Agricultural Internet of Things and Decision Support for Precision Smart Farming

By Dimitrios Moshou

Agricultural Internet of Things and Decision Support for Smart Farming reveals how a set of key enabling technologies (KET) related to agronomic management, remote and proximal sensing, data mining, decision-making and automation can be efficiently integrated in one system. Chapters cover how KETs enable real-time monitoring of soil conditions, determine real-time, site-specific requirements of crop systems, help develop a decision support system (DSS) aimed at maximizing the efficient use of resources, and provide planning for agronomic inputs differentiated in time and space. This book is ideal for researchers, academics, post-graduate students and practitioners who want to embrace new agricultural technologies.

• Presents the science behind smart technologies for agricultural management
• Reveals the power of data science and how to extract meaningful insights from big data on what is most suitable based on individual time and space
• Proves how advanced technologies used in agriculture practices can become site-specific, locally adaptive, operationally feasible and economically affordable

• Language: English
• Publisher: Academic Press
• Release date: Jan 9, 2020
• ISBN: 9780128183748

Book preview

Agricultural Internet of Things and Decision Support for Precision Smart Farming

Editors

Annamaria Castrignanò

Council for Agricultural Research and Economics, Bari, Italy

National Research Council of Italy, Water Research Institute, Bari, Italy

Gabriele Buttafuoco

National Research Council of Italy, Institute for Agricultural and Forest Systems in the Mediterranean, Rende, CS, Italy

Raj Khosla

Department of Soil & Crop Sciences, Colorado State University, Fort Collins, CO, United States

Abdul M. Mouazen

Department of Environment, Faculty of Bioscience Engineering, Ghent Univesity Gent, Belgium

Dimitrios Moshou

Agricultural Engineering Laboratory - Faculty of Agriculture, Aristotle University of Thessaloniki, Thessaloniki, Greece

Olivier Naud

Olivier Naud - ITAP, Univ Montpellier, Irstea, Montpellier, France

Table of Contents

Cover image

Title page

Copyright

Contributors

Preface

Chapter 1. Introduction to agricultural IoT

1.1. Introduction section: an integrated view on precision smart farming from a multidisciplinary perspective

Chapter 2. Monitoring

2.1. Introduction

2.2. Remote sensing

2.3. Proximal sensing

2.4. Conclusions

Chapter 3. Data processing

3.1. Introduction

3.2. Statistical approach to data fusion

Chapter 4. Support to decision-making

4.1. Introduction to decision support functions

4.2. From spatial data to site-specific decisions and action

4.3. Planning and optimization

4.4. Information systems for smart farms

Chapter 5. Smart action

5.1. Implementation of variable rate application

5.2. Smart collaborative robotics and CPS for smart agriculture

Chapter 6. Economic, environmental and social impacts

6.1. Introduction to economic, environmental and social impacts of smart farming

Chapter 7. Precision farming and IoT case studies across the world

Subchapter 7.1. France – The digital Mediterranean farm in the south of France: a model farm to facilitate the appropriation of precision farming tools and methods for wine growers and advisors

Subchapter 7.2. Greece – precision agriculture in Greece

Subchapter 7.3. Italy – nitrogen fertilization based on prescription maps and on-the-go variable rate crop sensors in northern Italy maize cultivation

Subchapter 7.4. Georgia, USA – smart irrigation in Georgia, USA. A case study on cotton

Subchapter 7.5. Argentina – evolution of precision agriculture in Argentina for the last 20 years

Subchapter 7.6. Tanzania – smart agro-farming in Africa

Subchapter 7.7. Japan – smart agriculture in Japan

Subject Index

Author Index

Copyright

Academic Press is an imprint of Elsevier

125 London Wall, London EC2Y 5AS, United Kingdom

525 B Street, Suite 1650, San Diego, CA 92101, United States

50 Hampshire Street, 5th Floor, Cambridge, MA 02139, United States

The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, United Kingdom

Copyright © 2020 Elsevier Inc. All rights reserved.

