AgriSmart: An IoT-enabled framework for agricultural resource optimization

Efficient use of farming resources (e.g., nitrogen, water, pesticides) is key to maximizing productivity and promoting sustainable agriculture. Traditional methods, such as fixed-rate applications or soil sampling, often fail to adapt to changing in-season conditions and specific nutrient demands, leading to inefficiencies and environmental harm. In this work, we propose AgriSmart, an IoT-enabled framework that optimizes resource application strategies to maximize crop yield while minimizing resource usage within a given budget. AgriSmart formulates an optimization problem solved periodically using an enhanced Differential Evolution (DE) algorithm that balances exploration and exploitation, following a Model Predictive Control (MPC) approach. Crop yield responses to varying application timings and rates are estimated using the process-based crop simulation model DSSAT (Decision Support System for Agrotechnology Transfer). To improve flexibility and reduce computational complexity, we introduce adjustable receding horizon that allows multiple actions to be applied before re-optimization, enabling adaptation to resources with different application frequencies (e.g., water vs. nitrogen). As the time horizon advances, AgriSmart dynamically adjusts the resource applications to better match crop needs at each growth stage, responding to evolving weather and field conditions. We evaluate AgriSmart in two use cases: irrigation scheduling for soybean and nitrogen management for maize. Results show that AgriSmart outperforms existing methods, achieving up to 21.4% water savings for soybean without yield loss, and increasing maize yield by 20% while reducing nitrogen use by up to 32%.