Towards a layered agent-modeling of IoT devices to precision agriculture Danilo Cavaliere Department of Information and Electrical Engineering and Applied Mathematics, University of Salerno, Fisciano (SA) 84084, Italy. e-mail: dcavaliere@unisa.it Vincenzo Loia Dipartimento di Scienze Aziendali, Management & Innovation Systems University of Salerno, Fisciano (SA) 84084, Italy. e-mail: loia@unisa.it Sabrina Senatore Department of Information and Electrical Engineering and Applied Mathematics, University of Salerno, Fisciano (SA) 84084, Italy. e-mail: ssenatore@unisa.it Abstract—Precision agriculture employs IoT devices to smartly monitoring plant vegetation and support food production. Preci- sion agriculture is highly required to improve product quality and better suit the requirements of the market. Among the IoT devices, Unmanned Aerial Vehicles (UAVs), can be equipped with many sensors that allow precise assessments of plant stress by flying over the plots. Notwithstanding the great benefits introduced, IoT devices may suffer from some issues. Many devices provide data in different formats on the same task, therefore they need solutions to integrate data and support a more thorough crop monitoring. This paper introduces a multi- tier architecture to deal with IoT-based intelligent monitoring, as well as an implementation of the architecture through multi- agent modeling of the IoT devices for precision agriculture. The introduced model allows data acquisition from various sources (i.e., IoT devices), an ontology-based integration of data provided by the devices and a knowledge integration process to deal with domain-specific applications. Index Terms—Precision agriculture, IoT, UAV, Multi-Agent Systems, Ontology I. I NTRODUCTION The spread of IoT technologies provided benefits in many distinct fields, from healthcare to business. In the agriculture field, the use of smart sensors paved the way to precision agriculture, which allows constant monitoring of plant growth and crop quality. Precision agriculture provides a food sup- ply chain that better suits the requirements of the market, it improves process quality, reduces production times and increases incomes. IoT devices help agriculturists to monitor threats and damages to their plots; equipped with actuators, they can also address the detected issues on time, such as specific environmental conditions, or the spread of viruses, which may have quick devastating effects on the crop. Among the various IoT devices, mobile devices, and more specifically, Unmanned Aerial Vehicles (UAVs) have been investigated to collect information on plant stress, environmental conditions [1]. UAVs are low-cost solutions; since they can be equipped with various sensors and easily take plots from the above, they are particularly suited to make precise assessments on vegetation or even take specific actions on selected plants. In UAV-driven path monitoring, recent studies [2] show how stochastic models support path planning in a UAV swarm and avoid collisions; in [3], a routing algorithm allows UAVs to avoid multi-obstacle areas in a plot. Geo-localization issues are faced in [4] determining the UAV position and orientation during its vision-based navigation or in [5] through data muling from acoustic sensor networks. Although the use of multiple IoT devices offers interesting benefits, they may suffer from various issues that can compro- mise their efficacy in precision agriculture. Issues are related to data communication (i.e., network stability issues) [6], control [7], acquisition and sharing of information [8]. Keeping a stable network is fundamental to guarantee the collaboration among the devices and quick replies to the human experts. The solutions in the literature focus on improving network stability in wide-range operations [6], swarm control [7], by building Fuzzy Neural Network models; in [8], an agent-based model allows UAV teams to build collective knowledge through Fuzzy Cognitive Maps to support surveillance applications. Vegetation monitoring and assessment are strictly based on indices from spectral data or vegetation measures as well as unsupervised classification techniques for fruit and plant detec- tion, and vegetation monitoring. The approach proposed in [9], for instance, applies spectral clustering to the collected images to detect tomatoes. In [10], sunflower growth is monitored through multi-temporal imaging. Since IoT data come from heterogeneous sensors and returned in different formats, the need for a homogeneous data reading claims new and challenging methods for data integration. In [11] a Markov-chain-based model is defined to integrate data in a IoT ecosystem; another study [12] designs an ontology-based architecture to support data integration and sharing among enterprises. Data integration is also required at agriculture domain level. Agriculture processes, indeed, consist of multiple phases (i.e., product, techniques, production processes), and knowledge-based solutions [13], [14] have been investigated to address inter-phases data integration. In knowledge-based approaches, ontologies map accurately the precision agriculture domain [13]; Agri Ont [14] is an ontology coming from domain-specific ontologies, such as IoT devices, precision agriculture, geo-location, and business aspects. Similarly, in [15], the fusion of two ontologies, one modeling food production domain, the other the agriculture processes, provides a rich knowledge base for applications on 978-1-7281-6932-3/20/$31.00 ©2020 IEEE