nAbout journal\nPublished in\n\nFrontiers in Sustainable Food Systems\n\nAgricultural and Food Economics\n4.1\nimpact factor\n6\ncitescore\n\nPart of a Research Topic\n\nStrengthening Community Governance for Resilient and Sustainable Food Futures\n\n15k views\n14 articles\nEditor & Reviewers\n\nEdited by\n\nIVAN ANTONIO GARCIA-MONTALVO\n\nReviewed by\n\nJessa Frida Festijo\n\nOliver Eboy\n\nOutline\n\nAbstract\n\n1 Introduction\n\n2 Literature review\n\n3 Research design\n\n4 Analysis of the online fresh produce market\n\n5 Market research and analysis\n\n6 Conclusions and recommendations\n\nData availability statement\n\nEthics statement\n\nAuthor contributions\n\nFunding\n\nConflict of interest\n\nGenerative AI statement\n\nPublisher’s note\n\nSupplementary material\n\nReferences\n\nFigures and Tables\n\nFigure 1\n\nView in article\n\nFigure 2\n\nView in article\n\nFigure 3\n\nView in article\n\nFigure 4\n\nView in article\n\nFigure 5\n\nView in article\n\nFigure 6\n\nView in article\n\nFigure 7\n\nView in article\n\nFigure 8\n\nView in article\n\nFigure 9\n\nView in article\n\nFigure 10\n\nView in article\n\nTable 1\n\nSatisfaction reliability test table.\n\nView in article\n\nTable 2\n\nSignificance and reliability test table.\n\nView in article\n\nTable 3\n\nKMO and Bartlett’s test of sphericity.\n\nView in article\n\nTable 4\n\nKMO and Bartlett’s test of sphericity.\n\nView in article\n\nTable 5\n\nThematic feature keywords.\n\nView in article\n\nTable 6\n\nWeight allocation table for fresh food theme.\n\nView in article\n\nTable 7\n\nSelf-adjoint fuzzy matrix table.\n\nView in article\n\nTable 8\n\nCorresponding numbers for online fresh food platforms.\n\nView in article\n\nTable 9\n\nImproved specification weighting matrix.\n\nView in article\n\nTable 10\n\nResults of the improved intuitive fuzzy TOPSIS evaluation.\n\nView in article\n\nTable 11\n\nProduct data characteristics and descriptive statistics.\n\nView in article\nORIGINAL RESEARCH article\n\nFront. Sustain. Food Syst., 02 July 2026\n\nSec. Agricultural and Food Economics\n\nVolume 10 - 2026 | https:\u002F\u002Fdoi.org\u002F10.3389\u002Ffsufs.2026.1866736\n\nFresh food retail satisfaction and sales forecasts: digital integration for sustainability in Guiyang, China\n\nChangpeng Yang 1†\n\nK\nH\n\nKeyan He 2†\n\nF\nH\n\nFu Hong 3†\n\nY\nZ\n\nYufeng Zhu 4\n\nLiming Chen 1*\n\n1. School of Tourism and Resource Environment, Qiannan Normal University for Nationalities, Duyun, China\n\n2. Party School of the Libo County Committee of the Communist Party of China, Libo, Guizhou, China\n\nSee more\n\nArticle metrics\n\nView details\n\n1,4k\n\nViews\n\n118\n\nDownloads\n\nAbstract\n\nIntroduction: \n\nDriven by the digital economy, China’s fresh food retail sector is accelerating its transition toward omnichannel integration. However, existing research has largely focused on first-tier cities in eastern China, paying insufficient attention to the heterogeneity of underdeveloped regions in southwestern China, and lacking a systematic analytical framework that integrates multi-source data and multiple methodologies.\n\nMethods: \n\nThis study takes six mainstream online fresh food platforms in Guiyang, China (including the local platform ‘Guihuihui’) as its research subjects. By integrating two rounds of stratified sampling questionnaire surveys (1,744 valid responses) with 16,321 online review data points, we constructed a multi-model integrated analytical framework comprising an LDA topic model, an improved intuitive fuzzy TOPSIS method, and the AT-DeepAR deep learning model. This framework was used to systematically investigate the mechanisms underlying the evolution of fresh food retail formats, the factors influencing consumer satisfaction, and market sales forecasting.\n\nResults: \n\nThe study found that: (1) transaction costs and service quality are the core drivers of digital transformation in fresh food retail, with consumers placing the greatest emphasis on delivery speed and shopping experience; (2) platform satisfaction rankings are as follows: Duoduo Grocery > Meituan Youxuan > JD NOW > Freshippo > Guihuihui > Dingdong Maicai, with national platforms enjoying significantly higher brand recognition than local platforms; (3) the AT-DeepAR model, incorporating an attention mechanism, can accurately capture the seasonal fluctuations in fresh produce sales, with predicted values closely matching actual figures (within a 95% confidence interval). It forecasts that the average monthly sales of fresh produce platforms in Guiyang will steadily increase to approximately 340 million yuan by 2025.\n\nDiscussion: \n\nAt the theoretical level, this paper fills a research gap on fresh food e-commerce in southwestern China by establishing an integrated analytical paradigm combining questionnaire data, text mining, and deep learning. At the practical level, it provides empirical evidence to support differentiated competition, transaction cost optimization, and the digital upgrading of the regional fresh food industry for local platforms in Guiyang. It holds significant practical value for promoting the sustainable development of fresh food retail, enhancing the efficiency of urban–rural food distribution, and facilitating the low-carbon transition of supply chains.\n\n1 Introduction\n\nThe rapid development of internet technology and shifts in consumer behavior have made online shopping an integral part of daily life. Against this backdrop, online fresh food shopping has emerged and is growing rapidly. As a key province in south-western China, Guizhou has seen sustained growth in the demand for fresh produce in recent years, driven by both urbanization and consumption upgrades (Guizhou Provincial Bureau of Statistics, Guizhou Provincial Office of the Leading Group for the Seventh National Population Census, 2021). Fueled by the pandemic, more consumers have turned to ‘online shopping’, presenting numerous challenges to traditional fresh food retail models, which suffer from low delivery efficiency, high costs and a lack of personalized service.\n\nIn recent years, the evolution of the fresh food retail sector has become increasingly evident. Traditional farmers’ markets and supermarkets have gradually shifted from offline to online operations, while new retail models such as front-end warehouses (e.g., Freshippo, Dingdong Maicai) and community group-buying platforms (e.g., Meituan Youxuan, Duoduo Grocery) have rapidly emerged. These platforms meet consumers’ demands for timeliness and convenience through an ‘online ordering + offline delivery’ model, making the integrated online-offline retail model increasingly prominent.\n\nAgainst this background, Guizhou Province launched the “Guihuihui” online fresh produce platform in 2022, integrating high-quality fresh produce resources from Guiyang and its surrounding areas (Zhongwang News, 2022). The platform breaks through the spatial and temporal barriers of traditional retail, centring on the digital reconstruction of the “people-goods-place” model to drive the transformation of supermarket retail formats. Through efficient service and high-quality products, it has rapidly gained market recognition and consumer trust, demonstrating strong growth momentum.