Controlled Environment Agriculture (CEA) demands precise, adaptive climate management across distributed infrastructure. This paper presents IOGRUCloud, a scalable three-tier IoT platform that integrates AI-driven control with edge computing for automated greenhouse climate regulation. The system architecture separates field-level sensing and actuation (L1), facility-level coordination (L2), and cloud-level optimization (L3-L4), enabling progressive autonomy from rule-based to fully autonomous operation. A Vapor Pressure Deficit (VPD) cascading control loop governs temperature and humidity with GRU-enhanced PID tuning, reducing manual calibration effort by 73%. Deployed across 14 production greenhouses totaling 47,000 m2, the platform demonstrates 23% reduction in energy consumption and 31% improvement in climate stability versus baseline. The system handles 2.3M daily sensor events with 99.7% uptime. We release the architecture specification and deployment results to support reproducibility in smart agriculture research.
China boasts abundant cultivated resources of pitahaya, with Guizhou Province being one of its core producing areas. Quality differences in red-fleshed pitahaya among local producing areas have not been fully clarified, and a standardized quantitative evaluation system for these differences remains lacking. This study seeks to identify the key factors influencing regional variations in quality and establish a comprehensive evaluation standard. In this study, 15 samples of red-fleshed pitahaya were collected from four major producing areas in Guizhou and used as research materials. Based on 15 quality characteristic indicators of the fruits, an analysis of quality differences and establishment of an evaluation system were carried out using multivariate statistical analysis. The results showed that 14 of the 15 quality indicators exhibited significant differences among pitahaya samples from different producing areas (<i>p</i> < 0.05), with the a* value being the sole exception. Cluster analysis classified the 15 samples into four groups. Principal component analysis (PCA) extracted four principal components, with a cumulative variance contribution rate of 81.07%, which clearly identified betacyanin, betaxanthin, 1,1-diphenyl-2-picrylhydrazyl (DPPH) free-radical scavenging rate, vitamin C, fruit shape index, and transverse diameter as the core evaluation indicators. This study systematically clarifies the differences in quality characteristics and the internal correlations among quality indicators of red-fleshed pitahaya from different major producing areas in Guizhou. It further provides an important scientific basis for pitahaya variety breeding, cultivation regulation, and market positioning in this region and fills the research gap existing in the field of comprehensive quality evaluation of pitahaya. This is of significant practical importance for promoting the standardized upgrading of local specialty fruit industries, enhancing the market competitiveness of products, and facilitating the high-quality development of the agricultural economy.
Teresa M. Anderson, Kirsty J. Verhoek, Brian T. Dela Rue
et al.
Many regions across New Zealand experience cold, wet winters with low pasture growth. Consequently, farmers often rely on winter forage crops to feed cows, resulting in challenges with animal welfare, environmental damage and operational difficulties. Despite the potential of off‐paddock facilities to overcome negative outcomes of crop wintering, only 15% of New Zealand dairy farmers have off‐paddock facilities. This study aimed to document farmers’ experiences when planning, building and using off‐paddock infrastructure and involved interviews with five dairy farmers who had recently built infrastructure. Early, strategic planning to include sufficient time for thorough consent preparation, establishing values, visiting farms with infrastructure, financial planning and alignment with farm operations was essential for success. Siting considerations, such as visual appearance and proximity to housing, were important for community acceptance. Farmers should consider appropriate long‐term supply of feed and bedding materials, understand effluent management and allow for flexibility in use. Finally, selecting a reputable project team with strong communication and problem‐solving skills was vital. The findings highlight the importance of comprehensive planning and stakeholder involvement from the outset. Insights from the project can help farmers “build it once and build it right,” avoiding costly mistakes with financial, environmental, social and animal welfare implications.
Kate E. Fransen, Sarah M. Gard, Ina Pinxterhuis
et al.
