Search

A novel hybrid AI-driven optimization architecture is proposed to advance climbing perch cultivation in intelligent aquaculture environments, aiming for simultaneous gains in yield and cost performance. By merging Taguchi experimental design with reinforcement learning and metaheuristic strategies, the system pinpoints the key environmental and operational drivers of fish growth and economic return. A multi-objective regression technique is deployed to fine-tune crucial variables, namely water temperature, dissolved oxygen, pH, feeding regimes, and stocking density. The RL-AMIS approach, leveraging Pareto front evaluation and TOPSIS, effectively reconciles the tension between expenditure and output. Empirical outcomes document a 15.3% drop in operating expenses and a 17.8% boost in growth efficiency relative to traditional practice. These insights confirm that AI-based optimization can strengthen scalability and flexibility within aquaculture operations, thereby advancing sustainable production agendas.

This investigation puts forward a novel intelligent farming platform that merges Internet of Things (IoT) capabilities with predictive modeling and self-regulating command systems to streamline watering processes and amplify resource productivity. The suggested design employs soil wetness, thermal, and atmospheric moisture probes interfaced with an Arduino-centered microcontroller unit, allowing irrigation to be triggered autonomously according to live sensor readings. Agricultural operators can remotely oversee and adjust watering cycles through portable devices, delivering both exactness and operational simplicity. The predictive model fuses historical data with current measurements to anticipate future irrigation demands, guaranteeing moisture delivery exclusively when truly required. Empirical on-farm assessments indicated a 30 % decline in water expenditure versus standard irrigation techniques, whilst upholding ideal earth moisture parameters. In addition, the framework markedly diminished workforce and power expenses, bolstering overall functional capability. By confronting the issue of hydrological shortage and advancing conscientious resource stewardship, this intelligent platform possesses the capability to enhance harvest volumes, curtail greenhouse gas discharges, and further eco-conscious cultivation methodologies. Within water-limited geographies, deploying this technological solution constitutes a pivotal stride towards enduring agricultural viability, indicating considerable scope for extensive implementation.

The present investigation tackles several pressing difficulties in modern poultry operations, namely pathogen transmission, workforce deficits, inadequate surveillance capacity, and intensifying requirements for superior animal welfare standards. To meet these demands, an elaborate machine vision framework based on deep learning is put forward, wherein the YOLOv11 algorithm is deployed to autonomously observe and administer poultry. A collection of 330 chicken photographs was gathered from Thai agricultural facilities under a range of environmental circumstances and subsequently increased to 1,716 images by means of rotation, along with alterations to brightness and saturation parameters. Training of the YOLOv11 architecture was carried out on the NVIDIA Jetson Orin Nano Developer Kit, with performance gauged through precision, recall, and mean average precision (mAP) metrics. The architecture attained remarkably high fidelity, registering precision of 0.964, recall of 0.938, and mAP of 0.963, correspondingly. The fusion of optical with thermal imaging permitted the apparatus to circumvent limitations imposed by variable illumination and ambient settings. Outcomes from practical deployment evidenced real-time observation and tracing of numerous birds, thereby supporting disease mitigation, feed optimization, and holistic farm governance. This inquiry advances the domain of intelligent agriculture by furnishing a wholly automated, expandable, and exceedingly accurate poultry oversight infrastructure designed to foster ecological responsibility, operational productivity, and elevated animal welfare within contemporary poultry cultivation.

