Inference System for Irrigation Scheduling with an Intelligent Agent

Abstract

The design and implementation of an inference method to schedule irrigation tasks in agriculture, based on the concept of rational agent is explained in this paper. Through the use of a raspberry-pi and xbee devices, the agent interacts with its environment acquiring soil moisture, soil temperature, luminosity, air temperature and rain data. Membership functions and a Mamdani inference methodology were implemented on the raspberry-pi to define irrigation time. Furthermore solenoid valves were used as actuators to apply the amount of water prescribed on the field. The inference system uses luminosity and ambient temperature data to define periods with high levels of evapotranspiration and soil moisture sensors to determine the Volumetric Water Content (VWC). As result of this implementation, temperature membership functions were defined, including the annual averages of minimum (9C) and maximum (20C) temperatures, according with weather data obtained in Nobsa, Boyacá, Colombia. Values between 0 and 2500 lx, were defined in the universe of discourse of brightness and soil moisture membership was defined using field capacity and permanent wilting points. The system developed allows to maintain the VWC in the soil near to the field capacity value, wherewith, soil moisture is maintained at the optimum level necessary for the correct development of a crop.