Smart AI-Based Agriculture Decision Support System (AgriSens)

Abstract

Traditional farming relies heavily on empirical guesswork and legacy ancestral knowledge, frequently leaving crops vulnerable to uncalibrated chemical fertilizer application, unpredictable micro-climates, and delayed foliar disease detection. This paper introduces AgriSens, a comprehensive, full-stack smart farming platform developed to empower farmers with data-driven precision agriculture. The system integrates a multi-parameter crop recommendation engine using a mathematical Gaussian-decay environmental fitness model alongside a multimodal leaf disease diagnostic system powered by Google Gemini 3.7 Flash computer vision to identify foliar pathogens from smartphone imagery. Furthermore, the architecture deploys a soil-texture-aware fertilizer optimization engine that calculates precise N-P-K (Nitrogen, Phosphorus, Potassium) nutrient deficits, a real-time agro-meteorological forecasting stream driven by the Open-Meteo API for chemical spray safety alerts, and a closed-loop automated IoT irrigation telemetry simulator. To eliminate structural literacy barriers within rural farming communities, the platform features native voice synthesis and a 7-language localized translation module. Simulation benchmarks demonstrate that the platform achieves rapid visual diagnostic latencies under 2.5 seconds and optimizes water conservation by 25% to 35% compared to baseline scheduled flooding methods.

Introduction

The text introduces AgriSens, an integrated, mobile-first smart agriculture decision-support platform designed to address major agricultural challenges, including declining groundwater levels, soil degradation, inefficient fertilizer use, and crop diseases. The system combines real-time environmental parameters, mathematical modeling, cloud-based artificial intelligence, and IoT-based automation to provide farmers with accurate, localized, and data-driven agricultural recommendations.

AgriSens uses a three-tier modular architecture consisting of a frontend presentation layer, an application and backend layer, and connected intelligence services. The frontend is developed using React 18, TypeScript, and Tailwind CSS, providing a responsive user interface with multilingual and text-to-speech support in seven languages: English, Hindi, Spanish, Marathi, Telugu, Punjabi, and Tamil.

The platform incorporates four primary functional modules:

• Crop Recommender: Evaluates soil nutrients, pH, temperature, humidity, and seasonal rainfall using a Gaussian decay function and weighted geometric mean to calculate crop suitability scores.

• Precision Fertilizer Planner: Estimates nitrogen, phosphorus, and potassium deficiencies while accounting for soil texture and nutrient retention efficiency. It calculates recommended quantities of Urea, DAP, and MOP to improve fertilizer utilization.

• Agro-Weather Service: Retrieves real-time weather information through external REST APIs and a secure Node.js and Express backend.

• Disease Detection Engine: Uses the Google Gemini API to analyze uploaded leaf images, identify potential crop diseases, and provide confidence indicators and treatment recommendations.

For future physical implementation, AgriSens proposes an IoT architecture based on the ESP32-WROOM-32D microcontroller. The hardware includes soil nutrient sensors, moisture sensors, environmental monitoring devices, and relay-controlled irrigation systems. The automated irrigation mechanism activates when soil moisture falls below 35% and continues until the target moisture level of 75% is reached.

The reported experimental results indicate that the crop recommendation module aligns with established agricultural guidelines, while the disease detection engine achieves diagnostic response times of approximately 2.5 seconds. Simulation results also suggest a potential 25%–35% reduction in water consumption compared with conventional timer-based irrigation methods.

Conclusion

The AgriSens platform demonstrates how precision web architecture, artificial intelligence, and simulated IoT networks can combine into a powerful, accessible agronomic dashboard. By replacing subjective, traditional farming methods with calculated mathematical assessments, the system lowers input costs, prevents sudden crop failures, and saves valuable water resources. Future enhancement phases will focus on integrating live Sentinel-2 NDVI multispectral satellite data to track field-wide crop health over time and implementing localized TensorFlow Lite models for offline visual diagnostics in deep rural farming areas with poor internet connectivity.

References

[1] Author(s), \"Smart AI-Based Agriculture Decision Support System (AgriSens),\" Major Project Phase-2 Report, Department of Computer Science & Engineering, Mysuru Royal Institute of Technology, Mandya, Oct. 2026. [2] Google Generative AI Team, \"Gemini 3.7 Flash: Multimodal Vision and High-Speed Inference Capabilities,\" Tech. Rep., 2025. [3] Open-Meteo API Documentation, \"Agro-Meteorological Forecasts and Real-Time Soil Telemetry Integration Data,\" Available: https://open-meteo.com.

Copyright

Copyright © 2026 Harshini G. K., Monisha B. P., Manoj G., Chandana K. M., Prof. Sushma S.. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.