\n\nAs a critical link in the food supply chain, the digital transformation and service optimization of fresh food retail have a direct bearing on food security, resource conservation, the reduction of wastage and losses, and the improvement of the quality of life for urban and rural residents. This study focuses on consumer groups in Guiyang, China, conducting an empirical analysis of consumer satisfaction through a questionnaire survey (Hongxin and Yuhui, 2015). It examines online fresh produce platforms in Guiyang, analyses the evolution of the fresh produce retail sector, explores the extent to which these platforms meet consumer needs and their impact on market development, reviews the current state of the fresh produce market, identifies shortcomings, and offers constructive suggestions and references for the market’s future development.\n\n2 Literature review\n\nDriven by the dual forces of the digital economy and consumption upgrades, China’s fresh food retail sector is rapidly shifting from traditional offline models toward diverse business formats such as omnichannel integration, community group buying and forward-depot models. Online fresh food retail has become the mainstream consumption channel due to its convenience, efficiency and value for money. The evolution of fresh food retail formats is the result of the combined influence of transaction costs, digital technology, cold-chain logistics and consumer behavior, whilst consumer satisfaction serves as a core indicator influencing repeat purchases and brand loyalty. Existing research has yielded substantial findings regarding the mechanisms of business model evolution, factors influencing satisfaction, evaluation models and sales forecasting; however, studies have generally focused on first-tier cities in the east and national platforms. There remains a notable lack of integrated research that targets south-western China, focuses on provincial capitals, compares local and national platforms, and combines multi-source data with multiple methodologies. This paper systematically reviews the relevant literature and identifies research gaps, thereby providing a theoretical foundation for subsequent empirical analysis.\n\n2.1 Research on the evolution of fresh produce retail formats\n2.1.1 The evolutionary trajectory of fresh produce retail formats\n\nExisting literature has established a relatively systematic research framework concerning the evolution of fresh produce retail formats. This framework primarily unfolds along three main threads: the iteration of new retail models, the integration of digital technologies, and consumer behavior-driven trends. It provides a theoretical foundation for understanding the transformation of China’s fresh produce distribution system from traditional channels toward an integrated online-offline model characterized by intelligence, efficiency and diverse coexistence. Regarding new retail models and business formats, the digital economy has driven the rapid proliferation of models such as forward-depot logistics, store-warehouse integration and community group buying, which have become the core directions for business format upgrading. Relevant studies, using cases such as Freshippo, have revealed the organizational logic, supply chain coordination and fulfilment efficiency characteristics of new retail models (Chen et al., 2022; Zhang et al., 2024). Omnichannel integration is regarded as a core feature of new retail, capable of significantly enhancing consumer engagement and channel loyalty (Chen et al., 2022) Meanwhile, traditional farmers’ markets have achieved modernisation through semi-formalization, forming a complementary and co-existing relationship with e-commerce platforms (Yao, 2025; Yuan et al., 2026). The retail sector’s overall evolution toward digitalisation, sustainability, platformisation and ecosystem-based models aligns closely with the reality of multi-platform coexistence in Guiyang (Akhtoian, 2023). In terms of digital technology integration, big data, artificial intelligence and supply chain optimization are deeply intertwined, driving the fresh produce retail sector toward precision and intelligence. Blockchain and green low-carbon technologies help reduce coordination costs and enhance supply chain transparency (Chen and Yin, 2023); the integration of online and offline channels significantly improves service quality and customer loyalty (Zhang and Liang, 2023); deep learning models based on LSTM and GWO-LSTM have been widely applied to sales forecasting and pricing optimization (Wang et al., 2024). Innovations in logistics models and carbon reduction practices play a crucial role in ensuring fresh produce quality and stabilizing supply (He and Sun, 2025). On the consumer-driven front, consumption preferences, value perceptions and channel choices directly shape the retail landscape. Research indicates that segmentation of consumption motives, quality demands, freshness preferences and price sensitivity significantly influence channel decisions (Kamal, 2025; Naveuler, 2025; Sarker et al., 2024), whilst consumption upgrades are driving the fresh produce supply chain toward greater health-consciousness, experiential value and immediacy (White, 2022). These findings provide the theoretical foundation for this study’s examination of business format evolution, consumer behavior and service experience in Guiyang. At the same time, digital transformation not only improves operational efficiency but also reduces wastage of fresh produce through accurate forecasting, optimizes delivery routes, and facilitates low-carbon logistics and the sustainable use of resources.\n\n2.1.2 Factors influencing the evolution of fresh food retail formats\n\nExisting research has systematically analyzed the driving mechanisms behind the evolution of fresh food retail formats from multiple perspectives, including digitalisation, consumer behavior, product quality, supply chain efficiency, physical retail environments and the policy landscape. Digital transformation and the O2O model are the core drivers of format upgrading, whilst the optimization of digital processes can reduce wastage and enhance fulfilment efficiency (Ma et al., 2022; Yang et al., 2022; Belwal et al., 2024). Consumer perceived value, service quality and trust in food safety significantly influence purchase intention and channel loyalty (Činčalová et al., 2021; Xing et al., 2024; Thuannadee and Praneetpholkrang, 2025), whilst product safety, traceability and supply chain efficiency constitute key constraints on the sustainable development of business formats (Cheng et al., 2026; Lestari et al., 2024). The physical environment, spatial layout and location selection of brick-and-mortar stores also influence the consumer experience (Maciel et al., 2023). At the macro level, policy regulations, urbanization and consumption upgrading collectively drive the transformation of traditional retail formats (Yuan et al., 2026).\n\n2.2 Theoretical framework and influencing factors of consumer satisfaction with online fresh food purchases\n2.2.1 Theoretical foundations of consumer satisfaction\n\nThe theoretical framework of consumer satisfaction is well-established, providing a solid academic foundation for research into satisfaction with online fresh produce. Cardozo (Cardozo, 1965) pioneered the revelation of the role satisfaction plays in promoting repeat purchases (Cardozo, 1965); the Expectation-Confirmation Theory (ECT) is widely used to explain the mechanisms underlying the formation of satisfaction (Shen et al., 2022). Satisfaction is a multidimensional construct influenced by multiple factors such as product attributes, demographic characteristics and context (Yılmaz and Balcıoğlu, 2026), and its application continues to expand across scenarios including omnichannel retail, service interactions and green consumption (Yaqub et al., 2022; Xu et al., 2022; Wang et al., 2023). In marketing and logistics theory, satisfaction often serves as a core mediating or outcome variable, significantly driving customer loyalty and willingness to continue using a product or service (Kerschbaumer et al., 2022; El Moussaoui et al., 2022; Kovalskii, 2024). This theoretical framework lays the foundation for the use of LDA and the improved intuitive fuzzy TOPSIS method in this study to evaluate satisfaction.