ABSTRACT In their review: An examination of the ability of plantain (Plantago lanceolata L. ) to mitigate nitrogen leaching from pasture systems , Eady et al. (2024) dispute both the historic estimates of typical urine patch nitrogen (N) load and leaching and the evidence for the N leaching reduction mechanisms of plantain; and question the recommended levels of plantain required to achieve N leaching reduction. We reject the suggestion that the urine patch has little influence on N leaching, and that average annual N leaching from dairy farms is 6 kg N/ha. We agree that the low dry matter content of plantain is the dominant and best documented effect of plantain on urine N dilution. We reject that there is no evidence for the effect of plantain on nitrogen partitioning to urine, and on potential nitrification rate in the urine patch. We point to empirical evidence of statistically significant reductions in nitrate leaching from plantain at paddock scale, at levels as low as 21% plantain of dry matter eaten. Current research will improve understanding of the mechanisms and magnitude of the effect of plantain on N loss, paving the way for recognition of other forage‐based N loss reduction options, and ongoing development of mechanistic models that are adaptable to other forages.
The Transfer Matrix Method (TMM) stands as the ubiquitous computational backbone for analyzing 1D wave propagation in layered media, underpinning critical product designs in photonics, seismology, and acoustics -- industries collectively valued in the tens of USD billions. Despite its essential role, legacy implementations of TMM create significant technical (and therefore strategic) bottlenecks, primarily due to a lack of straightforward differentiability and high computational costs associated with Uncertainty Quantification (UQ). This white paper assesses the current market footprint of TMM, identifies the economic "hidden costs" of traditional workflows, and outlines an emerging industrial alternative -- Differentiable Programming and Neural Surrogates -- and their own limitations.
Mari Ashiga, Vardan Voskanyan, Fateme Dinmohammadi
et al.
Recent advancements in Large Language Models (LLMs) for code optimization have enabled industrial platforms to automate software performance engineering at unprecedented scale and speed. Yet, organizations in regulated industries face strict constraints on which LLMs they can use - many cannot utilize commercial models due to data privacy regulations and compliance requirements, creating a significant challenge for achieving high-quality code optimization while maintaining cost-effectiveness. We address this by implementing a Mixture-of-Agents (MoA) approach that directly synthesizes code from multiple specialized LLMs, comparing it against TurinTech AI's vanilla Genetic Algorithm (GA)-based ensemble system and individual LLM optimizers using real-world industrial codebases. Our key contributions include: (1) First MoA application to industrial code optimization using real-world codebases; (2) Empirical evidence that MoA excels with open-source models, achieving 14.3% to 22.2% cost savings and 28.6% to 32.2% faster optimization times for regulated environments; (3) Deployment guidelines demonstrating GA's advantage with commercial models while both ensembles outperform individual LLMs; and (4) Real-world validation across 50 code snippets and seven LLM combinations, generating over 8,700 variants, addresses gaps in industrial LLM ensemble evaluation. This provides actionable guidance for organizations balancing regulatory compliance with optimization performance in production environments.
The widespread application of wireless communication technology has promoted the development of smart agriculture, where unmanned aerial vehicles (UAVs) play a multifunctional role. We target a multi-UAV smart agriculture system where UAVs cooperatively perform data collection, image acquisition, and communication tasks. In this context, we model a Markov decision process to solve the multi-UAV trajectory planning problem. Moreover, we propose a novel Elite Imitation Actor-Shared Ensemble Critic (EIA-SEC) framework, where agents adaptively learn from the elite agent to reduce trial-and-error costs, and a shared ensemble critic collaborates with each agent's local critic to ensure unbiased objective value estimates and prevent overestimation. Experimental results demonstrate that EIA-SEC outperforms state-of-the-art baselines in terms of reward performance, training stability, and convergence speed.
