140x Filetype PDF File size 0.58 MB Source: hal.archives-ouvertes.fr
On Designing and Implementing Agro-ecology IoT Applications: Issues from Applied Research Projects Sandro Bimonte, Amina Belhassena, Christophe Cariou, Jean Laneurit, Rim Moussa, Gerard Chalhoub, Robert Wrembel, Gauthier Picard, Ladjel Bellatreche, Alexandre Journaux, et al. To cite this version: Sandro Bimonte, Amina Belhassena, Christophe Cariou, Jean Laneurit, Rim Moussa, et al.. On Designing and Implementing Agro-ecology IoT Applications: Issues from Applied Research Projects. International Workshop on Latest Advances in Enterprise Architectures in the IoT Era (EAIoT’2021)/International Entreprise Distributed Object Computing Workshop (EDOCW), IEEE Comp Soc; Griffith Univ; Destinat Gold Coast, Oct 2021, Gold Coast, Australia. 10.1109/EDOCW52865.2021.00050. hal-03479790 HALId: hal-03479790 https://hal.science/hal-03479790 Submitted on 14 Dec 2021 HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. Copyright On Designing and Implementing Agro-ecology IoT Applications: Issues from Applied Research Projects ∗ ∗ ∗ ∗ † ‡ Sandro Bimonte , Amina Belhassena , Christophe Cariou , Jean Laneurit , Rim Moussa , Gerard Chalhoub , § ¶ k ∗∗ Robert Wrembel , Gauthier Picard , Ladjel Bellatreche , Alexandre Journaux , ∗∗ †† ‡‡ ´x Thierry Heirman , Ali Hassan , Stefano Rizzi , Jean-Pierre George ∗ ´ Universite Clermont Auvergne, TSCF, INRAE, France name.surname@inrae.fr †University of Carthage, Tunisia rim.moussa@enicarthage.rnu.tn ‡ ´ Universite Clermont Auvergne, LIMOS-CNRS, Clermont-Ferrand, France gerard.chalhoub@uca.fr §Poznan University of Technology, Poland robert.wrembel@cs.put.poznan.pl ¶ ´ ONERA/DTIS, Universite de Toulouse, France gauthier.picard@onera.fr k ISAE-ENSMA, France bellatreche@ensma.fr ∗∗GenPhySE, INRAE, France alexandre.journaux@inrae.fr, thierry.heirman@inrae.fr ††Umanis, France ahassan@umanis.com ‡‡DISI, University of Bologna, Italy stefano.rizzi@unibo.it xIRIT, University of Toulouse III, France jean-pierre.george@irit.fr Abstract—The design and implementation of agro-ecology IoT Variety, and Velocity. The usage of IoT in the agricultural applications is a non-trivial task since the data processed in such business is needed and promising. Indeed, one recent report1 applications are typically complex and heterogeneous. Moreover, estimates that in 2027 this sector will reach 34 billion USD. these applications are implemented using different systems and Agro-ecology aims to develop new cultural practices that technologies, over complex IoT communication network layers respect the environment and at the same time save production (edge, fog, cloud). The existing system design methods fail to effectively represent data in such a scenario. In this position and biodiversity [3]. Agro-ecology has been recognized by all paper we report and discuss the open issues for a new, dedicated governmental, economic, social and environmental institutions design method, based on our initial experience in implementing as one the main challenges of humanity for the next 30 years. an agro-ecology IoT system. Data used by agro-ecology models are very diverse, including Index Terms—Internet of Things, Big data, smart farming, environmental, agricultural, and socio-economic data, at dif- Agriculture robots ferent (micro and macro) spatial and temporal scales and also I. INTRODUCTION data coming from mobile autonomous robots and drones. In the context of agriculture, IoT has been successfully In the recent years, the Internet of Things (IoT) has been employed for different applications, e.g. agronomic surveil- successfully applied in several different application domains, lance and livestock production. IoT leads to a revolutionary as for example healthcare, environment, mobility, and even approach for agro-ecology since it provides the stakeholders agriculture [1]. IoT is the set of physically connected devices with more precise, complete, and innovative data and their that support computation and communication by means of dif- associated analysis. In particular, at the crossroads between ferent communication networks (e.g., ZigBee, Wi-Fi, ADSL). As described in [2], IoT produces Big Data, which are data 1https://www.marketsandmarkets.com/Market-Reports/ mainly characterized by (at least) the 3Vs, namely Volume, iot-in-agriculture-market-199564903.html breeding and agro-ecology, two main research topics emerge: provided by end-users. However, to the best of our knowledge, image recognition via neural networks to detect and recognize data modeling methods for IoT agricultural applications have the parasites on the legs of a grazing animal, and then the not been deeply investigated so far (see [5] for a complete