\n\n2.2.2 Factors influencing satisfaction with online fresh food\n\nIn the context of online fresh food, satisfaction is driven by multiple factors, including logistics services, product quality, service quality and brand image. Confirmed that logistics services are the most significant factor influencing satisfaction (Guan et al., 2022); product quality, freshness and safety also play a crucial role (Kang and Namkung, 2024); service quality indirectly influences behavioral intention through perceived value (Wang and Theerakittikul, 2022; Thuannadee and Praneetpholkrang, 2025); and brand image has a significant positive effect on satisfaction and purchase intention (Erkmen and Turegun, 2022). These findings provide direct justification for the selection of evaluation indicators in this study, including product source safety, product integrity, delivery service, and after-sales response. Furthermore, from a sustainability perspective, high satisfaction levels help to improve user retention, reduce customer acquisition costs and minimize waste of resources, and are central to the platform’s long-term sustainable operation.\n\n2.3 Research progress on evaluation methods and sales forecasting models for fresh food E-commerce\n2.3.1 Satisfaction evaluation models and methods\n\nTraditional satisfaction evaluation has primarily relied on methods such as IPA and TOPSIS; in recent years, it has been upgraded through the integration of positive text mining and machine learning. Developed a food quality evaluation mode (Valderrama et al., 2021); emphasized the necessity of a multi-dimensional service quality framework (Yuan et al., 2026); and (Mao, 2025) highlighted the superior accuracy and adaptability of intelligent models. With the digital transformation, technologies such as AI-powered recommendations, dynamic pricing and sentiment analysis have been employed to enhance satisfaction (Dai and Liu, 2024; Kumari et al., 2024; Mittal, 2025). Text mining and LDA topic modeling can efficiently extract consumer concerns and sentiment tendencies, thereby enhancing the objectivity of evaluations (Singh and Glińska-Neweś, 2022; Song et al., 2022; Arifin and Purnama, 2023), providing methodological support for the adoption of LDA and the improved intuitive fuzzy TOPSIS in this study.\n\n2.3.2 Deep learning models for sales forecasting\n\nDeep learning has demonstrated superior capabilities in capturing time-series patterns and volatility in sales forecasting. LSTM and its variants are widely used in retail sales forecasting (Hu et al., 2022; Samonte et al., 2022), and the introduction of self-attention mechanisms can further enhance the capture of key information (Kossen et al., 2021). Confirmed the accuracy advantages of deep learning in time series forecasting (Messaoudi et al., 2023); however, research also suggests a risk of overfitting in scenarios involving small sample sizes and high volatility (Peng et al., 2022; Tilshi and Punsisi, 2022). Accordingly, this paper introduces the AT-DeepAR model—which utilizes an attention mechanism to enable accurate sales forecasting, thereby significantly reducing inventory wastage in the fresh produce sector and lowering energy consumption in the cold chain. It serves as a key technological enabler for achieving low-carbon, efficient and sustainable operations in fresh produce retail.\n\n2.4 Research on the sustainable development of fresh food\n\nExisting research provides a solid foundation for the sustainable development of fresh food and highlights the intrinsic links between food safety, environmental sustainability and economic viability. Enhancing the environmental sustainability of food supply chains is crucial to achieving several of the United Nations Sustainable Development Goals (SDGs), particularly Zero Hunger (Goal 2) and Responsible Consumption and Production (Goal 12); They emphasize that food safety measures must be balanced against environmental factors to avoid adverse effects (López-Gálvez et al., 2021). Further research supports this view, highlighting the necessity of establishing a sustainable fresh food cold chain system that balances environmental, social and economic benefits; a systematic review indicates that sustainable practices in this field can enhance market viability and resource utilization efficiency (Liao et al., 2021). Furthermore, innovations in food packaging, such as biodegradable materi","Frontiers in Sustainable Food Systems 2026-07-02","2026-07-02T01:00:00Z","论文",10,false,0,{"impact":15,"substance":15,"depth":15,"authority":15,"freshness":15,"relevant":15,"comment":17},"标题与正文不符，正文为生鲜零售满意度研究，非农业机器人路径规划，不相关。",[19],{"name":10,"url":6},[21,22,23],"农业机器人","路径规划","丘陵果园",{"code":25,"slug":26,"name":27,"path":28,"parent_code":29},"B2","fruit","果树","\u002Ftopic\u002Fhorticulture\u002Ffruit","B",[],{"status":32,"locked":14,"version":33,"method":34,"evidence":35},"classified","2026-10-22.2","rules",{"candidates":36,"agricultural_evidence":39},[37,44,50],{"code":25,"score":38,"supported":39,"title":40,"abstract":42,"tags":43},8,true,[41],"orchard",[],[],{"code":45,"score":46,"supported":39,"title":47,"abstract":48,"tags":49},"F6",4,[],[21],[21],{"code":51,"score":52,"supported":14,"title":53,"abstract":54,"tags":55},"F5",1,[],[],[22],[57,58],"农业机器人 丘陵果园 路径规划","农业机器人 丘陵果园","农业机器人丘陵果园路径规划-111",11,"10.3389\u002Ffsufs.2026.1866736\u002Ffull","2026-07-31T00:00:56.026421Z",{"total":64,"page":52,"page_size":64,"items":65},6,[66,154,218,279,325,389],{"id":67,"title":68,"url":69,"summary":70,"summary_zh":71,"content":8,"source_name":72,"source_url":69,"published_at":73,"category":12,"cover_url":8,"hotness":13,"is_selected":14,"score":74,"score_detail":75,"sources":80,"tags":82,"primary_topic":87,"secondary_topics":88,"classification":89,"search_phrases":115,"slug":118,"view_count":52,"doi":119,"paper":120,"created_at":153},4889,"A Distributed Multirobot Architecture for Decoupled Inspection and Localized Treatment in Precision Agriculture","https:\u002F\u002Fdoi.org\u002F10.1002\u002Faisy.70593","This work presents a distributed multirobot architecture for autonomous inspection and precision treatment in agriculture, validated under real vineyard conditions. The proposed system integrates a heterogeneous fleet of four specialized ground robots (two inspection and two treatment platforms), coordinated through a cloud‐based mission control center (MCC). The MCC acts as the cognitive layer of the system, integrating CNN‐based disease detection, mission planning, and fleet management, while the robotic platforms operate as autonomous agents responsible for sensing and actuation. Inspection and treatment processes are separated through an event‐driven, grid‐based coordination mechanism maintained by the MCC, enabling asynchronous and independent operation of both robot types without direct inter‐robot communication. Georeferenced images acquired by the inspection robots are transmitted in real time to the MCC, where the disease detection model identifies affected areas and updates the treatment grid accordingly. Experimental validation in a commercial vineyard demonstrated the feasibility of the approach: the model achieved a recall of 95.2% against expert annotations at a 16% infection prevalence, no data losses were recorded across communication streams, and 21.1% of acquired images generated treatment events. The proposed architecture is not tied to a specific task and can be extended to other precision agriculture scenarios.","本文提出了一种用于农业自主巡检与精准处理的分布式多机器人架构，并在真实葡萄园条件下进行了验证。