No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher. Details on how to seek permission, further information about the Publisher’s permissions policies and our arrangements with organizations such as the Copyright Clearance Center and the Copyright Licensing Agency, can be found at our website: www.elsevier.com/permissions.

This book and the individual contributions contained in it are protected under copyright by the Publisher (other than as may be noted herein).

Notices

Knowledge and best practice in this field are constantly changing. As new research and experience broaden our understanding, changes in research methods, professional practices, or medical treatment may become necessary.

Practitioners and researchers must always rely on their own experience and knowledge in evaluating and using any information, methods, compounds, or experiments described herein. In using such information or methods they should be mindful of their own safety and the safety of others, including parties for whom they have a professional responsibility.

To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors, assume any liability for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions, or ideas contained in the material herein.

Library of Congress Cataloging-in-Publication Data

A catalog record for this book is available from the Library of Congress

British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library

ISBN: 978-0-12-818373-1

For information on all Academic Press publications visit our website at https://www.elsevier.com/books-and-journals

Publisher: Charlotte Cockle

Acquisition Editor: Nancy Maragioglio

Editorial Project Manager: Sara Pianavilla

Production Project Manager: Omer Mukthar

Cover Designer: Miles Hitchen

Typeset by TNQ Technologies

Contributors

Marco Acutis,     University of Milan, Agricultural and Environmental Sciences Production, Landscape, Agroenergy Department, Milan, Italy

Thomas Alexandridis,     Laboratory of Remote Sensing, Spectroscopy and GIS, School of Agriculture, Aristotle University of Thessaloniki Thessaloniki, Greece

Evangelos Anastasiou,     Agricultural University of Athens, Department of Natural Resources Management and Agricultural Engineering, Athens, Greece

Carmelo Ardito,     Computer Science Department, University of Bari Aldo Moro, Bari, Italy

Avital Bechar,     The Institute of Agriculture Engineering, Agriculture Research Organization, Volcani Center, Bet-Dagen, Israel

Eric Bourreau,     LIRMM, Univ Montpellier, CNRS, Montpellier, France

Henning Buddenbaum,     Environmental Remote Sensing and Geoinformatics, Trier University Trier, Germany

Gabriele Buttafuoco,     National Research Council of Italy, Institute for Agricultural and Forest Systems in the Mediterranean, Rende, CS, Italy

Danilo Caivano,     Computer Science Department, University of Bari Aldo Moro, Bari, Italy

Annamaria Castrignanò

Council for Agricultural Research and Economics, Bari, Italy

National Research Council of Italy, Water Research Institute, Bari, Italy

Jean-Pierre Chanet,     Irstea - Centre de Clermont-Ferrand - UR TSCF, Aubière, France

Yafit Cohen,     Institute of Agricultural Engineering, Agricultural Research Organization (Volcani Center), Rishon LeZion, Israel

Lucio Colizzi,     Computer Science Department, University of Bari Aldo Moro, Bari, Italy

Thomas Crestey,     ITAP, Univ Montpellier, Irstea, Montpellier, France

Giuseppe Desolda,     Computer Science Department, University of Bari Aldo Moro, Bari, Italy

Puwadol Oak Dusadeerungsikul,     PRISM Center and School of Industrial Engineering, Purdue University, West Lafayette, IN, United States

Guido Fastellini,     Topcon Agriculture, Turin, Italy

Spyros Fountas,     Agricultural University of Athens, Department of Natural Resources Management and Agricultural Engineering, Athens, Greece

Kadeghe G. Fue,     Centre for ICT, Sokoine University of Agriculture, Morogoro, Tanzania

Rodolphe Giroudeau,     LIRMM, Univ Montpellier, CNRS, Montpellier, France

Serge Guillaume,     ITAP, Univ Montpellier, Irstea, Montpellier, France

Kun-Mean Hou,     ISIMA, LIMOS, University of Clermont-Ferrand, Aubière, France

Gao Hui,     University Clermont Auvergne, LIMOS UMR 6158 CNRS, France

Ahmed Kayad,     University of Padua, Department of Land, Environment, Agriculture and Forestry