Interest in sustainable farming approaches has surged, especially within hyper-arid environments where acute water shortages make food security a pressing priority. This increased focus arises from the goal of agricultural independence in light of geopolitical volatility, rising population numbers, climate disruption, and the drive to decrease reliance on imported food. The present work investigates novel, data-centric methods specifically engineered for agriculture under such demanding circumstances, concentrating on the exploitation of CO₂ enrichment to amplify resource efficiency by accelerating plant development and fine-tuning input application within closed greenhouse structures. Within this study, a data-driven model predictive control (MPC) system is deployed in a greenhouse context to refine irrigation timing. The primary emphasis lies in harnessing the capability of an extreme gradient boosting (XGBoost) model to project real-time transpiration fluxes, accounting for the complex interdependencies between microclimatic variables and physiological shifts that influence transpiration. The XGBoost framework is designed to ingest microclimate observations—namely solar radiation, internal air temperature, internal relative humidity, and internal CO₂ molar fraction—as well as vegetation indices extracted from hyperspectral reflectance measurements, including NDVI, WBI, and PRI, all serving as forecasting inputs. The predictive model achieved strong fidelity, registering an R² of 97.1% alongside an RMSE of 0.417 mmol/m²/s for transpiration rate estimation. The trained XGBoost is subsequently embedded into the MPC scheme to steer irrigation delivery while upholding target soil water status. This integrated controller is then applied to govern watering tactics across two differing CO₂ supplementation settings: 400 ppm and 1000 ppm. Outcomes from this work reveal that the MPC-driven irrigation regime yields a 42.2% reduction in water use across a one-week forecast horizon when contrasted with the baseline irrigation protocol, irrespective of CO₂ level. Moreover, operating the MPC controller under the distinct CO₂ regimes shows a 34% decrease in water consumption at the 1000 ppm enrichment level relative to the 400 ppm condition. These findings highlight the value of MPC within enclosed greenhouse facilities, stressing the gains offered by advanced predictive analytics, multi-source data fusion, and continuous receding-horizon optimization for realizing superior irrigation regulation. The work equally accentuates the role of CO₂ fertilization in sealed agricultural greenhouses—particularly in zones subject to intense solar irradiance—as a potent measure for curbing water expenditure.

Precise categorization of avocado ripeness plays a vital role in strengthening post-harvest operations and curbing waste throughout agricultural supply networks. The present study centers on constructing a resilient ensemble classification framework employing spectral readings from 120 kilograms of Buccaneer avocados procured from the Royal Project in Chiang Mai, Thailand. The specimens were assessed via near-infrared (NIR) spectroscopy at 18 wavelengths. Five distinct machine learning models—Random Forest, Decision Tree, XGBoost, Gradient Boosting, and Gaussian Mixture Model—were separately trained and subsequently fused into a unified ensemble. Four distinct algorithms were deployed to fine-tune the weighting of each model: Bayesian optimisation, differential evolution, particle swarm optimisation, and grid search. Evaluation metrics covering accuracy, precision, recall, F1-score, confusion matrices, and ROC curves were applied. Grid Search yielded the strongest classification outcomes, securing an accuracy of 82.5% alongside an F1-score of 85.3%, emphasizing the value of weight-refined ensemble learning relative to standalone classifiers. This investigation delivers a scalable and transparent pipeline for non-destructive ripeness identification. The outcomes, notwithstanding constraints related to overfitting and dependency on spectral data integrity, pave the way for eventual real-time field deployment in agriculture.

Rice blast (RB) and brown spot (BS) represent two economically impactful diseases in rice cultivation, responsible for substantial annual yield reductions. Both pathogens infect the same host plant and produce lesions of remarkably similar appearance, leading to frequent misidentification by growers. Correct identification is vital to ensure effective disease control. Visual diagnosis requires trained personnel, while laboratory-based molecular assays, although precise, are costly and time-consuming. The present study examined whether lesions from these two diseases could be distinguished using proximally sensed digital RGB images and calculated colour indices. A smartphone camera was used to capture digital photographs of the lesions. A set of thirty-six colour indices was assessed via k-means clustering for disease separation, utilising three colour channel systems: RGB, HSV, and Lab. In summary, image backgrounds were masked so that only the leaf spot lesion remained, and colour indices were then computed as features from the lesion’s central portion, reflecting the diagnostic practice commonly employed by plant pathologists. The outcomes indicated that the complete set of 36 indices separated the two diseases with an accuracy of 84.3%. It was also noted, however, that accuracy was largely driven by indices related to the R, G, and B profiles, with the remaining indices contributing little. The indices G/R, NGRDI, (R + G + B)/R, VARI, (G + B)/R, R/G, Nor_r, G-R, Mean_A, and Logsig were singled out as providing the greatest contribution to disease discrimination. Accordingly, these specific RGB-derived colour indices can be applied alongside the k-means algorithm to separate blast and brown spot diseases. The study’s outcomes deliver a non-destructive, alternative objective means of recognising RB and BS disease symptoms. On the strength of these results, a mobile application named Blast O spot was created to perform field-level disease differentiation.