所提出的系统集成了由四台专用地面机器人（两台巡检平台和两台处理平台）组成的异构机器人集群，并通过基于云的任务控制中心（MCC）进行协调。MCC作为系统的认知层，集成了基于CNN的病害检测、任务规划和集群管理，而机器人平台则作为负责感知与执行的自主智能体。巡检与处理过程通过MCC维护的事件驱动、基于网格的协调机制实现分离，使两类机器人能够异步独立运行，无需机器人之间的直接通信。巡检机器人采集的地理参考图像实时传输至MCC，由病害检测模型识别受影响区域并相应更新处理网格。在商业葡萄园中的实验验证表明了该方法的可行性：在16%感染率下，模型相对于专家标注达到了95.2%的召回率，各通信流中未记录到数据丢失，21.1%的采集图像触发了处理事件。所提出的架构不局限于特定任务，可扩展至其他精准农业场景。","Advanced Intelligent Systems","2026-10-12T00:00:00Z",79,{"impact":76,"substance":77,"depth":76,"authority":78,"freshness":38,"relevant":52,"comment":79},18,22,13,"多机器人分布式架构在真实葡萄园验证，检出召回率95.2%，对精准农业机器人协同有参考价值。",[81],{"name":72,"url":69},[83,21,84,85,86],"智慧农业","精准施药","病害检测","多机器人协同",{"code":25,"slug":26,"name":27,"path":28,"parent_code":29},[],{"status":32,"locked":14,"version":33,"method":34,"evidence":90},{"candidates":91,"agricultural_evidence":39},[92,100,106,111],{"code":93,"score":94,"supported":39,"title":95,"abstract":97,"tags":99},"I1",15,[96],"precision agriculture",[98,96],"精准农业",[83],{"code":25,"score":101,"supported":39,"title":102,"abstract":103,"tags":105},3,[],[104],"葡萄",[],{"code":107,"score":52,"supported":14,"title":108,"abstract":109,"tags":110},"F4",[],[],[84],{"code":45,"score":52,"supported":14,"title":112,"abstract":113,"tags":114},[],[],[21],[116,117],"多机器人 葡萄园 病害检测","精准农业 机器人 架构","多机器人葡萄园病害检测-4889","10.1002\u002Faisy.70593",{"doi":119,"openalex_id":121,"authors":122,"venue":72,"cited_by_count":15,"oa_url":145,"card":146,"direction":150,"ingested_from":152},"W7221635124",[123,126,129,132,135,137,139,142],{"name":124,"orcid":125},"Jose Maria Bengochea-Guevara","https:\u002F\u002Forcid.org\u002F0000-0003-4081-7325",{"name":127,"orcid":128},"Héctor Montes","https:\u002F\u002Forcid.org\u002F0000-0001-8638-9966",{"name":130,"orcid":131},"María Nuria Conejero","https:\u002F\u002Forcid.org\u002F0000-0002-9372-3739",{"name":133,"orcid":134},"Laura Garrido-Rey","https:\u002F\u002Forcid.org\u002F0009-0002-1782-5175",{"name":136,"orcid":8},"Carlos Ranz",{"name":138,"orcid":8},"Angelo Vincenzo Barra",{"name":140,"orcid":141},"Dionisio Andújar","https:\u002F\u002Forcid.org\u002F0000-0002-5801-0944",{"name":143,"orcid":144},"Ángela Ribeiro","https:\u002F\u002Forcid.org\u002F0000-0001-5807-8132","https:\u002F\u002Fonlinelibrary.wiley.com\u002Fdoi\u002Fpdfdirect\u002F10.1002\u002Faisy.70593",{"tldr":147,"method":148,"finding":149,"direction":150,"opportunity":151},"提出分布式多机器人架构，在葡萄园实现异步巡检与精准治疗。","四台异构地面机器人、云任务控制中心、CNN病害检测与网格协调。","病害检测召回率95.2%，通信无数据丢失，21.1%图像触发治疗事件。","智慧农业 \u002F 农业物联网","可探索去中心化协同、多作物迁移及边缘计算下的实时病害决策优化。","openalex","2026-10-14T23:30:13.144306Z",{"id":155,"title":156,"url":157,"summary":158,"summary_zh":159,"content":8,"source_name":160,"source_url":157,"published_at":161,"category":12,"cover_url":8,"hotness":13,"is_selected":14,"score":162,"score_detail":163,"sources":168,"tags":170,"primary_topic":174,"secondary_topics":175,"classification":176,"search_phrases":194,"slug":197,"view_count":101,"doi":198,"paper":199,"created_at":217},3965,"Design and evaluation of a hybrid fin-ray finger with inflatable soft pads for apple harvesting under leaf interference","https:\u002F\u002Fdoi.org\u002F10.1016\u002Fj.compag.2026.112477","This paper presents the design, modelling, and experimental validation of a Hybrid Fin-Ray Finger (HFRF) integrating segmented inflatable Dragon Skin 30 silicone membrane pads for adaptive and bruise-free robotic apple harvesting. Harvesting fruits within cluttered canopy environments demands a balance between gentle contact to prevent bruising and sufficient pulling and detachment torque. However, conventional Fin-Ray fingers often exhibit limited torsional resistance, non-uniform force distribution, and reduced contact area, leading to slip and localized pressure concentrations on the fruit surface. To overcome these limitations, we integrate inflatable soft membrane pads into the Fin-Ray structure, increasing the effective finger width and improving surface conformity and pressure distribution. An analytical model is developed to predict membrane deformation and contact angle as a function of internal pressure, and the results are validated through finite element simulations and experimental testing. Pressure-sensitive film experiments demonstrate that Dragon Skin 30 inflation at 30 kPa increased contact area by up to 40% and reduced mean contact pressure by 11% relative to the unpressurised configuration, while EcoFlex 00–30 exhibited volumetric expansion under pressure that reduced conformal contact across the tested range. Performance evaluations were conducted under three standardized leaf interference conditions (0%, 50%, and 100% occlusion) using real Pink Lady apple leaves, with statistical analysis performed using one-way ANOVA and Tukey HSD post-hoc comparisons. At 30 kPa without leaf interference, normal detachment force increased by 148% relative to the no-pad configuration and torsional moment increased by 58% relative to the unpressurised condition, while maintaining performance above the detachment threshold. Under full leaf occlusion, torsional performance collapsed regardless of inflation pressure, identifying full canopy occlusion as the primary limiting factor for torsional resistance. These results confirm that the proposed HFRF provides pressure-adaptive, slip-resistant, and bruise-mitigating grasping performance suitable for robotic apple harvesting in unstructured orchard environments.","本文介绍了一种混合鳍条手指（Hybrid Fin-Ray Finger，HFRF）的设计、建模与实验验证，该手指集成了分段式充气Dragon Skin 30硅胶膜垫，用于自适应且无瘀伤的机器人苹果采摘。在杂乱的树冠环境中采摘果实，需要在轻柔接触以防止瘀伤与足够的拉拔和分离力矩之间取得平衡。然而，传统鳍条手指通常表现出有限的抗扭性能、不均匀的力分布以及减小的接触面积，导致滑移和果实表面的局部压力集中。为克服这些局限，我们将充气软膜垫集成到鳍条结构中，增大了有效手指宽度，并改善了表面贴合性和压力分布。我们建立了一个解析模型来预测膜变形和接触角随内部压力的变化关系，并通过有限元仿真和实验测试对结果进行了验证。压敏膜实验表明，Dragon Skin 30在30 kPa充气压力下，相较于未充气配置，接触面积增大了40%，平均接触压力降低了11%；而EcoFlex 00–30在压力下表现出体积膨胀，在测试范围内降低了贴合接触。性能评估在三种标准化叶片干扰条件（0%、50%和100%遮挡）下使用真实Pink Lady苹果叶片进行，并采用单因素方差分析（one-way ANOVA）和Tukey HSD事后比较进行统计分析。在30 kPa且无叶片干扰条件下，法向分离力相较于无膜垫配置增大了148%，扭转力矩相较于未充气条件增大了58%，同时保持了高于分离阈值的性能。在完全叶片遮挡条件下，无论充气压力如何，扭转性能均崩溃，表明完全树冠遮挡是抗扭性能的主要限制因素。这些结果证实，所提出的HFRF提供了压力自适应、抗滑移和减轻瘀伤的抓取性能，适用于非结构化果园环境中的机器人苹果采摘。","Computers and Electronics in Agriculture","2026-10-01T00:00:00Z",82,{"impact":76,"substance":164,"depth":76,"authority":165,"freshness":166,"relevant":52,"comment":167},23,14,9,"提出充气软垫混合Fin-Ray手指，量化提升接触面积与脱附力，对机器人采收有实质参考价值。",[169],{"name":160,"url":157},[83,21,171,172,173],"果园自动化","苹果采摘","柔性夹爪",{"code":25,"slug":26,"name":27,"path":28,"parent_code":29},[],{"status":32,"locked":14,"version":33,"method":34,"evidence":177},{"candidates":178,"agricultural_evidence":39},[179,186,190],{"code":25,"score":180,"supported":39,"title":181,"abstract":183,"tags":185},17,[182],"apple",[184,41,182],"苹果",[],{"code":45,"score":52,"supported":14,"title":187,"abstract":188,"tags":189},[],[],[21],{"code":93,"score":52,"supported":14,"title":191,"abstract":192,"tags":193},[],[],[83],[195,196],"苹果采摘 机器人 柔性夹爪","Fin-Ray 手指 苹果 采收","苹果采摘机器人柔性夹爪-3965","10.1016\u002Fj.compag.2026.112477",{"doi":198,"openalex_id":200,"authors":201,"venue":160,"cited_by_count":15,"oa_url":157,"card":211,"direction":215,"ingested_from":152},"W7215051321",[202,205,208],{"name":203,"orcid":204},"Ali Yasar","https:\u002F\u002Forcid.org\u002F0000-0003-0114-4471",{"name":206,"orcid":207},"Dilshan Ranasinghe","https:\u002F\u002Forcid.org\u002F0009-0000-5293-8584",{"name":209,"orcid":210},"Chao Chen","https:\u002F\u002Forcid.org\u002F0000-0002-1201-2039",{"tldr":212,"method":213,"finding":214,"direction":215,"opportunity":216},"设计并验证了一种集成充气软垫的混合Fin-Ray手指，用于苹果采摘时自适应、防损伤抓取。","在Fin-Ray手指中嵌入充气Dragon Skin 30硅胶膜垫，结合解析建模","30 