M. Kernecker,     Program Area Land Use and Governance, Leibniz Centre for Agricultural Landscape Research, Müncheberg, Brandenburg, Germany

Raj Khosla,     Department of Soil & Crop Sciences, Colorado State University, Fort Collins, CO, United States

Knierim,     Program Area Land Use and Governance, Leibniz Centre for Agricultural Landscape Research, Müncheberg, Brandenburg, Germany and Rural Sociology, University of Hohenheim, Stuttgart, Germany

Vasileios Liakos,     College of Agriculture and Environmental Sciences University of Georgia, Athens, GA, United States

Francesco Marinello,     University of Padua, Department of Land, Environment, Agriculture and Forestry

Maristella Matera,     Dipartimento di Elettronica, Informazione e Bioingegneria (DEIB), Politecnico di Milano, Milano, Italy

Andrés Méndez,     Instituto Nacional de Tecnología Agropecuaria INTA, Córdoba, Argentina

Francesco Morari,     Department of Agronomy, Food, Natural Resources, Animals and Environment, University of Padova, Padova, Italy

Eiji Morimoto,     Tottori University, Faculty of Agriculture Laboratory of Bio-systems Engineering

Dimitrios Moshou,     Agricultural Engineering Laboratory - Faculty of Agriculture, Aristotle University of Thessaloniki, Thessaloniki, Greece

Abdul M. Mouazen,     Department of Environment, Faculty of Bioscience Engineering, Ghent Univesity Gent, Belgium

David Mulla,     Department of Soil, Water & Climate, University on Minnesota, St. Paul, MN, United States

Olivier Naud,     ITAP, Univ Montpellier, Irstea, Montpellier, France

Said Nawar,     Department of Environment, Faculty of Bioscience Engineering, Ghent Univesity Gent, Belgium

Shimon Y. Nof,     PRISM Center and School of Industrial Engineering, Purdue University, West Lafayette, IN, United States

J.E. Ørum,     Department of Food and Resource Economics, University of Copenhagen, Frederiksberg, Denmark

S.M. Pedersen,     Department of Food and Resource Economics, University of Copenhagen, Frederiksberg, Denmark

M.F. Pedersen,     Department of Food and Resource Economics, University of Copenhagen, Frederiksberg, Denmark

François Pinet,     Irstea - Centre de Clermont-Ferrand - UR TSCF, Aubière, France

Camilius A. Sanga,     Centre for ICT, Sokoine University of Agriculture, Morogoro, Tanzania

Luigi Sartori,     University of Padua, Department of Land, Environment, Agriculture and Forestry

Calogero Schillaci,     University of Milan, Agricultural and Environmental Sciences Production, Landscape, Agroenergy Department, Milan, Italy

Hongling Shi,     University Clermont Auvergne, LIMOS UMR 6158 CNRS, France

Kenneth A. Sudduth,     USDA Agricultural Research Service, Columbia, MO, United States

James Taylor,     ITAP, Univ Montpellier, Irstea, Montpellier, France

Bruno Tisseyre,     ITAP, Univ Montpellier, Irstea, Montpellier, France

Siza D. Tumbo,     Ministry of Agriculture, Dodoma, Tanzania

F van Egmond,     Wageningen Environmental Research, Wageningen University & Research, Wageningen, The Netherlands

F.K. van Evert,     Agrosystems Research, Wageningen University & Research, Wageningen, The Netherlands

Juan Pablo Vélez,     Instituto Nacional de Tecnología Agropecuaria INTA, Córdoba, Argentina

George Vellidis,     College of Agriculture and Environmental Sciences University of Georgia, Athens, GA, United States

Preface

Supplying high-quality, safe food to a very complex and diverse marketplace is a very demanding challenge for our farming industries across the globe. Our biological systems are being further challenged in terms of long-term environmental sustainability and increasing demand for produce from a reducing resource base. Accelerated climate uncertainty brought about by global warming will further exacerbate the problem. If we are to avoid being overtaken by these challenges, we must deploy improved methods derived from enriched understandings. One aspect of that is the application and use of a range of well-integrated technologies to better inform our decision-making.