Achieving sustainability in agriculture requires resolving multiple optimization challenges. Among these, the harvest scheduling problem stands out. This paper explores scheduling tea harvests. The problem is to devise a harvest timetable that accounts for farmer quotas while respecting the constraints of both the collection center and the processing plant. The tea harvest is a cooperative effort between the grower and the factory. Factory administration is focused on resource utilization. Consequently, the process must incorporate the factory’s capacity, the collection centers’ capacities, and the frequency of transport trips. From the grower’s standpoint, it is observed that farming is often not their sole profession. On harvest days, growers are frequently absent from their main occupations. Integrating their preferred harvest days into the schedule is critical to the longevity of agricultural practices. Two separate mathematical formulations are introduced to tackle this problem. The initial formulation reduces the weekly transport frequency of factory vehicles, operating within the bounds of factory and collection center capacities. The subsequent formulation both curtails transport frequency and endeavors to satisfy grower preferences to the greatest extent feasible. A practical implementation was conducted in an area encompassing 12 collection centers, 988 growers, and 3392 decares of tea plantations. The outcomes demonstrate that the alignment rate with grower harvest preferences could be increased from 52% to 97% without altering the factory’s transport frequency. This finding reveals that accommodating grower preferences on harvest day imposes no negative consequences on the factory. Instead, it was determined that this practice bolsters both sustainability and incentive within agriculture. Additionally, the findings confirm the models’ effectiveness in resolving the problem.

Achieving sustainability and efficiency in maintaining high crop yields while minimizing agricultural chemical pollution requires precise weed mapping to fully deploy site-specific weed management—a persistent, major hurdle in contemporary farming. This investigation tested the resilience of training epochs for the You Only Look Once (YOLO) v5s convolutional neural network (CNN) model in building an automated system for classifying crops and weeds from UAV-acquired imagery. The images were labeled with bounding boxes and trained on Google Colab across 100, 300, 500, 600, 700, and 1000 epochs. The network identified and sorted five distinct categories: sugarcane (Saccharum officinarum), banana trees (Musa), spinach (Spinacia oleracea), pepper (Capsicum), and weeds. Training proceeded over multiple epochs to pinpoint the peak test-set performance and was terminated once the test metrics (overall accuracy, precision, and recall) began to decline. The findings reveal a marked improvement in classifier capability as the epoch count climbed from 100 to 600. However, a modest reduction surfaced at 700 epochs, where overall accuracy, weed precision, and recall registered 65, 43, and 43%, respectively, contrasted with the 67, 78, and 34% achieved on these same metrics at 600 epochs. This erosion continued at 1000 epochs, producing an overall accuracy of 65%, a weak precision of 45%, and a recall of 40%. These outcomes confirm that the training epoch length in YOLOv5s substantially governs the model’s reliability for automated crop and weed discrimination, singling out 600 epochs as the optimal performance point.

In some Asian countries, jute is a fundamental source of revenue and contributes to Gross Domestic Product (GDP). Like other crops, jute is vulnerable to pest invasions, and detecting them often relies on visual inspection in countries such as Bangladesh, India, Myanmar, and China. This practice is also slow, difficult, and somewhat inaccurate, posing a significant economic risk. To mitigate this challenge, the investigation proposes a highly accurate and resilient transfer learning (TL)- based JutePestDetect model for detecting jute pests at their initial stages. To begin with, we assembled a jute pest collection comprising 17 categories and roughly 380 images per category, which were scrutinized after manual and automated preprocessing and refinement, including background removal and scaling. Following this, five distinguished pre-trained architectures—DenseNet201, InceptionV3, MobileNetV2, VGG19, and ResNet50—were picked from an earlier work to formulate the JutePestDetect model. Every architecture was adjusted by replacing the classification layer with a global average pooling layer and inserting a dropout layer for regularization. For performance measurement, a range of indicators, including precision, recall, F1 score, ROC curve, and confusion matrix, were used. These examinations provided a deeper understanding of the models’ capabilities. Within this set, the proposed JutePestDetect model, based on a tailored regularized DenseNet201, outperformed its counterparts, achieving an outstanding accuracy of 99%. Hence, our suggested method and tactics deliver a refined solution for pest recognition specific to jute, with the potential to substantially assist cultivators worldwide.