kPa充气使接触面积增40%、压力降11%，无叶干扰时脱附力增148%，但全叶遮挡下扭转性能崩","农业人工智能与决策模型","可探索基于视觉的遮挡程度感知与充气压力自适应调控策略，以提升全遮挡下的扭转抗滑性能。","2026-10-02T23:30:01.662869Z",{"id":219,"title":220,"url":221,"summary":222,"summary_zh":223,"content":8,"source_name":224,"source_url":221,"published_at":225,"category":12,"cover_url":8,"hotness":13,"is_selected":14,"score":226,"score_detail":227,"sources":229,"tags":231,"primary_topic":234,"secondary_topics":235,"classification":236,"search_phrases":257,"slug":260,"view_count":46,"doi":261,"paper":262,"created_at":278},3780,"Epistemological Friction: Embodied Expertise and the ‘Last Mile’ of Agricultural Automation in the US Tree‐Fruit Industry","https:\u002F\u002Fdoi.org\u002F10.1111\u002Fntwe.70047","ABSTRACT Autonomous and AI‐driven agricultural technology is often framed as a panacea for labour dependency. While proponents envision autonomous technologies seamlessly optimising farming, this rationalised model misunderstands the reality of on‐farm integration. Based on an ethnographic and comparative study of 16 operations in the US West Coast Tree Fruit Industry (California and Washington), this article examines the ‘last mile’ of agricultural automation. The study reveals that integrating AI and robotics is a contested negotiation shaped by epistemological friction—the tension between the algorithmic logic of developers and the embodied, site‐specific expertise of farmers and farm workers. Rather than resulting in seamless labour replacement, we find that agricultural automation reconfigures work, creating an invisible reliance on ‘human middleware’ to navigate the biological recalcitrance of the orchard. Demonstrating how regional contexts and participatory implementation shape technological trajectories, this study extends Labour Process Theory. We argue for the need for a relational approach recognising the essential cognitive labour of farmers and farm workers that is demanded by digital agriculture. Ultimately, this study shows the ‘farm of the future’ depends on reconciling rigid systems with the situated expertise of the farm workforce.","自主与人工智能驱动的农业技术常被描绘为解决劳动力依赖问题的万能药。支持者设想自主技术能够无缝优化农业生产，然而这种理性化模型误解了农场整合的现实。本文基于对美国西海岸果树产业（加利福尼亚州和华盛顿州）16家经营主体的民族志比较研究，考察农业自动化的“最后一英里”。研究发现，人工智能与机器人技术的整合是一场充满争议的协商过程，其受到认识论摩擦的塑造——即开发者的算法逻辑与农民及农场工人具身的、场所特定的专业知识之间的张力。我们发现，农业自动化并未实现无缝的劳动力替代，而是重构了工作形态，制造了对“人类中间件”的隐性依赖，以应对果园的生物性阻力。本研究通过展示区域情境与参与式实施如何塑造技术轨迹，拓展了劳动过程理论。我们主张需要一种关系性方法，承认数字农业所要求的农民和农场工人不可或缺的认知劳动。最终，本研究表明“未来农场”取决于刚性系统与农场劳动力情境化专业知识之间的调和。","New Technology Work and Employment","2026-09-27T00:00:00Z",80,{"impact":76,"substance":77,"depth":76,"authority":165,"freshness":38,"relevant":52,"comment":228},"基于美西16家果园民族志研究，揭示AI与机器人落地的“最后一公里”认识论摩擦与“人类中间件”隐性依赖，对智慧农业人机协同具参考价值。",[230],{"name":224,"url":221},[232,83,233,21,171],"数字农业","农业人工智能",{"code":25,"slug":26,"name":27,"path":28,"parent_code":29},[],{"status":32,"locked":14,"version":33,"method":34,"evidence":237},{"candidates":238,"agricultural_evidence":39},[239,244,248,252],{"code":93,"score":38,"supported":39,"title":240,"abstract":241,"tags":243},[],[232,242],"digital agriculture",[232,83],{"code":25,"score":64,"supported":39,"title":245,"abstract":246,"tags":247},[],[27,41],[],{"code":45,"score":52,"supported":14,"title":249,"abstract":250,"tags":251},[],[],[21],{"code":253,"score":52,"supported":14,"title":254,"abstract":255,"tags":256},"I2",[],[],[233],[258,259],"美国 果树 自动化","果园 机器人 劳动力","美国果树自动化-3780","10.1111\u002Fntwe.70047",{"doi":261,"openalex_id":263,"authors":264,"venue":224,"cited_by_count":15,"oa_url":271,"card":272,"direction":150,"ingested_from":152},"W7214618881",[265,268],{"name":266,"orcid":267},"M. Anne Visser","https:\u002F\u002Forcid.org\u002F0000-0001-5047-7702",{"name":269,"orcid":270},"Alex Bladimir Hernandez","https:\u002F\u002Forcid.org\u002F0009-0001-5320-7203","https:\u002F\u002Fonlinelibrary.wiley.com\u002Fdoi\u002Fpdfdirect\u002F10.1111\u002Fntwe.70047",{"tldr":273,"method":274,"finding":275,"direction":276,"opportunity":277},"通过美西果树业民族志研究，揭示农业自动化落地中算法逻辑与农户经验知识的冲突。","对加州和华盛顿16个果园开展民族志与比较研究。","自动化未无缝替代劳动，反而依赖‘人类中间件’协调生物不确定性。","数字乡村与农业信息化","可探索参与式设计如何将农户隐性知识嵌入AI系统，弥合‘最后一英里’鸿沟。","2026-09-29T23:30:12.281947Z",{"id":280,"title":281,"url":282,"summary":283,"summary_zh":8,"content":8,"source_name":284,"source_url":8,"published_at":285,"category":12,"cover_url":8,"hotness":13,"is_selected":14,"score":226,"score_detail":286,"sources":288,"tags":290,"primary_topic":293,"secondary_topics":294,"classification":295,"search_phrases":311,"slug":314,"view_count":315,"doi":8,"paper":316,"created_at":324},3253,"Visuomotor Robotic Pruning in Planar Orchills Using Hybrid Reinforcement Learning（基于混合强化学习的平面果园视觉运动机器人修剪）","https:\u002F\u002Farxivtldr.org\u002Fabs\u002F2609.24906","arXiv发表研究：休眠期树木修剪是劳动密集型但对维持现代高产果园至关重要的工作。本文针对现代平面树形训练系统（V型棚架苹果、UFO樱桃）的修剪任务，提出端到端管线学习闭环视觉运动控制器。控制器完全使用仿真和合成生成数据进行训练，以零样本方式部署到真实果园。管线包括平面果园树网格的合成生成、基于物理的果园仿真器构建、通过运动规划自动收集成功修剪轨迹，以及结合离线演示与在线仿真推出相结合的新型混合强化学习算法。控制器使用腕部相机的光流输入，避免完整三维重建需求，连续引导切割器穿越杂乱分支环境到指定切割点并在真实果园38次物理试验中展示零样本仿真到真实迁移。在V-Trellis苹果上达到49.9%成功率，UFO樱桃上46.0%。","arXiv (Abhinav Jain, Cindy Grimm, Stefan Lee)","2026-09-20T00:00:00Z",{"impact":76,"substance":77,"depth":76,"authority":78,"freshness":166,"relevant":52,"comment":287},"仿真到真实零样本迁移的果园修剪机器人研究，方法新颖、数据扎实，对智慧果园机械化有参考价值。",[289],{"name":284,"url":282},[83,233,21,291,292],"强化学习","果园管理",{"code":25,"slug":26,"name":27,"path":28,"parent_code":29},[],{"status":32,"locked":14,"version":33,"method":34,"evidence":296},{"candidates":297,"agricultural_evidence":39},[298,302,307],{"code":25,"score":101,"supported":39,"title":299,"abstract":300,"tags":301},[],[184],[],{"code":253,"score":303,"supported":14,"title":304,"abstract":305,"tags":306},2,[],[],[233,291],{"code":93,"score":52,"supported":14,"title":308,"abstract":309,"tags":310},[],[],[83],[312,313],"V-Trellis 苹果 机器人修剪","UFO 樱桃 零样本迁移","V-Trellis苹果机器人修剪-3253",5,{"doi":8,"openalex_id":8,"authors":317,"venue":8,"cited_by_count":15,"oa_url":8,"card":318,"direction":215,"ingested_from":323},[],{"tldr":319,"method":320,"finding":321,"direction":215,"opportunity":322},"提出端到端视觉运动控制器，用仿真合成数据训练，零样本迁移到真实果园完成修剪。","合成数据生成、物理仿真器、运动规划收集轨迹、混合强化学习、腕部相机光流。","真实果园38次试验零样本迁移成功，苹果49.9%、樱桃46.0%成功率。","可探索多季节、多树种泛化及真实数据微调，提升复杂冠层下的鲁棒性。","agent","2026-09-23T00:04:33.854148Z",{"id":326,"title":327,"url":328,"summary":329,"summary_zh":330,"content":8,"source_name":160,"source_url":328,"published_at":331,"category":12,"cover_url":8,"hotness":13,"is_selected":14,"score":332,"score_detail":333,"sources":337,"tags":339,"primary_topic":340,"secondary_topics":341,"classification":342,"search_phrases":357,"slug":360,"view_count":101,"doi":361,"paper":362,"created_at":388},1590,"Semantic risk-aware motion planning for apple young-fruit thinning manipulators","https:\u002F\u002Fdoi.org\u002F10.1016\u002Fj.compag.2026.112374","Semantic risk-aware motion planning for apple young-fruit thinning manipulators。Computers and Electronics in Agriculture","苹果疏果机械臂的语义风险感知运动规划。