This book presents a vision of SMART agriculture where linked sensors can be used to gather meaningful real-time data to be utilized within a number of approaches to data processing and modelling to produce high-quality and reliable information for decision-making. The book follows a logical pathway to demonstrate how data contribute to a converging flow of information towards decision support system and how it can be transformed into actionable steps.

Much of the SMART vision is presented with the idea of being strongly integrated round concepts of intelligent, fast, technologically, economically, politically and culturally sustainable practices and systems. The book further develops ideas surrounding strong integrating ICT and IoT to manage rural assets, arguing that the core of SMART farming is the adoption of technologies and methods to deliver improved economic and environmental performance in a spatially and temporarily variable environment. A novel Collaborative Control Protocol for Robotics is also presented, aimed at ensuring that humans, robot and sensors perform the agricultural tasks in harmony.

A large number of well-known authors have contributed with their work to this book and it is presented in logical sequence demonstrating how data can be transformed into action. There are a large number of interesting examples of SMART farming presented from around the world.

The book is aimed at a wide audience including researchers, academics, postgraduate students and practitioners; indeed, anyone who is considering the use and application of a range of technologies in our food production system would greatly benefit from this stimulating work.

Chapter 1

Introduction to agricultural IoT

Lucio Colizzi ¹ , Danilo Caivano ¹ , Carmelo Ardito ¹ , Giuseppe Desolda ¹ , Annamaria Castrignanò ² , ³ , Maristella Matera ⁴ , Raj Khosla ⁵ , Dimitrios Moshou ⁶ , Kun-Mean Hou ⁷ , François Pinet ⁸ , Jean-Pierre Chanet ⁸ , Gao Hui ⁹ , and Hongling Shi ⁹       ¹ Computer Science Department, University of Bari Aldo Moro, Bari, Italy      ²Council for Agricultural Research and Economics, Bari, Italy      ³ National Research Council of Italy, Water Research Institute, Bari, Italy      ⁴ Dipartimento di Elettronica, Informazione e Bioingegneria (DEIB), Politecnico di Milano, Milano, Italy      ⁵ Department of Soil & Crop Sciences, Colorado State University, Fort Collins, CO, United States      ⁶ Agricultural Engineering Laboratory - Faculty of Agriculture, Aristotle University of Thessaloniki, Thessaloniki, Greece      ⁷ ISIMA, LIMOS, University of Clermont-Ferrand, Aubière, France      ⁸ Irstea - Centre de Clermont-Ferrand - UR TSCF, Aubière, France      ⁹ University Clermont Auvergne, LIMOS UMR 6158 CNRS, France

Abstract

Significant challenges will have to be overcome to achieve the level of agricultural productivity necessary to meet the predicted world demand for food, feed, fibre and fuel in 2050. Although agriculture has met significant challenges in the past, targeted increases in productivity will have to be made by 2050, in the face of stringent constraints including limited resources, less skilled labour, limited amount of arable land and changing climate, among others. Currently, agriculture production accounts for over 70% of freshwater consumption and unsustainable levels of chemical consumption for crop production. In the hyperconnected world, where people, computers and physical objects cooperate to solve complex tasks, a big amount of data and information rises rapidly and a critical aspect is to manage that knowledge to make the right decision at the right time and the right place. Also, farming has to become SMART adopting a new vision of the primary production sector where the development processes are based on the integration of information and communications technologies and Internet of Things technologies in a secure fashion to manage the rural assets and optimization of agronomic inputs such as water, fertilizer, agrochemical or soil tillage and to enhance input use efficiency, output or production and profitability in a sustainable manner. In this vision, the land becomes a substrate where different kinds of sensors could acquire heterogeneous data. Those sensors are connected in a sort of rural network in turn linked to the Internet network. The real-time streaming data are stored in complex database containing all the necessary knowledge about the land characteristics. Intelligent programmes connected with the knowledge base run to make real-time decisions, sending acting messages to the domotic back-end system or suggestions to the farmer.