This paper delivers an interview-based investigation and literature analysis examining the potential benefits and barriers to adopting Artificial Intelligence (AI) in farming enterprises. Three branches of agriculture are examined: crop cultivation, dairy farming, and beef farming. As groundwork for the study, a literature analysis re-evaluates earlier work on intelligent farming. Subsequently, an interview study with 21 participants examines receptivity and readiness for intelligent farming tools and provides technical detail on three selected AI applications in agriculture. Though centered on circumstances in Sweden, the results apply to the wider European agricultural context and beyond. The study uncovers a broad range of elements that both promote and hinder the transition. Chief among the recognized opportunities are the strategic weight intelligent farming carries on the agendas of numerous sector players, the overall shift toward software-as-a-service models through co-owned equipment, the enormous stockpiles of existing data, and the substantial eagerness among farmers for novel technology. Conversely, the study highlights primary obstacles, including technical and legal difficulties related to data ownership, looming cybersecurity threats, the need for a clearly articulated economic rationale, and the occasional lack of technical competence within the sector. The paper finds that the overarching direction signals a shift toward intelligent farming, but the pace of change will depend on how well the identified barriers are resolved.

Artificial intelligence is increasingly positioned as a transformative force for climate-smart agriculture, promising simultaneous gains in productivity, efficiency, and sustainability. Across agricultural policy, innovation systems, and digital farming markets, AI is presented as a solution to climate uncertainty, resource scarcity, labour pressures, and food-security demands. This review examines that promise through a critical lens. The dominant narrative surrounding AI-enabled climate-smart agriculture often treats optimization as a neutral technical process. Yet algorithmic systems do not merely improve farming decisions; they also define what counts as efficiency, which outcomes are prioritised, and whose knowledge becomes authoritative. This raises important questions about power, agency, and sustainability. The objective of this critical review is to interrogate the politics of AI in climate-smart agriculture. It synthesises evidence on the tensions between productivity, efficiency, sustainability, farmer agency, data sovereignty, and governance. The review asks how algorithmic optimization reshapes agricultural futures and who benefits from this transformation. The review draws on  peer-reviewed articles published across agricultural innovation, sustainability science, digital agriculture, artificial intelligence, and critical data studies. Rather than reporting new empirical data, it develops a critical synthesis of existing debates and identifies conceptual gaps in current scholarship. The review finds that algorithmic farming is not simply a technical evolution of precision agriculture. It is a contested political terrain in which sustainability goals, commercial interests, data infrastructures, and farmer autonomy are negotiated. A critical research agenda is needed to ensure that AI contributes to equitable, democratic, and ecologically resilient agricultural futures.

Climate stress is reshaping the operational conditions under which agricultural decisions are made. Drought, heat, flooding, salinity, and increasingly variable weather patterns intensify uncertainty around irrigation, fertilisation, pest management, crop selection, and yield planning. Precision agriculture has therefore become a central domain in which artificial intelligence is expected to support adaptive, data-driven decision-making. This systematic review examines how artificial intelligence decision models are used in precision agriculture under climate stress. The review focuses on three connected dimensions: the model families used to support decisions, the data sources on which these models depend, and the sustainability outcomes that are reported. The objective is to clarify whether current evidence supports the claim that artificial intelligence can deliver more resilient and sustainable agricultural management. The review synthesises peer-reviewed journal articles published. The study selection followed a PRISMA-oriented logic, using predefined search strings, inclusion criteria, and screening stages to identify relevant work on machine learning, deep learning, remote sensing, sensor-based agriculture, irrigation optimisation, yield prediction, crop disease detection, and sustainability assessment. No new empirical field data were generated. The evidence base shows rapid expansion in machine learning and deep learning applications, especially for yield prediction, irrigation management, crop monitoring, plant disease detection, and smart farming decision support. Satellite imagery, UAV data, IoT sensors, weather variables, soil information, and farm-management records appear as the dominant data sources. However, the review also identifies uneven attention to climate stress conditions, limited integration of long-term climate projections, and inconsistent reporting of sustainability outcomes.The review concludes that artificial intelligence for precision agriculture under climate stress is technically promising but methodologically fragmented. Most studies emphasise model accuracy, yield improvement, or input optimisation, while fewer evaluate environmental trade-offs, social impacts, farmer livelihoods, or long-term resilience. A more rigorous, transparent, and sustainability-centred research agenda is needed to move the field beyond predictive performance toward climate-resilient agricultural decision systems.