Feeding a larger and wealthier global population without transgressing ecological limits is increasingly challenging, as rising food demand (especially for animal products) intensifies pressure on ecosystems, accelerates deforestation, and erodes biodiversity and soil health. We develop a stylized, spatially explicit global model that links exogenous food-demand trajectories to crop and livestock production, land conversion, and feedbacks from ecosystem integrity that, in turn, shape future yields and land needs. Calibrated to post-1960 trends in population, income, yields, input use, and land use, the model reproduces the joint rise of crop and meat demand and the associated expansion and intensification of agriculture. We use it to compare business-as-usual, supply-side, demand-side, and mixed-policy scenarios. Three results stand out. First, productivity-oriented supply-side measures (e.g. reduced chemical inputs, organic conversion, lower livestock density) often trigger compensatory land expansion that undermines ecological gains-so that supply-side action alone cannot halt deforestation or widespread degradation. Second, demand-side change, particularly reduced meat consumption, consistently relieves both intensification and expansion pressures; in our simulations, only substantial demand reductions (on the order of 40% of projected excess demand by 2100) deliver simultaneous increases in forest area and declines in degraded land. Third, integrated policy portfolios that jointly constrain land conversion, temper input intensification, and curb demand outperform any single lever. Together, these findings clarify the system-level trade-offs that frustrate piecemeal interventions and identify the policy combinations most likely to keep global food provision within ecological limits.
The Entrance Dependent Vehicle Routing Problem (EDVRP) is a variant of the Vehicle Routing Problem (VRP) where the scale of cities influences routing outcomes, necessitating consideration of their entrances. This paper addresses EDVRP in agriculture, focusing on multi-parameter vehicle planning for irregularly shaped fields. To address the limitations of traditional methods, such as heuristic approaches, which often overlook field geometry and entrance constraints, we propose a Joint Probability Distribution Sampling Neural Network (JPDS-NN) to effectively solve the EDVRP. The network uses an encoder-decoder architecture with graph transformers and attention mechanisms to model routing as a Markov Decision Process, and is trained via reinforcement learning for efficient and rapid end-to-end planning. Experimental results indicate that JPDS-NN reduces travel distances by 48.4-65.4%, lowers fuel consumption by 14.0-17.6%, and computes two orders of magnitude faster than baseline methods, while demonstrating 15-25% superior performance in dynamic arrangement scenarios. Ablation studies validate the necessity of cross-attention and pre-training. The framework enables scalable, intelligent routing for large-scale farming under dynamic constraints.
Extant research has widely acknowledged the role of digital innovation as a facilitator of digital transformation, presenting solutions for various challenges in various industries. However, prior research demonstrates inadequate discussions on the determinants of digital innovation adoption for digital transformation in developing countries, particularly in the agricultural sector. To address this gap, this study investigates the effect of food security awareness, innovation characteristics, and the moderating role of agricultural experience on behavioral intention to adopt digital innovation in the agricultural sector. A dyadic model based on diffusion innovation theory and the technology acceptance model is proposed to investigate the phenomenon. This study employed a cross-sectional quantitative approach to investigate the phenomenon based on survey data collected from 207 study participants in Ghana's agricultural sector and the partial least square structural equation modeling technique. The study's findings revealed that personal innovativeness significantly affects food security awareness (β = 0.574; p < 0.000), relative advantage (β = 0.699; p < 0.000), compatibility (β = 0.687; p < 0.000), and complexity (β = 0.312; p < 0.000). In addition, food security awareness (β = 0.336; p < 0.000), compatibility (β = 0.257; p < 0.000), and agricultural experience (β = 0.238; p < 0.003) significantly affect behavioral intention to adopt digital innovation. Furthermore, the study revealed that agricultural experience (β = −0.145; p < 0.036) moderates the relationship between compatibility and behavioral intention. Together, these variables explain 78.9 % of the variance in behavioral intention to adopt digital innovation in the agricultural sector in Ghana. The study contributes to the literature on digital innovation adoption in the agricultural sector in developing countries and proffers actionable insights for practitioners.
