Agent-Based Crop Model for Smart Irrigation: Design of a State Estimator

In the design of smart irrigation systems, there are several open challenges, among which: i) the modeling of heterogeneity of cropping land, and ii) the estimation of non-measured state variables to control crop development. This work addresses both challenges by an agent-based model (ABM) of a discretized field and by using state estimation techniques. For the last challenge, two software sensors, i.e., an extended Kalman filter (EKE) and an unscented Kalman filter (UKF) are used and compared to estimate on-line the states of homogeneous portions of land assigned to the agents of an ABM model. The agent-based crop model is presented and simulated under two different climatic scenarios to assess the performance of the estimation techniques. Simulation results of a testbed in Colombia shows the advantages of UKF over the EKF.