Digital twin technology has moved from industrial cyber-physical systems into agriculture as part of the wider shift toward intelligent, data-intensive farming. In farming contexts, digital twins are expected to connect physical production systems with virtual models that can monitor, simulate, and support decisions in near real time. Yet the agricultural literature remains fragmented across crop, soil, water, livestock, sensor, modelling, and innovation studies. This scoping review maps the emerging evidence on digital twins in intelligent farming systems. Its objective is to classify the applications, architectures, and translation barriers reported in peer-reviewed research. The review focuses on how digital twin concepts are adapted to agriculture rather than treating the term as a generic label for digital monitoring. The review follows a scoping logic to chart the breadth of the field rather than evaluate intervention effectiveness. A total of 29 peer-reviewed articles were used to map conceptual definitions, technical architectures, agricultural applications, and implementation constraints. The evidence was organised around three guiding questions: what architectures are proposed, what agricultural domains are covered, and what barriers restrict translation. The mapped literature shows that agricultural digital twins are most often framed around IoT-enabled sensing, cloud or edge data processing, simulation models, decision dashboards, and feedback loops. Applications are concentrated in crop management, greenhouse production, irrigation, autonomous machinery, and livestock monitoring, while soil-system and whole-farm twins remain less mature. The literature also reveals persistent barriers related to data interoperability, connectivity, cost, skills, governance, standards, and uncertain return on investment. The review concludes that digital twins hold significant promise for intelligent farming but remain concentrated at the prototype and demonstration stages. Moving from research concepts to practical farm systems requires coordinated progress in data infrastructure, modular architectures, validation studies, standards, and farmer-centred implementation models. The review provides a structured map for future digital agriculture research and practice.

Agri-food supply chains are increasingly exposed to overlapping disruptions arising from climate variability, geopolitical instability, pandemic effects, price volatility, labour shortages, and changing consumer demand. These pressures have intensified interest in resilience-oriented technologies that can help supply chains anticipate disruption, absorb shocks, adapt operations, and recover service continuity. Artificial intelligence has emerged as a prominent candidate because it can transform fragmented supply chain data into forecasts, alerts, recommendations, and operational decisions. This systematic review examines how artificial intelligence is being applied to strengthen agri-food supply chain resilience. The review focuses on four functional domains: forecasting, traceability, risk detection, and decision support. It asks whether current AI applications merely improve isolated technical tasks or whether they provide demonstrable evidence of resilience improvement across agri-food supply systems. The review follows a systematic design informed by PRISMA reporting principles and structured literature review procedures. A corpus of 32 peer-reviewed journal articles published between 2017 and 2025 was synthesised. The selected studies cover AI forecasting models, blockchain-enabled traceability, quality and safety monitoring, logistics optimisation, supply chain decision support, and review-methodology guidance. The synthesis shows that forecasting research is technically advanced, especially in price prediction, demand forecasting, and quality-sensitive perishable supply chain management. Traceability and risk detection studies show strong growth around blockchain, IoT, big data, computer vision, and food safety monitoring. Decision support research connects AI to resilience more directly, but most evidence remains conceptual, model-based, or function-specific rather than validated through real disruption events. The review concludes that AI has substantial potential to support agri-food supply chain resilience, but the evidence base remains fragmented. Existing studies tend to focus on single functions such as forecasting accuracy, provenance visibility, or routing efficiency, while system-level resilience outcomes are rarely measured. Future research should therefore move from isolated AI applications toward integrated, empirically validated, resilience-focused agri-food intelligence systems.