Understanding the impacts of climate change on crop production and irrigation water demand is crucial for adapting to global warming. This study evaluated the effects of shifting planting dates on irrigated and rainfed crop yields and irrigation water demand under the latest Shared Socio-economic Pathways (SSPs) climate scenarios using the AquaCrop-OS model in Alberta, Canada. The results indicate: (1) climate change generally benefits irrigated crop yields while reducing rainfed yields under low mitigation scenarios (SSP585 and SSP370). (2) The impacts of planting date shifts on crop yields vary spatially and temporally across different SSPs. Early planting improves both rainfed and irrigated crop yields and reduces irrigation water demand under SSP585 in the latter half of the 21st Century, suggesting it is a viable strategy for mitigating heat and water stress in agricultural systems. However, this strategy does not guarantee yield increases under other SSPs. (3) The irrigated yields of spring wheat and canola are expected to increase under all scenarios, while rainfed yields decline under SSP585 and SSP370, with only marginal increases under SSP126. Annual irrigation demand will increase in the future, with the monthly irrigation peak occurring earlier. The most irrigation demand is under SSP585, followed by SSP370 and SSP126. (4) Early planting results in reduced irrigation water demand.
In recirculating aquaculture systems (RASs), degassers maintain optimal water quality by removing dissolved carbon dioxide (CO2). The performance of a degasser is generally evaluated based on its standard stripping efficiency (SSE), which is affected by its operating parameters. The present study aimed to optimize the air flow rate (QA), water flow rate (QW), and packing media height (PMH) to enhance degasser performance. To achieve this, an artificial neural network (ANN) and particle swarm optimization (PSO) were combined for parametric optimization and the predictive modeling of the SSE. The ANN model was trained using experimental data to predict the SSE, and PSO was then employed to optimize the operational parameters to achieve the maximum SSE. The optimal QA, QW, and PMH were found to be 355 m³/h, 35 m³/h, and 0.65 m, respectively, generating a maximum SSE of 0.188 kg CO2/kWh. The hybrid ANN-PSO approach was then validated by comparing experimental and predicted SSE values, with a difference between the two of only ±2.08 %. This confirms that the proposed optimization technique can reliably improve the SSE of degassers in RASs.
Preety Baglat, Ahatsham Hayat, Sheikh Shanawaz Mostafa
et al.
This study focuses on improving the automation of banana harvesting decisions for farmers with artificial intelligence assistance. Traditionally, experienced harvesters manually inspect fields to determine the optimal harvesting time, a process that is both labor-intensive and increasingly unsustainable due to a shortage of skilled workers. To address this challenge, this work proposes a computer vision-based approach for detecting banana bunches in images captured by mobile phones, as a preliminary step towards a comprehensive harvesting decision pipeline. To achieve this, a dataset was collected with 2179 photos of multiple Cavendish banana bunches in different light and exposure conditions, and a comparative analysis of You Only Look Once (YOLO) object detection models was conducted, from version 1 to 12, to identify the most accurate and efficient solution for banana bunch detection, ensuring compatibility with mobile-based applications. Among all models evaluated, YOLOv12n achieved the most balanced performance on five-fold cross-validation, with 93 % Average Precision (AP50test), 51 % AP50–95test, and 5.1 ms latency, making it well-suited for real-time deployment on resource-constrained edge devices.
Chrysanthos Maraveas, George Kalitsios, Marianna I. Kotzabasaki
et al.