geo-localization of parasites on a plot or a territory. The survey). In this position paper, we motivate the need for a monitoring of crops development and agricultural practices new methodology for agro-ecology applications’ design and using autonomous robots is another hot research topic. Agro- implementation (Sec. II), then we present the modelling and ecological animal and plant breeding in the era of IoT and implementation requirements and some envisaged solutions. Artificial Intelligence implies the usage of wireless sensors, Some relevant works are presented in Sec. IV. This contri- drones, satellite images, multimedia data, and classical data in bution is based on our findings while realizing some agro- an integrated, coherent, and effective way. ecology research projects. AnexampleofaclassicalIoTarchitecture in the agricultural II. MOTIVATION context is illustrated in Figure 1, which shows the data and the network connections involved. IoT in the agro-ecological context comes with new issues Data is collected, and sometimes computed, by IoT devices that we discuss in this section. (such as autonomous robots, tractors, meteorological sensors, A. What drones, etc.) deployed in the field. These IoT devices produce Agro-ecology IoT data have a spatio-temporal nature since real-time stream data which, when combined with other data all agronomic and bravery phenomena are geolocalized (e.g., (such as farm data, geospatial data, images, etc.), can be used plots, positions of animals and robots). These data are com- for online analyses at the farm level. Moreover, historical plex, ranging from images and videos to time series pro- IoT data and other external data can be used to provide duced by sensors and autonomous robots. Therefore, they more complex analyses (such as prediction models, OLAP, need ad-hoc conceptual representations and implementations. etc.). Therefore, an IoT agriculture application is usually fed Indeed, IoT systems typically rely on relational or NoSQL with data coming from the field and historical external data database managment systems (DBMSs), data stream manage- in a real-time way. All these data are deployed in different mentsystems(DSMSs),andothercomponentsimplementedin data management systems (sensors devices, tractors’ laptops, different technologies and supporting different programming classical PCs, distributed servers, etc.). These data manage- languages, which run on heterogeneous hardware (IoT devices, ment systems are deployed at different levels of the network personal computers, cloud servers). Here, such complex, het- architecture (directly on the field, in the farm, in the cloud, erogeneous, and changing data will be called polyglot data. etc), and they communicate by means of various network Quality of Service (QoS) features (such as latency, data loss, communication protocols (for example, ADSL, Wi-Fi, etc.). etc.) play a major role in IoT data architectures. Data provided Overall, agricultural IoT applications require: by a system and their QoS features are strictly related. For • the use of complex spatio-temporal data (e.g., robot example, it is likely to send images from animal drinkers trajectories, meteorological data); with different resolutions; robots can send one aggregated • the use of stream data (e.g., from sensors deployed in odometry data per minute instead of one data per second, fields) and historical data (e.g., warehoused data on all according to the available network bandwidth. This means the aspects of an IoT system). that, for each piece of data, different reliable representations Moreover, agro-ecology IoT applications seem to be more must be considered by end-users. Thus, IoT data can be challenging than in Industry 4.0 in the following aspects: represented in different ways and at different abstraction • the use of autonomous robots and vehicles that operate levels (multi-representation data) according to the physical in an uncontrolled environment; constraints imposed by the network architecture. Clearly, each • the limited computation and communication resources representation can be implemented in different ways in its (ADSL networks, low-quality Wi-fi connections, small corresponding system (e.g. DBMS, DSMS, sensors). laptops) deployed in rural areas; These polyglot and multi-representation data must be cor- • the involvement of stakeholders (such as farmers, re- related to provide a global data-centric representation of IoT searchers, managers, etc.) who have heterogeneous pro- data. These correlations raise several design and implementa- files with different knowledge and experience in smart tion issues since they can involve different data management farming (from farmers not skilled in IT to researchers