The digitalisation of agri-food networks is increasingly presented as a pathway toward more intelligent, efficient, transparent, and resilient food systems. In this narrative, artificial intelligence functions as the analytical brain of the food system, converting dispersed data from farms, supply chains, markets, and consumers into actionable intelligence. This promise has become central to contemporary visions of smart farming, precision logistics, demand forecasting, and waste reduction. Yet the language of food system intelligence often conceals a more troubling political-economic transformation. The infrastructures that collect, process, and monetise agri-food data are frequently controlled by proprietary platforms, agribusiness firms, technology corporations, and data intermediaries. As a result, digitalisation may expand not only analytical capacity but also platform dependency, data concentration, and asymmetrical control over decision-making. This critical review interrogates the tension between food system intelligence and platform dependency in digital agri-food networks. It examines how AI, data infrastructures, and platform governance reshape farmer autonomy, market structure, and public authority. The review draws on critical data studies, political economy, agri-food governance, and platform studies to question the assumption that more data automatically produces more democratic or sustainable food systems. The review concludes that the central policy question is not whether AI should be used in agri-food systems, but who governs it, who benefits from it, and whose knowledge counts within it. Without deliberate governance intervention, digital agri-food networks risk becoming extractive data monopolies rather than intelligent public goods. A more democratic digital food future requires data sovereignty, farmer-centred infrastructures, public-interest governance, and stronger collective control over agri-food data systems.

Ensuring food quality, safety, and authenticity has become a central challenge for contemporary food systems because risks now emerge across production, processing, distribution, retail, and consumption. Artificial intelligence is increasingly used to interpret complex food data, including spectral signals, images, sensor outputs, molecular profiles, traceability records, and supply-chain alerts. These developments suggest that AI can support a shift from reactive food inspection toward predictive and preventive food integrity assurance. Despite this promise, AI applications in food quality, safety, and authenticity often remain separated across disciplinary and technological domains. Sensing studies focus on detection accuracy, prediction studies focus on model performance, traceability studies focus on chain-of-custody infrastructure, and explainable AI studies focus on interpretability. This fragmentation limits the field’s ability to understand how these components could function together as an integrated assurance system. This integrative review synthesises 31 peer-reviewed articles published between 2017 and 2025 across food science, analytical chemistry, sensing technology, food informatics, and artificial intelligence. The review examines four interconnected domains: sensing technologies and data sources, prediction models for quality and safety, traceability and authenticity verification, and explainable decision systems. Its purpose is not to report new empirical data, but to integrate diverse research streams into a coherent conceptual understanding of AI-enabled food integrity assurance. The review shows that AI has already demonstrated value in optical sensing, spectroscopy, imaging, authenticity testing, shelf-life prediction, risk assessment, fraud detection, and traceability analysis. However, most applications are still designed as isolated models rather than as connected decision systems. Three tables classify sensing technologies, categorise prediction models, and present an integrative framework linking detection, prediction, provenance, and explainable decisions. The central conclusion is that the technical components of AI-enabled food integrity assurance already exist, but they remain insufficiently integrated. Future research should move beyond stand-alone classification or prediction models and toward interoperable, explainable, and operationally validated systems. Such systems should connect sensing, prediction, traceability, and human decision-making across food supply chains.

Generative artificial intelligence is rapidly entering the food sector as a family of technologies capable of producing text, images, formulations, molecular structures, process scenarios, and synthetic training data. Its relevance to food manufacturing and product development lies in its capacity to accelerate experimentation, support creativity, and extend data-driven decision-making beyond conventional prediction. Yet the evidence base remains early, fragmented, and unevenly validated. This rapid evidence review maps the emerging literature on Generative AI in food manufacturing and product development from 2017 to 2025. The objective is to identify where applications are being explored, what opportunities are being claimed, and which risks and governance needs are becoming visible. The review focuses on Generative AI as a food system technology rather than as a general digital innovation. A rapid evidence review approach was used to synthesise 29 peer-reviewed articles relevant to Generative AI, food manufacturing, food quality control, product design, reformulation, digitalisation, food safety, and rapid review methodology. The review does not present new empirical data. Instead, it offers an accelerated mapping of a fast-moving field to inform research, practice, and policy. The evidence indicates that Generative AI is most visible in synthetic image generation, large language model support for food knowledge extraction, recipe and formulation assistance, hyperspectral data augmentation, and broader Industry 4.0 food manufacturing contexts. However, much of the literature remains conceptual, proof-of-concept, or adjacent to Generative AI rather than demonstrating mature industrial deployment. Risks are concentrated around validation, hallucinated outputs, safety claims, intellectual property, biological plausibility, and weak governance alignment. The review concludes that Generative AI has significant transformative potential in food manufacturing and product development, but this potential is conditional on rigorous validation and anticipatory governance. Without domain-specific standards, human oversight, transparent documentation, and regulatory engagement, Generative AI could amplify uncertainty rather than improve food system performance. A proactive research and practice agenda is therefore essential.