Over recent decades, consumer expectations for food quality and freshness have steadily increased. To meet these standards, fresh fruits and fresh-cut vegetables in supermarkets and other commercial outlets undergo rigorous sorting processes. Quality assessments typically focus on visible characteristics such as color, ripeness, shape uniformity, defect-free skin and flesh, and texture features like firmness, toughness, and tenderness. To automate real-time quality assurance of perishable agricultural products, we have developed a user-friendly smartphone application that enables freshness assessment of apples and lettuces using RGB data at multiple stages of the supply chain. This app utilizes image recognition technology, allowing for precise freshness assessment and estimated product lifespan. Nine deep algorithms were compared in the research for image classification including Vision Transformer (ViT), Swin Transformer, Residual Networks (ResNet), EfficientNet, ConvNeXt, DeiT, MobileNetV3, MaxViT, and TNT (Transformer in Transformer). The comparison considered three metrics, including accuracy ( %), parameters (millions), and inference time (ms). Based on the findings, the MobileNetV3 was identified as the optimal deep learning architecture for the apple and lettuce classification because it maintained a good compromise between classification accuracy and mobile device resource constraints - (99.95 % and 2.5 ms for apple; 99.17 % and 2.5 million for lettuce). Such advancements offer valuable insights for policymakers, farmers, and stakeholders in making more informed decisions, thus supporting sustainable agricultural practices and improving food security across supply chains.
Eveleen A. Dawood, Thamer J. Mohammed, Buthainah Ali Al-Timimi
et al.
The disposal of wastewater resulting from petroleum industries presents a major environmental challenge due to the presence of hard-to-degrade organic pollutants, such as oils and hydrocarbons, and high chemical oxygen demand (COD). In this study, an efficient and eco-friendly method was developed to treat such wastewater using a photocatalyst composed of biochar derived from pistachio shells and loaded with zinc oxide (ZnO) nanoparticles. The biochar-ZnO composite was prepared via a co-precipitation-assisted pyrolysis method to evaluate its efficiency in the photocatalytic degradation of petroleum wastewater (PW). The synthesized material was characterized using various techniques, including scanning electron microscopy (SEM), X-ray diffraction (XRD), and Fourier transform infrared (FTIR) spectroscopy, to determine surface morphology, crystal structure, and functional groups present on the catalyst surface. Photocatalytic degradation experiments were conducted under UV and sunlight for 90 h of irradiation to evaluate the performance of the proposed system in removing oil and reducing COD levels. Key operational parameters, such as pH (2–10), catalyst dosage (0–0.1) g/50 mL, and oil and COD concentrations (50–500) ppm and (125–1252) ppm, were optimized by response surface methodology (RSM) to obtain the maximum oil and COD removal efficiency. The oil and COD were removed from PW (90.20% and 88.80%) at 0.1 g/50 mL of PS/ZnO, a pH of 2, and 50 ppm oil concentration (125 ppm of COD concentration) under UV light. The results show that pollutant removal is slightly better when using sunlight (80.00% oil removal, 78.28% COD removal) than when using four lamps of UV light (77.50% oil removal, 75.52% COD removal) at 0.055 g/50 mL of PS/ZnO, a pH of 6.8, and 100 ppm of oil concentration (290 ppm of COD concentration). The degradation rates of the PS/ZnO supported a pseudo-first-order kinetic model with R<sup>2</sup> values of 0.9960 and 0.9922 for oil and COD. This work indicates the potential use of agricultural waste, such as pistachio shells, as a sustainable source for producing effective catalysts for industrial wastewater treatment, opening broad prospects in the field of green and nanotechnology-based environmental solutions in the development of eco-friendly and effective wastewater treatment technologies under solar light.
Omar Al‐Marashdeh, H. M. Gayani P. Herath, Yvonne V. Lee
et al.
