SMARTKISAAN: An AI-Driven, IoT-Based Smart Irrigation Monitoring System for Small and Marginal Farmers

Small and marginal farmers often irrigate by observation and fixed schedules, which can lead to over-irrigation, under-irrigation, and wasted water. This paper describes SmartKisaan, a low-cost, solar-powered smart irrigation prototype that combines an ESP32 microcontroller, soil-moisture, temperature-humidity, rain, and water-tank sensors, Firebase Realtime Database, a Random Forest irrigation model served through a Flask API, and a Flutter mobile application with weather-based scheduling, leaf-image disease screening, and an English and Kannada voice-enabled assistant. Every pump command passes through a layered safety hierarchy that is also implemented in the ESP32 firmware. On a held-out test set of 700 records, the Random Forest achieved 99.00% accuracy and an F1-score of 97.73%. However, the 3,500 records used for training were generated by simulation from agronomic rules, so these results show that the model learns the intended irrigation policy and do not represent measured performance in a real field. Four of five hardware bench tests passed and one failure was recorded; all 12 automated decision-logic tests passed. These checks do not establish water savings, crop-yield improvement, or long-term field reliability. SmartKisaan is a decision-support prototype and not a validated commercial irrigation controller.