10th International Congress on Information and Communication Technology

Authors - Maryem MELHAOUI, Abderrahim BAJIT, Khalid BOUALI, Youness ZAHID, Hicham Essamri, Rachid EL BOUAYADI
Abstract - To assure sustainability and optimize production and resource management in the agricultural sector while reducing the impact on the environment, we are currently using precision agriculture, which is mainly based on advanced technologies, including IoT platforms. In this context, the efficiency and accuracy of these IoT platforms depends heavily on the performance of its electronic resources and the distribution of its nodes. In the literature, the architecture of IoT platforms often consists of static nodes designated for the collection and communication of data to the computing machine. This distribution can also be mobile using self-driving robots on which nodes are hosted. In this perspective, we have studied, designed and deployed an IoT platform with dynamic distribution of nodes that is the subject of discussion and analysis in this manuscript. The idea is based on the intelligent deployment of the node operating mode, namely: activated or deactivated. To achieve this idea, we used two types of intelligent learning: supervised learning and unsupervised learning that both use a battery of intelligent models capable of analyzing and predicting the voluntary or involuntary missing data within the IoT platform. To recover missing data, two types of predicting solutions are available either by taking advantage of the arrival of certain new PAYLOAD’s components, or by using features’ history. To accomplish this we deployed Random Forrest (RF) and LSTM models. The paper’s objective consists of an intelligent and sustainable green solution to optimize the resource management of a dynamic IoT platform dedicated to greenhouse supervision.