Background: Water scarcity poses a critical challenge to global food security, with agriculture consuming approximately 70% of freshwater resources worldwide. Small-scale farms in developing regions face particular challenges in implementing smart irrigation technologies due to limited internet connectivity, high costs, and complexity of existing cloud-based solutions. This paper presents an AI-powered smart irrigation system designed specifically for resource-constrained environments, featuring offline operation, edge-based intelligence, and low-cost hardware implementation. Materials and Methods: The proposed system employs a three-node architecture based on ESP32 microcontrollers communicating via the ESP-NOW protocol. The Brain Node executes the Hargreaves-Samani evapotranspiration model and implements a hybrid AI decision engine called TinyAdjuster. The TinyML model occupies only 65.3 KB with 18.7 ms inference time. Simulation-based validation was conducted across three crops over 30 days. Results: The AI Adaptive system achieves 24.1% water savings versus traditional irrigation and 8.8% versus threshold IoT systems, with 96.2% efficiency. R² > 0.87 for all crops. Learning converges in 25 cycles, reducing error from 2.8 to 0.45 mm/day (84% improvement). Conclusion: The system features offline AI operation, MAD-based triggering, variable irrigation amounts, and adaptive learning. These innovations make it suitable for remote agricultural areas with limited infrastructure.
Moreira et al. (Fri,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: