Edge-enabled smart agriculture framework: Integrating IoT, lightweight deep learning, and agentic AI for context-aware farming
Bibliographic record
Abstract
• Edge-first framework unites IoT, dual vision models, and rule-based actions. • MiT-B0 at 128 × 128 enables low-power on-device weather and crop inference. • Achieves 88% weather and 93% crop accuracy with robust error metrics. • Agentic layer closes perception-to-actuation loop for autonomous control. • CPU-only Python case studies validate real-time decisions without cloud. • Proposed a dual-vision deep learning architecture that performs simultaneous crop and weather classification using lightweight, edge-optimized models for real-time operation. • Developed a rule-based agentic AI decision layer that fuses multimodal predictions to autonomously drive IoT-enabled agricultural interventions. • Demonstrated real-time responsiveness through seamless integration with field-deployable IoT devices such as irrigation controllers, drones, and environmental sensors. • Validated the system’s scalability and deployment-readiness for Agriculture 4.0 through rigorous experimentation, applied scenarios, and quantitative performance metrics. • Addressed major limitations in existing smart agriculture frameworks—including lack of context-awareness, server dependency, and resource inefficiency—by enabling autonomous, interpretable, and low-cost actuation in real-world environments. Smart farming in connectivity-limited, energy-sensitive environments demands on-device perception and decision-making to reduce latency and cloud dependence. This article proposes an edge-enabled smart agriculture framework that integrates lightweight deep learning, rule-based agentic AI, and Internet of Things (IoT) devices for real-time, autonomous farming decisions. The system features two vision-based models—one for weather classification and one for crop identification—built on the MiT-B0 Vision Transformer architecture and optimized for low-resolution (128 × 128) image inputs. These models run on resource-constrained hardware suitable for rural deployment and support efficient, on-device processing. Weather prediction spans 11 classes (e.g., frost, lightning, rain, sandstorm), while crop classification covers 5 major crops. The system achieves an accuracy of 88% for weather and 93% for crops, with high F1-scores and low MAE, Kappa, and Hamming loss values. Predictions are interpreted by a rule-based agentic AI layer that triggers actions across multiple IoT actuators, such as smart irrigation, NDVI sensors, frost alarms, drones, and pest detectors. The decision engine supports both joint rule logic (e.g., activating hail protectors when hail is detected in maize fields) and fallback single-condition rules. Python-implemented case studies show seamless model–AI–IoT interaction in combined and separate scenarios. By minimizing cloud dependency, reducing communication overhead, and enabling low-power operation, the proposed framework addresses critical challenges in connectivity-limited, energy-sensitive agricultural contexts. It demonstrates the potential for scalable and intelligent smart farming, aligning with the goals of sustainable Agriculture 4.0.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".