An IoT based Smart Irrigation Management System using Reinforcement Learning modeled through a Markov Decision Process

In this paper, the proposal of an irrigation system exploiting IoT is provided. The proposed system, based on IoT sensors and smart platforms such as Raspberry PI and Arduino, is able to manage farm operations in term of irrigation. The control of water pumps for irrigation is driven by two important parameters: soil moisture and evapotranspiration. The system uses a Reinforcement Learning approach based on Markov Decision Process in order to learn the right water amount needed by the plants. This approach allows to reduce both water and energy consumption. The proposal has been compared with a conventional irrigation system that normally is based on a soil humidity threshold and takes its decision considering the threshold setting. Conducted experiments show the water and energy saving by the use of the proposed Smart Irrigation system.