《农业计算机与电子》","2026-09-03T00:00:00Z",61,{"impact":334,"substance":94,"depth":94,"authority":334,"freshness":335,"relevant":52,"comment":336},12,7,"针对苹果疏果机器人的语义风险感知运动规划研究，提升作业安全与效率，细分领域进展。",[338],{"name":160,"url":328},[83,21,292],{"code":25,"slug":26,"name":27,"path":28,"parent_code":29},[],{"status":32,"locked":14,"version":33,"method":34,"evidence":343},{"candidates":344,"agricultural_evidence":39},[345,349,353],{"code":25,"score":165,"supported":39,"title":346,"abstract":347,"tags":348},[182],[184,182],[],{"code":45,"score":52,"supported":14,"title":350,"abstract":351,"tags":352},[],[],[21],{"code":93,"score":52,"supported":14,"title":354,"abstract":355,"tags":356},[],[],[83],[358,359],"农业机器人 智慧农业 果园管理","农业机器人 智慧农业","农业机器人智慧农业果园管理-1590","10.1016\u002Fj.compag.2026.112374",{"doi":361,"openalex_id":363,"authors":364,"venue":160,"cited_by_count":15,"oa_url":382,"card":383,"direction":215,"ingested_from":152},"W7166343784",[365,367,369,371,373,375,377,380],{"name":366,"orcid":8},"谢胜旺",{"name":368,"orcid":8},"T B Zhang",{"name":370,"orcid":8},"Xiaoning He",{"name":372,"orcid":8},"Yanan Hu",{"name":374,"orcid":8},"Shuqi Shang",{"name":376,"orcid":8},"Z G Zhu",{"name":378,"orcid":379},"Xia Gao","https:\u002F\u002Forcid.org\u002F0000-0001-7139-7883",{"name":381,"orcid":8},"Shengda Li","https:\u002F\u002Fdoi.org\u002F10.2139\u002Fssrn.7009815",{"tldr":384,"method":385,"finding":386,"direction":215,"opportunity":387},"提出语义风险感知运动规划方法，用于苹果疏果机械臂，提高作业安全与效率。","结合语义信息与风险度量，优化机械臂运动规划算法。","该方法能有效降低机械臂作业风险，提升疏果精度。","可探索将语义风险感知扩展到其他精细农业操作，如采摘、修剪，或结合实时传感器数据。","2026-09-04T23:30:01.595949Z",{"id":390,"title":391,"url":392,"summary":393,"summary_zh":8,"content":394,"source_name":395,"source_url":8,"published_at":396,"category":12,"cover_url":8,"hotness":13,"is_selected":14,"score":397,"score_detail":398,"sources":401,"tags":403,"primary_topic":408,"secondary_topics":409,"classification":410,"search_phrases":431,"slug":434,"view_count":46,"doi":8,"paper":8,"created_at":435},742,"卫星—无人机两层组网葡萄园巡检与NDVI自适应路径规划(原标题:Satellite-Enabled Two-Tier UAV Vineyard Inspection with NDVI-Adaptive Path Planning)","https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1753","提出高空多光谱卫星+低空RGB无人机两层协同葡萄园巡检方案,基于NDVI兴趣点识别引导低空病检任务;引入LEO卫星通信模拟,带宽受限下路由优化可减少任务时间约15%,为丘陵果园精准植保提供端到端解决方案。","## 1. Introduction\n\nViticulture is a high-value agricultural sector with both economic and cultural significance in many regions. Vineyards are exposed to a range of pressures, including disease, water stress, and climatic variability. Efficient monitoring is essential for informing timely interventions and management decisions, while keeping labour costs and effort low. Automated and data-driven monitoring approaches offer an increasingly practical path to vineyard monitoring, combining sensing, communication, and analytics for decision support. Relevant technologies include satellite imagery and communications, drone mapping, multispectral vegetation indices, and computer vision.\n\nSatellite technologies can support crop monitoring both directly, through Earth observation, and indirectly, through communication. Satellite imagery can provide repeated, wide-area observations of vineyard condition and has been used to support precision viticulture by mapping vegetation vigour, water status, spatial variability, and other relevant indicators for improved sustainability and operational efficiency [[1](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1753#B1-agriculture-16-01753)]. As data-driven approaches gain traction in agriculture, satellite communication is also becoming increasingly relevant, especially for remote regions where terrestrial connectivity may be unreliable or unavailable [[2](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1753#B2-agriculture-16-01753)].\n\nDrones provide a flexible and high-resolution sensing platform for precision agriculture, complementing the wider coverage of satellite imagery with detailed observations for crop mapping, plant-health assessment, irrigation support, yield estimation, spraying, and other monitoring and management tasks [[3](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1753#B3-agriculture-16-01753)]. Drone-based multispectral mapping can provide vegetation indices, such as NDVI, which are widely used to quantify vegetation condition, growth, and stress [[4](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1753#B4-agriculture-16-01753)]. Previous vineyard-focused research has examined such data for specific applications including water-status variability mapping and disease progression monitoring [[5](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1753#B5-agriculture-16-01753),[6](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1753#B6-agriculture-16-01753)]. Nevertheless, UAV monitoring involves an inherent trade-off between coverage and detail: high-altitude flights can map larger areas efficiently, while low-altitude inspection can provide more detailed visual information but requires additional flight time and produces larger data volumes. At a broader monitoring scale, drone imagery is therefore often considered alongside satellite imagery, combining wide-area observation with finer spatial detail [[7](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1753#B7-agriculture-16-01753)].\n\nComputer vision and machine learning further expand the role of image-based crop monitoring by enabling automated interpretation of RGB, multispectral, and other data. In agriculture, these methods have been applied to a broad range of tasks, including plant and canopy segmentation, stress detection, disease identification, and yield estimation. For disease monitoring specifically, recent reviews highlight the growing use of UAV imagery, deep learning, and remote or proximal sensing tools for early detection and field-scale assessment [[8](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1753#B8-agriculture-16-01753),[9](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1753#B9-agriculture-16-01753),[10](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1753#B10-agriculture-16-01753)]. In grapevines, recent work includes both leaf-level disease classification using convolutional neural networks and field-level disease monitoring using UAV multispectral data [[6](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1753#B6-agriculture-16-01753),[11](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1753#B11-agriculture-16-01753)].\n\nBuilding on these developments, this work presents a satellite-enabled, two-tier UAV workflow for semi-automated vineyard inspection. The proposed approach combines high-altitude multispectral scanning, identification of points of interest from vegetation index analysis, live flight path adaptation, low-altitude RGB inspection, and vision-based disease detection, classification, and severity estimation. Processing tasks are offloaded to a remote server reached through an emulated satellite communication environment, allowing the UAV-side system to remain lightweight while receiving multispectral analysis results during the mission. These results are used to adapt the flight path and guide selective low-altitude inspection of areas that may indicate plant stress or disease.\n\nUnlike existing approaches that typically address multispectral crop monitoring, close-range disease inspection, communication constraints, or UAV path planning separately, this work integrates these elements into a single adaptive inspection workflow. The main contributions of this paper are as follows:\n\n*   A two-tier UAV inspection architecture combining high-altitude multispectral sensing, NDVI-based smart sampling, low-altitude RGB inspection, and remote image processing.\n\n*   On-the-fly path adaptation for inserting newly identified waypoints into an ongoing UAV mission, together with a comparative evaluation of alternative insertion and route-optimisation strategies.\n\n*   An experimental analysis of how satellite-link bandwidth affects image-transfer latency, waypoint availability, and total mission duration during adaptive inspection.\n\nThe experimental evaluation in [Section 3](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1753#sec3-agriculture-16-01753) focuses on the effect of different path-adaptation strategies and communication bottlenecks on mission efficiency. A detailed evaluation of point-of-interest identification as a smart sampling mechanism for low-altitude image acquisition is beyond the scope of this work and will be addressed in future research. The disease-detection component is included as a downstream analysis module based primarily on established computer vision approaches, and is therefore not benchmarked as a separate contribution in this paper.