Artificial intelligence is increasingly used across agriculture and food systems for precision farming, yield prediction, pest detection, supply-chain optimisation, quality control, decision support, and resource management. These applications are often promoted as tools for increasing productivity, reducing uncertainty, improving resource efficiency, and strengthening food-system resilience. As AI becomes embedded in farms, supply chains, laboratories, and food governance infrastructures, its consequences extend beyond technical performance. The central problem is that the benefits of agri-food AI are frequently assessed through accuracy, automation efficiency, profitability, or optimisation metrics, while broader ethical, equity, environmental, and governance implications receive less systematic attention. AI systems may alter farmer autonomy, labour relations, data ownership, market participation, environmental decision-making, and institutional accountability. Treating these consequences as secondary risks understates the degree to which AI changes relationships among people, technologies, land, food, and power. This critical review examines responsible AI in agriculture and food systems through four connected dimensions: ethics, equity, sustainability, and governance. It argues that responsible AI cannot be reduced to technical reliability or model performance because agri-food systems are socially embedded, ecologically consequential, and institutionally fragmented. The article therefore evaluates responsible AI as a socio-technical and governance challenge rather than as a purely computational problem. The article contributes a critical argument that responsible agri-food AI requires accountable design, inclusive access, sustainability-aware evaluation, and multi-level governance. It emphasises that trustworthy AI in agriculture and food systems depends on who controls data, who benefits from automation, who bears environmental and social risks, and who participates in decision-making. Responsible AI is therefore positioned as a core condition for equitable, sustainable, and publicly legitimate agri-food transformation.

Machine learning is increasingly applied across agri-food systems to support prediction, classification, optimisation, monitoring, and decision-making. These applications now extend from farm-level production to processing, logistics, consumption, and food-waste management. As data-intensive agriculture expands, machine learning is often presented as an enabling technology for sustainable agri-food transitions. Despite this promise, the sustainability evidence base remains fragmented. Many studies report improvements in model accuracy, prediction performance, or operational decision support, but they do not always specify how these technical gains translate into environmental, economic, or social outcomes. This creates uncertainty about whether machine learning contributes to sustainability transitions directly, indirectly, or only under particular implementation conditions. This systematic review examines how machine learning contributes to sustainable agri-food transitions through environmental, economic, and social impact pathways. It focuses on the relationship between machine learning methods, agri-food applications, and sustainability-relevant outcomes rather than treating model performance as an end in itself. The review asks where evidence is strongest, where it remains indirect, and how different pathways vary across production, processing, distribution, consumption, waste reduction, and governance. The review finds that environmental and economic pathways are more developed than social and equity pathways. Evidence is strongest where machine learning supports precision input management, yield forecasting, irrigation decision support, operational efficiency, supply-chain analytics, and food-waste reduction. However, stronger longitudinal evaluation, causal assessment, life-cycle integration, participatory design, and equity-sensitive governance are needed before machine learning can be judged as a reliable driver of sustainable agri-food transitions.

Foundation models and multimodal artificial intelligence are beginning to reshape agricultural AI by enabling systems that can integrate images, text, sensor streams, remote-sensing data, environmental records, and management information. These approaches differ from earlier narrow models because they can reuse learned representations across related tasks and support richer forms of agricultural interpretation. Their potential relevance spans crop monitoring, disease detection, advisory services, automation, forecasting, and agri-food decision support. Despite this promise, the transfer of large-scale AI systems into agriculture remains complex. Agricultural environments are biologically variable, spatially heterogeneous, seasonally dynamic, and shaped by management decisions that are often poorly captured in available datasets. As a result, questions remain about data quality, annotation adequacy, domain adaptation, validation, explainability, governance, and deployment risk. This scoping review maps emerging applications, data requirements, and deployment risks associated with foundation models and multimodal AI in agriculture and agri-food systems. It focuses on how the reviewed literature frames model architectures, data modalities, application domains, validation needs, and governance concerns. The purpose is not to estimate pooled effectiveness but to chart the evidence landscape and clarify where knowledge is advancing or remains unsettled. The review uses a scoping approach to organise literature around key concepts, model types, data inputs, agricultural tasks, implementation settings, and responsible deployment concerns. It charts relationships among remote-sensing foundation models, vision-language systems, large language models, generative AI, multimodal data fusion, benchmark datasets, and agricultural decision-support systems. It also identifies gaps in field validation, interoperability, low-resource contexts, participatory governance, and trustworthy deployment. The review contributes a structured synthesis of how foundation models and multimodal AI may extend agricultural sensing, prediction, automation, and decision support. It highlights that technical capability alone is insufficient for responsible agricultural use. Robust data infrastructures, context-aware adaptation, transparent validation, and accountable governance will be central to whether these systems support sustainable and equitable agri-food transformation.

