Monitoring plant traits such as chlorophyll content and stress response is essential for agricultural research and crop management. However, traditional phenotyping methods are often destructive, labor-intensive, and lack temporal resolution. Continuous imaging using digital cameras offers a non-invasive approach to tracking plant development over time. This study evaluated trail cameras commonly used in ecological research for their effectiveness in continuous, nondestructive plant monitoring in an agricultural setting. Eight cameras were deployed at canopy height in a cranberry field to capture images and temperature data at 30 min intervals over a 15-day period. Based on operational reliability and data quality, four cameras were shortlisted for detailed analysis. Key performance parameters included image quality (blur, color cast, and noise), temperature measurement accuracy, and battery longevity. Temperature readings were compared with data from a canopy-level probe sensor and a dike-level weather station. Cameras 3, 5, and 7 each demonstrated strengths across different performance metrics, with Camera 3 excelling in battery longevity (maintaining 100% capacity over 15 days) and temperature correlation (R2 up to 0.983), Camera 5 showing superior color accuracy (lowest color cast score: 9.679 ± 0.06), and Camera 7 achieving the lowest noise levels (0.906 ± 0.002) along with highly consistent imaging performance. These models also demonstrated reliable temperature measurement accuracy, with coefficients of determination (R2) ranging from 0.935 to 0.983 when compared with canopy-level probe data. Overall, the results indicate that each camera offers distinct advantages depending on the specific monitoring requirements, such as image clarity, color accuracy, noise reduction, or continuous field deployment. The results highlight trail cameras as affordable and reliable tools for remote, non-destructive, and continuous plant monitoring, offering valuable guidance for researchers in plant phenotyping and crop management.
Hydroponic production enables precise control of nutrient supply, irrigation, lighting and root-zone conditions, creating a need for rapid, reliable and biologically meaningful monitoring. Optical and spectroscopic approaches can non-destructively assess nutrient solutions, roots, plant physiology and crop quality. This review critically evaluates VIS–NIR and NIR spectroscopy, multispectral and hyperspectral imaging, chlorophyll fluorescence, Raman spectroscopy, FTIR/MIR spectroscopy, laser-induced breakdown spectroscopy (LIBS) and optical sensing of nutrient solutions. Rather than ranking technologies by predictive accuracy, we classify them by the information they provide: direct or near-direct measurement, calibrated quantitative prediction, physiological proxy sensing, and classification or anomaly detection. Despite substantial proof-of-concept progress, evidence remains constrained by homogeneous datasets, information leakage, limited external validation, calibration transfer barriers and scarce commercial evaluation. Reliable deployment requires representative calibration, validated reference methods, biologically independent validation, uncertainty assessment and continuous quality assurance. Progress will depend less on increasingly sophisticated algorithms than on stronger analytical and biological foundations and integration of complementary solution, environmental and physiological measurements. Spectroscopic sensing should therefore progress from isolated sensor-to-trait models towards validated decision support systems linking measurement, state estimation, diagnosis, intervention and outcome verification, evaluated for analytical validity, transferability, biological interpretability, deployment readiness and demonstrated management value.
This work presents the design and the first quantitative greenhouse assessment of an ultrasonic device that pollinates tomato flowers without physical contact. Greenhouse tomato production relies on commercial bumblebee hives, which entail recurring costs, management and regulatory constraints, and biosecurity risks. A portable handheld airborne ultrasonic device integrating two synchronized open-hardware phased-array modules (2 × 256 elements, 40 kHz carrier) was developed, with electronic phase control for focal-point steering and amplitude modulation (20 Hz–1 kHz) to excite floral vibrations. A coupled finite-element/boundary-element model indicated that reducing the acoustic wavelength toward millimeter-scale floral dimensions increases acoustic–structure coupling. Performance was evaluated in a greenhouse on 102 tomato plants divided into three groups (untreated controls, a commercial vibrating wand, and ultrasonic actuation) at operating distances of 0.1–0.3 m, using a single module aimed by one operator. Focused ultrasound produced sound pressure levels of 155–160 dB (RMS, re 20 µPa) at the focal point, and ultrasonic pollination consistently achieved fruit-set rates above 90% (up to 96.5%), approaching the commercial vibrating wand (94%), whereas untreated controls remained at 53–65%. Under the greenhouse conditions and working distances tested, focused, modulated airborne ultrasound may therefore offer a promising non-contact alternative to contact-based pollination. Occlusion sensitivity, acoustic attenuation, and operator exposure remain to be addressed.