\n\n## 2. Materials and Methods\n\n### 2.1. System Architecture\n\nThe overall system architecture and information flow are shown in [Figure 1](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1753#fig_body_display_agriculture-16-01753-f001). The drone captures the images from the vineyard and sends them to the remote server for processing. The system incorporates high-altitude multispectral scanning, dynamic mission adaptation, and computer vision-based disease assessment. It is deployed as two interacting parts:\n\n**Figure 1.** System Architecture and data flow.\n\n1.\nA locally deployed application that interfaces with—and controls—the drone: For missions with a real drone, this is an Android app leveraging an SDK to communicate with the drone and deployed on the drone remote controller; for the tests presented in this work, a simulator was used instead, to facilitate repeat testing, as described in [Section 2.6](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1753#sec2dot6-agriculture-16-01753).\n\n2.\nA remote server that handles image processing and computer vision tasks: In this work, a server hosted in a VM within the satellite emulation testbed presented in [Section 2.2](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1753#sec2dot2-agriculture-16-01753) was used.\n\nFurthermore, the system comprises three main software modules, each with a distinct input and output, as shown in [Table 1](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1753#table_body_display_agriculture-16-01753-t001).\n\n**Table 1.** System software modules.\n\nThe flow of information between the modules, during the course of a mission, is as follows:\n\nInitially, a mission is created according to user input, comprising only high-altitude waypoints, placed evenly on a grid so as to cover the whole vineyard. At each high-altitude waypoint, the drone captures a multispectral image. The local application sends the pair of Red and NIR bands to the remote server. There, the POI identification module ([Section 2.3](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1753#sec2dot3-agriculture-16-01753)) receives and processes them, and sends back the coordinates of any POIs identified. The Path adaptation module ([Section 2.4](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1753#sec2dot4-agriculture-16-01753)) inserts these POIs into the existing path, adapting and optimizing it to minimise its total length. When a POI is visited, the drone captures a close-up RGB photo of the grapevine. All RGB close-ups are sent to the remote server after the mission for Disease detection ([Section 2.5](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1753#sec2dot5-agriculture-16-01753)).\n\nThe satellite constellation is transparent and serves the purpose of providing connectivity between the drone and the ground node hosting the remote server.\n\n### 2.2. Satellite Emulation Testbed\n\nThe satellite emulation testbed is capable of emulating a configurable number of satellites in an orbital plane. In this work, it is used to reproduce the behaviour of a Low Earth Orbit (LEO) satellite constellation with Inter-Satellite Links (ISLs), providing the communication environment between the drone-side system and the remote processing server.\n\nThe testbed was developed by DLR for research purposes using Mininet [[12](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1753#B12-agriculture-16-01753)], a network emulator that can create networks of virtual hosts, switches, controllers, and links. In contrast to purely analytical or discrete-event simulators, Mininet networks run real code, including standard Unix\u002FLinux network applications and the Linux kernel network stack. As a result, software developed and tested in Mininet can be transferred to real systems with minimal changes, making it suitable for network virtualisation, performance evaluation, and deployment-oriented testing.\n\nIn the emulated satellite network, ground nodes connect to the nearest visible satellite. Depending on the constellation configuration, communication between two ground nodes may traverse more than one satellite, making ISLs important for end-to-end connectivity. The performance of the communication path depends on the links between ground nodes and satellites, as well as on the ISLs between satellites. For this reason, the testbed supports multiple configurable parameters, including the following:\n\n*   The height of the orbital plane, which affects both the delay between a ground node and the nearest satellite and the duration of satellite visibility with respect to the ground node;\n\n*   The number of satellites between the source and destination ground nodes;\n\n*   Ground-to-satellite link parameters, such as bandwidth and packet loss rate;\n\n*   ISL parameters, including delay, bandwidth, and packet loss rate.\n\nAlthough the testbed can support two orbital planes, only one orbital plane was considered in this work for simplicity. The testbed can also emulate orbits other than LEO, provided that the relevant parameters are selected carefully to avoid unrealistic scenarios. Communication between nodes is based on Internet Protocol (IP) routing, and both Transmission Control Protocol (TCP) and User Datagram Protocol (UDP) are supported.\n\nThe experimental setup is shown in [Figure 2](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1753#fig_body_display_agriculture-16-01753-f002). The emulated satellite network runs inside a Virtual Machine (VM) on Ubuntu 22.04, which is accessed through a Virtual Private Network (VPN) connection. The drone-side system connects to this VPN and sends data to the VM. Upon receiving the data packets, the VM forwards them to the satellite emulation framework. Packets are then forwarded through the emulated satellite path, from the nearest satellite toward the gateway or destination ground node. In this work, end-to-end communication was carried out using TCP, and four satellites were modelled along the path between the drone and the ground node. The ground node hosts the Docker container running the remote processing service. After the images are processed, the results are returned through the same emulated communication path in the reverse direction.\n\n**Figure 2.** Experimental testbed configuration.\n\nSatellite emulation is considering a Starlink-like constellation. Hence, the ground nodes have continuous connectivity since at least one satellite is visible. This ensures that the ground nodes can always communicate without the need of a storage mechanism on the ground nodes. Considering that TCP is employed in the transport layer it is essential that the end-to-end connectivity is always present. Otherwise, the docker present on the other end of the constellation will not be readily reachable, which can adversely affect the overall performance.\n\nThe links between satellites and ground nodes, namely the drone and gateway, are modelled based on the orbital height of the satellite constellation and the relative movement between satellites and ground nodes. Link delay is highest when the satellite appears at the horizon of the ground node. It then decreases as the satellite moves toward the point above the ground node, before increasing again until the satellite is no longer visible. Satellite visibility duration also depends on orbital height, with higher orbital planes resulting in longer visibility duration. In the scenario considered in this work, the orbital height was set to 1221 km.