Digital agriculture technologies are increasingly positioned as tools for strengthening food security through improved sensing, prediction, automation, decision support, and resource management. These technologies are promoted as ways to help agricultural systems respond to climate stress, resource constraints, production volatility, and disruptions across agri-food value chains. Their potential relevance extends from farm-level monitoring to broader food-system coordination. Evidence on digital agriculture is distributed across multiple technology domains, including artificial intelligence, Internet of Things systems, robotics, remote sensing, and decision-support platforms. Food-security implications are often reported indirectly through productivity, efficiency, sustainability, resilience, risk reduction, or decision-quality outcomes. This dispersion makes it difficult to judge where the evidence is strongest and where food-security claims remain underdeveloped. This umbrella review synthesises review-level evidence on digital agriculture technologies in relation to food-security pathways. It examines how the reviewed literature connects digital agriculture with food availability, access, utilisation, stability, resilience, sustainability, productivity, and agri-food decision-making. The article focuses on review evidence rather than primary empirical findings. The approach compares review-level evidence across technology domains, outcome pathways, implementation conditions, and evidence gaps. It gives particular attention to whether digital agriculture evidence demonstrates direct food-security relevance or supports more indirect pathways through resource efficiency, monitoring capacity, operational timeliness, or improved decision support. It also distinguishes technical potential from demonstrated system-level benefit. The contribution of this umbrella review is to clarify where digital agriculture evidence is most developed, where food-security links remain indirect, and what future research is needed to evaluate digital agriculture as a contributor to resilient and inclusive food systems. The synthesis shows that digital agriculture should be assessed not only by technical performance but also by its capacity to support equitable, sustainable, and context-sensitive food-security outcomes.

Intelligent agriculture is increasingly shaped by automation, data platforms, decision-support tools, robotics, sensors, and algorithmic advisory systems. These technologies are promoted as ways to improve productivity, precision, efficiency, and sustainability in farming. Yet they also alter how agricultural decisions are made, who participates in those decisions, and whose knowledge is treated as authoritative. The central problem addressed in this article is that intelligent agriculture is often presented as a technical improvement while its implications for farmer agency remain contested. Farmer autonomy, judgement, dependency, skill, and voice can be strengthened by digital tools, but they can also be narrowed when systems become opaque, prescriptive, proprietary, or difficult to contest. This tension makes farmer agency a core question for responsible intelligent agriculture. This critical interpretive review examines how automation, platformization, and participatory design reshape farmer agency. It interprets agency not as an individual trait that farmers either possess or lose, but as a socio-technical capacity formed through relations among farmers, machines, platforms, data infrastructures, advisors, markets, and governance arrangements. The review therefore focuses on how agency is enabled, constrained, redistributed, and contested. The article examines key concepts, tensions, and assumptions in the reviewed literature. It does not report new primary empirical data, but critically synthesises existing scholarship to clarify how intelligent agriculture reorganises decision-making, knowledge, dependency, participation, power, and accountability. Particular attention is given to tensions between efficiency and autonomy, automation and situated judgement, platform convenience and dependency, data-driven advice and local knowledge, and participation and tokenism. The review argues that farmer agency is redistributed through socio-technical arrangements rather than simply preserved or removed by technology adoption. Responsible intelligent agriculture requires accountable automation, fair platform governance, participatory design, recognition of local knowledge, and institutional mechanisms that give farmers meaningful voice. Farmer agency should therefore be treated as a condition of technological legitimacy, not as an afterthought to innovation.