ABSTRACT The aim of this study was to investigate the effect of blood urea nitrogen (N) breeding value (BUNBV) on the concentration of urinary N (UN) and urinary urea N (UUN) of Hereford heifers. Thirty two heifers with divergent BUNBV: 16 low (BUNBV – 1.95 to −0.58, liveweight 208 ± 4.4 kg and age 303 ± 16.7 d; LBUN) and 16 high (BUNBV 0.32–1.91, liveweight 204 ± 4.8 kg and age 302 ± 7.9 d; HBUN) were studied over a 14‐day period during which they were fed lucerne silage. Individual urine, faecal, and blood spot samples were collected in the morning (0900 h) and afternoon (1400 h) on four non‐consecutive days during the second experimental week. Intakes of dry matter and N, and liveweight were similar across treatments. Positive relationships existed between BUNBV and UN (R 2 = 0.51; P < 0.001), and BUNBV and UUN (R 2 = 0.57; P < 0.001). One unit increase in BUNBV increased heifer UN by 0.60 ± 0.254 g/L and UUN by 0.54 ± 0.221 g/L across both sampling times (AM and PM). This suggests that breeding for low BUN may reduce the environmental impact of pastoral beef production systems. However, further research is needed to compare N balance and urination behaviours of beef heifers divergent in BUNBV.
To mitigate soil degradation and decrease dependency on chemical inputs in agriculture, this study examined the joint effects of coconut shell biochar and Bacillus strain Ya-1 on soil fertility, rhizosphere bacterial communities, and the growth of chili (Capsicum annuum L.). A controlled pot experiment with four treatments was conducted: control (CK), biochar only (C), Bacillus strain Ya-1 only (B), and a combination of both (BC). The BC treatment significantly enhanced the soil carbon and available phosphorus contents by approximately 20% and the soil nitrogen content and pH by 18% and 0.3 units, respectively, compared to the control. It also increased microbial biomass carbon and nitrogen by 25% and 30%, respectively, indicating improved soil microbial diversity as shown by the highest Pielou evenness index and Shannon index values. The combined application of biochar and the Ya-1 strain resulted in a 15% increase in chili plant height and a 40% improvement in root dehydrogenase activity, suggesting enhanced nutrient uptake and metabolism. Metabolic profiling showed shifts in stress response and nutrient assimilation under different treatments. Collectively, these results indicate the potential of biochar and microbial inoculants to significantly promote soil and plant health, providing a sustainable strategy to improve agricultural productivity and reduce reliance on chemical inputs.
Rikhiya Ghosh, Oladimeji Farri, Hans-Martin von Stockhausen
et al.
The healthcare industry is currently experiencing an unprecedented wave of cybersecurity attacks, impacting millions of individuals. With the discovery of thousands of vulnerabilities each month, there is a pressing need to drive the automation of vulnerability assessment processes for medical devices, facilitating rapid mitigation efforts. Generative AI systems have revolutionized various industries, offering unparalleled opportunities for automation and increased efficiency. This paper presents a solution leveraging Large Language Models (LLMs) to learn from historical evaluations of vulnerabilities for the automatic assessment of vulnerabilities in the medical devices industry. This approach is applied within the portfolio of a single manufacturer, taking into account device characteristics, including existing security posture and controls. The primary contributions of this paper are threefold. Firstly, it provides a detailed examination of the best practices for training a vulnerability Language Model (LM) in an industrial context. Secondly, it presents a comprehensive comparison and insightful analysis of the effectiveness of Language Models in vulnerability assessment. Finally, it proposes a new human-in-the-loop framework to expedite vulnerability evaluation processes.
Lars Nieradzik, Henrike Stephani, Jördis Sieburg-Rockel
et al.
In this study, we explore the explainability of neural networks in agriculture and forestry, specifically in fertilizer treatment classification and wood identification. The opaque nature of these models, often considered 'black boxes', is addressed through an extensive evaluation of state-of-the-art Attribution Maps (AMs), also known as class activation maps (CAMs) or saliency maps. Our comprehensive qualitative and quantitative analysis of these AMs uncovers critical practical limitations. Findings reveal that AMs frequently fail to consistently highlight crucial features and often misalign with the features considered important by domain experts. These discrepancies raise substantial questions about the utility of AMs in understanding the decision-making process of neural networks. Our study provides critical insights into the trustworthiness and practicality of AMs within the agriculture and forestry sectors, thus facilitating a better understanding of neural networks in these application areas.