Decarbonizing transport in agro-exporting economies requires tools that couple biofuel production with by-product valorization without displacing food. This study extends a low-data predictive model of sugarcane by-product bioenergy with three ethanol pathways—direct juice fermentation (1G-A), flexible sugar–ethanol production (1G-B) and simultaneous saccharification and fermentation of bagasse (2G)—coupled to anaerobic digestion, and applies it to Peru, a structurally net sugar-importing sector. The research question is how much fuel ethanol such a sector can supply without reducing its commercial sugar output, and which route carries that supply with the least sensitivity to the sugar balance. The sugar-preserving allocation coefficient fjuicemax,SP is derived in closed form and shown to depend on the industrial configuration mix only through the aggregate sucrose diverted. The sugar reference is the 2011–2018 mean of the official series, the same window as the cane baseline, and the recovery chain is calibrated on that pair (λ∗=0.9604). The 1G-A cascade reproduces the nominal capacity ratio of the only Peruvian juice-to-ethanol plant within 1.24%, a design-level benchmark rather than an industrial validation; the 2G coefficient (254.9 L t−1 dry bagasse) agrees within 10% with an independent process simulation but remains unvalidated for Peruvian bagasse. Without productive growth the constraint is exactly binding: no juice can be diverted and the 2030 potential is 297 ML yr−1 from residual molasses and bagasse, rising to 498 and 731 ML yr−1 under 2.5% and 5% annual cane growth, with 790–1231 GWh yr−1 of cogenerated electricity as a gross upper bound. Single-year sugar references move the no-growth margin between 0 and 0.067; the 0.104 of an earlier formulation is an artefact of pairing an uncalibrated recovery chain with the lowest year of the series. Analytically, the constrained yield is a hinge function of cane sucrose content with an 8.3-fold slope change where domestic sugar demand is met, and the elasticity of the diversion margin to every multiplicative input of the sugar balance equals (1−f)/f. This structure explains why sucrose content governs 85–90% of the flexible-pathway variance, why the null-margin probability is 52% at the binding point, and why, within the model, the lignocellulosic route is least exposed to the sugar balance, its absolute level remaining conditional on the transferability of the 2G coefficient.
The Internet of Things (IoT), robotics, and artificial intelligence have only recently converged in agriculture, driven by the need for sustainable, resilient food systems. Precision farming demands real-time physical action, not just data-driven insight, yet existing agricultural robotic systems rarely integrate omnidirectional mobility, dexterous manipulation, and environmental awareness for complex tasks in unstructured farming environments. This systematic literature review examines the state of the art in IoT-enabled smart agricultural management systems that combine Mecanum-wheel mobile platforms with multi-degree-of-freedom (DoF) robotic arm manipulators. Following the PRISMA 2020 guidelines, we analysed 81 peer-reviewed articles (2015–2025) from Scopus, Web of Science, and IEEE Xplore. By design, every included study employs Mecanum-wheel mobility, since this was a required inclusion criterion; this reflects our deliberately narrow scope rather than the adoption rate of Mecanum wheels in agricultural robotics generally, which the wider excluded literature suggests remains comparatively low. We synthesise key mathematical models, Mecanum inverse/forward kinematics, 6-DoF Jacobian matrices, LoRaWAN link budget analysis, and closed-loop PID control across sensing, mobility, manipulation, integration, and validation. This synthesis reveals five recurring limitations: poor terrain adaptability of Mecanum rollers in muddy fields, communication latency exceeding 1 s in long-range protocols, absent force feedback in most manipulator designs, and 28–42% performance degradation under field versus laboratory conditions. Few studies report true closed-loop operation where sensor data directly triggers robotic action, and smallholder farmers, particularly in developing economies such as Nigeria, remain excluded from these technologies due to cost, energy, and usability barriers. To address these gaps, we propose a modular system architecture featuring solar-assisted Mecanum mobility, a 6-DoF force-controlled arm with quick-change end-effectors, hybrid LoRaWAN/Wi-Fi 6 communication, edge AI (NVIDIA Jetson), and a multilingual cloud dashboard laying a foundation for scalable, inclusive agricultural robotics. This proposed architecture is a conceptual direction motivated by the review’s findings and is not itself an empirical outcome of the systematic review.