in systems (collection, storage, and computation). Noticeably, robotics). according to [6], no conceptual model allows representation When dealing with IT applications that process complex of polyglot and multi-representation data. and heterogeneous data, the adoption of a conceptual design At the conceptual design level, the main research questions step using formalisms such as UML or E/R has been widely to be faced are: proved to be necessary to grant the success of projects [4]. • “How to define an integrated, polyglot meta-model that Indeed, these formalisms make the implementation and tech- conceptually represents agro-ecological data together nical issues transparent, allowing database designers and IoT with data obtained from different kinds of computations experts to focus exclusively on the functional requirements independently of all implementation details?”, Fig. 1: An example of an agricultural IoT architecture. • “How to conceptually represent each agro-ecological system is deployed must take into account the QoS features. data entity at multiple abstraction levels, and what poli- Therefore, QoS features on every layer play a major role in IoT cies should be defined to seamlessly switch from one level data architectures. Indeed, they can respect some functional to another?” and non-functional requirements, such as bandwidth, which • “WhichQoSfeaturescanbespecifiedbyend-usersduring lead to a particular placement of data and computation over design, and how to integrate them with the meta-model the different layers. For example, in the context of hard real- and with the multi-representation policies?” time applications, data and computation can be deployed at the At the implementation level, the main research questions are: edge level (for example on a robot) to improve performances. • “How to generate (semi)automatic implementations of Therefore, the research questions associated to QoS are: these polyglot and multi-representation IoT data over • “Which are the relevant QoS performance indicators different data management (collection, storage, and com- to guide the deployment and the functioning of data putation) systems and programming languages?” management systems over the different network layers • “How to choose the most suitable technology for agro- (edge, fog, or cloud): access delay, data rate, packet loss ecological data management (collection, storage, and ratio?” computation) and its deployment locations over the net- • “How to obtain these indicators in a reliable way for all work?” layers of the network?” B. How • “How to exploit these QoS indicators with regard to user experience?”. Several different data management (collection, storage, and Moreover, agro-ecology IoT data can be implemented in computation) systems have been proposed to take into account different ways and locations in the IoT architecture, and only the particularities of the data stored and queries processed at run time the best data management system configuration of (data workload). For example, to handle very high volumes of each data can be chosen to make the overall system resilient to robots data, NoSQL DBMSs seem better suited than classical network problems. Therefore, the research question associated relational ones. Therefore, in order to select the system that to the run-time execution of the application is: best fits the type of data, we can consider the workload as a • “How to define and implement an algorithm for dynam- “metadata” that must be also represented in IoT systems. To ically choosing the most suitable configuration for the the best of our knowledge, only [7] introduces the workload at overall system at run-time, making it resilient according the conceptual level, but it addresses only NoSQL document to the QoS indicators?” DBMSs. • “Which configuration mode is most suited to enhance the Thus, the research questions associated to the workload are: user experience depending on the application use cases?” • “Which workload features are relevant at design time?” III. ISSUES FOR AGRO-ECOLOGY IOT APPLICATIONS’ • “How to integrate them in the conceptual meta-model?” DESIGN AND IMPLEMENTATION C. Where In this section we present two representative scenarios, as IoT applications are characterized by a geographically- well as the requirements and some envisaged solutions for a distributed deployment of (potentially moving) devices, and a method to design and implement agro-ecology IoT applica- communication network continuum over different layers (from tions. Some relevant works are cited that could be extended edge to cloud). The network layer where the data management to meet these requirements. Figure 2 shows an overview of
no reviews yet
Please Login to review.