Keywords

Arduino; Internet of things; Open source platform; Raspberry pi; Sensors; Smart farming; Smart object

1.1 Introduction section: an integrated view on precision smart farming from a multidisciplinary perspective

1.1.1 Internet of things architectures and paradigms

1.1.1.1 Physical devices and controllers

1.1.1.2 Connectivity

1.1.1.3 Edge computing or cloud computing

1.1.1.4 Data accumulation

1.1.1.5 Data abstraction

1.1.1.6 Application layer

1.1.1.7 Collaboration and processes

1.1.2 Open source internet of things platforms

1.1.2.1 Arduino

1.1.2.2 The solenoid valves

1.1.2.3 Relay

1.1.2.4 Moisture sensor

1.1.2.5 Rainfall sensor

1.1.3 From an object to a smart object

1.1.3.1 State of the art of smart objects

1.1.3.2 State of the art of cognitive techniques

1.1.3.3 State of the art of smart objects platform

1.1.3.4 Smart object platform and use cases

1.1.3.5 Smart object interoperability

1.1.4 Developing internet of things applications

References

Further reading

1.1. Introduction section: an integrated view on precision smart farming from a multidisciplinary perspective

According to the recent report by FAO, the world's population will surpass 9.0 billion people by year 2050 (FAO, 2009). Significant challenges will have to be overcome to achieve the level of agricultural productivity necessary to meet the predicted world demand for food, feed, fibre and fuel in 2050. Although agriculture has met significant challenges in the past, targeted increases in productivity will have to be made by 2050, in the face of stringent constraints including limited resources, less skilled labour, limited amount of arable land and changing climate, among others. For most of the 20th century, many key factors influenced increases in the rate of crop production, primarily mechanization, improved genetics and increased use of inputs. However, such increase in crop production came at a cost of overapplication of various agricultural inputs, i.e., irrigation, nutrients and pesticides. The use of resource-intensive, high-input agriculture around the world led to depletion of soils, water scarcity, widespread deforestation and high levels of greenhouse gas emissions (FAO, 2017; NASEM, 2019 ). Currently, agriculture production accounts for over 70% of freshwater consumption and unsustainable levels of chemical consumption for crop production. Hence, sustainability in agriculture is a must that is becoming a need due not only to the scarceness of natural resources and the growth of population but also for the growing attention deserved to well-being and green lifestyle. Agriculture needs to provide effective solutions to old and new challenges to embrace the insights from other disciplines and use them in an integrated way.

Precision agriculture (PA) presents itself as one among many solutions to the grand challenges that agriculture and our world are currently facing. PA has been around for the past three decades and has established itself as a management approach that harnesses the heterogeneity in both space and time and in production fields to deploy its simple yet effective approach of applying the right input at the right time, at the right place, in the right amount and in the right manner—the five 'R' concept of PA (Khosla, 2010). Over the years, PA has grown worldwide and is slowly embracing newer technologies that are autonomous, disruptive and data-intensive. The first decade of PA had a strong focus on Global Navigation Satellite Services (GNSS) and its ability to locate and quantify spatial variability in soils. The second decade focused on tractor automation and developing technologies that would allow precision management of inputs, such as crop nutrients. Now, in its third decade, there is an exponential increase in collection of location-based agricultural data via suite of sensors and sensing devices that created a new paradigm of making management decision based on evidence for higher degree of precision management. Hence, the success of future farming practices, i.e., output, efficiency and sustainability, would rely heavily on 'farming the data' as much as 'farming the land'.

Recently, the International Society of Precision Agriculture (ISPA) completed a global project to co-create the definition of PA. Input was received from about 50 notable researchers from around the world. According to the co-creation process, the top ranked definition of PA is 'Precision Agriculture is a management strategy that gathers, processes and analyses temporal, spatial and individual data and combines it with other information to guide site, plant or animal specific management decisions to improve resource efficiency, productivity, quality, profitability and sustainability of agricultural production'.