\n\n### 2.3. POI Identification\n\n#### 2.3.1. POI Identification Rationale and Objective\n\nScanning an area, or a vineyard in the present case, is a trade-off between flight altitude and effective ground resolution. The ground area captured by a single photo (also known as the footprint) depends on the altitude of flight, as well as the camera’s field of view (FOV). For simplicity and uniformity, in this paper we are considering only photos taken at nadir orientation (i.e., the camera is pointing straight down).\n\nFor a camera with\n\n$F O V_{H} \\times F O V_{V}$\nfield of view and the drone flying at altitude A, the footprint is a rectangle of dimensions\n\n$X \\times Y$\n:\n\n$$\nX & = 2 A tan \\left(\\right. F O V_{H} \u002F 2 \\left.\\right) \\\\ Y & = 2 A tan \\left(\\right. F O V_{V} \u002F 2 \\left.\\right)\n$$\n\n(1)\n\nThe ground sampling distance (GSD) describes the distance between adjacent pixels and is equal to the footprint size divided by the camera resolution. For a camera with a\n\n$W \\times H$\nresolution:\n\n$$\nG S D_{x} & = 2 A \\frac{tan \\left(\\right. F O V_{H} \u002F 2 \\left.\\right)}{W} \\\\ G S D_{y} & = 2 A \\frac{tan \\left(\\right. F O V_{V} \u002F 2 \\left.\\right)}{H}\n$$\n\n(2)\n\nFor a given drone model, the camera specifications are constant and GSD is reduced to a function of altitude, which can be derived from either the horizontal or the vertical GSD and resolution, equivalently. For the camera of a modern commercial drone (in this work, experiments were conducted using the DJI Mavic 3M (DJI, Shenzhen, China)), with a vertical resolution of 3956 pixels and a corresponding vertical FOV of 48°, the altitude required to achieve a desired GSD is:\n\n$$\nA = \\frac{1}{2} \\frac{tan \\left(\\right. 48 \u002F 2 \\left.\\right)}{3956} G S D \\approx 4443 \\cdot G S D\n$$\n\n(3)\n\nGrapevine leaves average about 20 cm in size, while discolouration spots that can point to disease are often about 1 cm. In disease detection datasets such as [[13](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1753#B13-agriculture-16-01753),[14](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1753#B14-agriculture-16-01753),[15](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1753#B15-agriculture-16-01753)], discolouration spots that signify disease are usually 7–20 pixels wide. Hence, targeting a similar resolution and assuming that spots would be clearly identifiable if they are at least 10 pixels wide, at a GSD of no more than 0.001 m per pixel would be required, pointing at a maximum altitude of about 4.4 m for disease detection. This GSD is also in line with relevant research on unhealthy leaf detection in the wild [[16](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1753#B16-agriculture-16-01753)]. This corresponds to a footprint roughly 5.2 m × 3.9 m, or an area of 20 m 2, meaning that even a relatively small vineyard of 3–4 hectares would require thousands of photos to scan exhaustively.\n\nNon-exhaustive scans, or sampling, can provide insights regarding disease presence while keeping the number of images, the drone flight time, and the energy expenditure manageable. POI identification’s goal is to maximise the efficiency of low-altitude image sampling by identifying the most interesting, from a plant-health perspective, points in the vineyard.\n\n#### 2.3.2. Approach\n\nPrevious research has shown that NDVI bears a strong correlation with plant health and can be used to identify relevant risks [[6](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1753#B6-agriculture-16-01753),[17](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1753#B17-agriculture-16-01753)]. Crucially, NDVI-based insights do not rely on specific details but on the average value of a plant or vineyard area, and hence can be acquired with a greater GSD (e.g., 0.02–0.05 m per pixel), at higher altitudes. Accounting for the comparatively lower resolution of multispectral cameras, this corresponds to an altitude of 45–110 m.\n\nHence, NDVI images constructed from high-altitude multispectral photos can provide insights regarding plant health. In this work, such photos are used to drive a smart sampling scheme, in which spots are identified as points of interest (POIs) based on their NDVI value, and are marked for low-altitude image acquisition.\n\n#### 2.3.3. NDVI Calculation and Correction\n\nThe NDVI is computed from the Red and Near Infrared (NIR) bands:\n\n$$\nN D V I = \\frac{N I R - R e d}{N I R + R e d}\n$$\n\n(4)\n\nWhile the NIR and Red bands are acquired as part of the multispectral photo, they each have slightly different fields of view, due to the placement of the sensors on the drone. In addition, both images have slight vignetting distortion. Before computing the NDVI, band images must be undistorted and aligned:\n\nTo compensate for vignetting distortion, the vignetting polynomial coefficients are extracted from each image’s metadata and used. For alignment, Oriented FAST and Rotated BRIEF (ORB) [[18](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1753#B18-agriculture-16-01753)] is used to detect keypoints and compute compact binary descriptors in the Red and NIR images. ORB was selected because it combines low computational cost with robustness to the small rotational differences between the two bands, making it suitable for repeated processing of high-resolution multispectral images, while its binary descriptors can be matched efficiently using Hamming distance. A brute-force matcher compares the descriptors between the two bands, with cross-checking applied to retain only mutual best matches. The matches are sorted by distance, and the best 50 correspondences are used to estimate an affine transformation that aligns the NIR band to the Red band.\n\n#### 2.3.4. Grapevine Row Identification and Segmentation\n\nFrom each NDVI image depicting a patch of the vineyard, individual segments of plants, corresponding to potential POIs, must be identified and the mean NDVI of each must be calculated:\n\nFirstly, the NDVI images show rows of grapevines as well as soil. To separate the grapevines from the background, a binary mask is applied to NDVI with a threshold of 0.35. This threshold is in line with previous research, which suggests thresholds between 0.35 and 0.4, depending on plant development [[19](https:\u002F\u002Fwww.mdpi.com\u002F2077-0472\u002F16\u002F16\u002F1753#B19-agriculture-16-01753)]. Grapevine rows are identified as continuous areas in the binary mask. Finally, each row is divided into smaller segments. Segmentation is performed along each row’s long axis, by lines perpendicular ","MDPI Agriculture 16(16):1753 2026-08-15","2026-08-15T00:00:00Z",62,{"impact":334,"substance":76,"depth":399,"authority":334,"freshness":46,"relevant":52,"comment":400},16,"论文提出卫星-无人机两层组网巡检与NDVI自适应路径规划，方法新颖，数据详实，对智慧农业有参考价值，但时效性稍差。",[402],{"name":395,"url":392},[22,404,405,406,407],"智慧种植","葡萄园","NDVI","卫星-UAV",{"code":25,"slug":26,"name":27,"path":28,"parent_code":29},[],{"status":32,"locked":14,"version":33,"method":34,"evidence":411},{"candidates":412,"agricultural_evidence":39},[413,417,422,426],{"code":25,"score":334,"supported":39,"title":414,"abstract":415,"tags":416},[104],[104],[405],{"code":107,"score":101,"supported":39,"title":418,"abstract":419,"tags":421},[],[420],"植保",[],{"code":51,"score":52,"supported":14,"title":423,"abstract":424,"tags":425},[],[],[22],{"code":427,"score":52,"supported":14,"title":428,"abstract":429,"tags":430},"I3",[],[],[406],[432,433],"智慧种植 路径规划 葡萄园 卫星-UAV","智慧种植 路径规划","智慧种植路径规划葡萄园卫星-UAV-742","2026-08-19T23:57:32.281962Z"]