Today's PA is a perfect example of intersection of agriculture and information technologies. It increasingly depends on the collection, transfer and management of information by information and communications technologies (ICT) to drive increased productivity. What was once a highly mechanical system is becoming a dynamic cyber-physical system (CPS) that combines the cyber or digital domain with the physical domain.

In the last decade, the scientific and industrial communities fostered the emergence of the so-called 'Smart Vision' where 'smart' means integrated, intelligent, fast, technologically, economically, politically and culturally sustainable, doing more with less, improving the quality of life of all. This new vision has determined a convergence of disciplines in the problem-solving process.

In the hyperconnected world, where people, computers and physical objects cooperate to solve complex tasks, a big amount of data and information rises rapidly and a critical aspect is to manage that knowledge to make the right decision at the right time and the right place.

The most significant example in this sense is the 'smart cities', an urban development vision to integrate ICT and Internet of Things (IoT) technologies in a secure fashion to manage a city's assets. These assets include local departments' information systems, schools, libraries, transportation systems, hospitals, power plants, water supply networks, waste management, law enforcement and other community services. A smart city is promoted to use urban informatics and technology to improve the efficiency of services. ICT allows city officials to interact directly with the community and the city infrastructure and to monitor what is happening in the city, how the city is evolving and how to enable a better quality of life. Through the use of sensors integrated with real-time monitoring systems, data are collected from citizens and devices and then processed and analyzed. The information and knowledge gathered are keys to tackling inefficiency. ICT is used to enhance quality, performance and interactivity of urban services, to reduce costs and resource consumption and to improve interaction between citizens, stakeholders and government. Smart city applications are developed to manage urban flows and allow for near real-time responses.

Also, farming has to become SMART to overtake the today's challenges, and the IoT model (that is now evolving toward Internet of Everything ((IoE)) may represent the right architecture to reorganize farming and all the disciplines and technologies involved in the smart way.

Transferring what is already realized in smart city to agriculture means introducing a set of innovative methodologies and connected technologies with the aim to optimize the agronomic inputs such as water, fertilizer, agrochemical or soil tillage and to enhance input use efficiency, output or production and profitability in a sustainable manner.

Smart farming means to adopt a new vision of the primary production sector where the development processes are based on the integration of ICT and IoT technologies in a secure fashion to manage the rural assets.

In this vision, the land becomes a substrate where different kinds of sensors could acquire heterogeneous data. Those sensors are connected in a sort of rural network in turn linked to the Internet network. The real-time streaming data are stored in complex database containing all the necessary knowledge about the land characteristics. Intelligent programmes connected with the knowledge base run to make real-time decisions, sending acting messages to the domotic system or suggestions to the farmer.

An example of CPS which covers an important role in the vision described above is 'precision farming' (Fig. 1.1).

The intent of PA is to match agricultural inputs and practices with the local conditions within a field; therefore, it becomes crucial to obtain more and better information so that farmers can achieve better decisions and accomplish their many different goals more efficiently.

The core of precision farming is the application of technologies and methods to effectively manage spatial and temporal variability associated with all aspects of agricultural production for the purpose of improving crop performance and environmental quality. Without variability, the concepts of precision farming would have little meaning (Mulla and Schepers, 1997) and would never have evolved. Therefore, any precision farming system must first focus on measuring and understanding both spatial and temporal variability.

The foundation of precision farming rests on geospatial data techniques for improving the management of inputs and documenting production outputs. The predominant control strategies for these systems are based on management maps developed by farmers and their crop consultants. PA techniques focus on the existence of in-field variability of natural components, including chemical leaching, runoff, drainage, water content, nutrients and soil components. The goal is to use new technologies, such as GNSS, satellites, aerial remote sensing and proximal sensors and sensor networks to assess the variations in a field more rapidly, reliably, in situ, nondestructively, in real time, at spatially dependent scales and accurately (NASEM, 2019). Accordingly, farming practices, including land preparation, sowing, irrigation, fertilization and management and pest control, can be scheduled autonomously with the aid of autonomous machines and robots, according to the assessment of the field. A competitive technology for map-based precision farming is on-the-go sensing systems based on the concept of machine-based sensing of agronomic properties (plant health, soil properties, presence of disease or weeds, etc.). The immediate use of these data drives control systems for variable rate applications. These sensor capabilities essentially turn the agricultural vehicle into a mobile recording system of crop attributes measured across the landscape. The concept of precision farming as a CPS consists of the following:

Figure 1.1 Precision farming as a cyber-physical system.

• wireless sensor networks (WSNs) (IoT),

• information and data fusion,

• decision support intelligence and

• actuators for applications of inputs.

While the technology can facilitate the application of precision farming, it is only the knowledge and interpretation of variability and its management, in terms of site-specific agronomic recommendations, that makes precision farming feasible. From what said above, it is clear that smart precision farming is much more than a mere set of even advanced technologies but essentially involves extensive changes in management style and strategy and causes a real revolution in mind.

Moreover, owing to the complexity of several factors affecting crop production, most of them still unknown, only an interdisciplinary systemic approach can be adopted.

To overcome the limitations of spatially scarce data, advances in proximal sensing technology and data processing techniques are now able to provide information on soil, crops and associated environmental properties. Currently, the number of proximal sensing techniques has increased because of the advantages of noninvasive techniques being time- and cost-efficient. These sensors will produce large volumes of data that have to be collected, stored, shared, processed, analyzed, fused and interpreted for translating data into new knowledge and action (Fig. 1.2). Sensor data produced by a single sensor will not provide the relevant information that can be used to fully understand the situation. Therefore, sensor data collected through multiple sensors need to be fused together, processed and understood later.

The integration of wireless sensors with agricultural mobile apps and cloud platforms helps in collecting relevant information pertaining to the environmental conditions—temperature, rainfall, humidity, wind speed, pest infestation, soil humus content or nutrients—of the farmland, which can be used to take informed decisions aimed at improving quality and quantity of production and minimizing risks and wastes. The app-based field or crop monitoring also lowers the hassles of managing crops at multiple locations. For example, farmers can now detect which areas have been fertilized (or mistakenly missed) or need to be irrigated and estimate impact of their practices on future yields.

Figure 1.2 Agricultural internet of things model.

The agricultural IoT model can be sketched as in Fig. 1.2.

Another key point of this model is that 'decision-making' process uses dynamic models that evolve according to the experience collected over time. This allows the growth of an agricultural body of knowledge according to the empirical evidence collected, so as to update the best practices consequently.

The book organization will follow the same logical path.

The issues of each chapter and its sections will be treated not only from the technological point of view but also it will be showed how the data contribute to the flow of information converging on decision support system (DSS) to be transformed into action or recommendation.

1.1.1. Internet of things architectures and paradigms

The IoT paradigm is based on the concept of a pervasive network capable of connecting not only people but also objects and systems.

The term was coined in 1999 by Kevin Ashton and represents a domain of technology in which it is possible to imagine a global network that makes connected, enabling their cooperation, millions of objects (wearable gadgets, logistics and transport systems, everyday used devices, appliances, buildings and their subsystems, home automation modules, sensors, actuators, medical devices, etc.). When the term was coined, it was thought to be a kind of interconnection between electronic devices. The concept has evolved over the years to converge into the more general paradigm defined as IoE where in the global network there are not only things and devices but also people, processes and data.

It is almost impossible to establish how many objects are connected today even if some studies report numbers like 17 billion that could become over 40 billion in 2020 with a double-digit annual growth rate that will bring the connected devices in 2030 to be over 125 billion (source IHS Markit). In reality, there are many other objects connected but not counted because they are cooperating devices in a local network not exposed to the Internet. In this area, it can only be said that there is a growing use of hundreds of network objects in constant increase.

Within the IoT paradigm, a connected object is also defined as Smart, if, in some way, it is able to process its state or the world around it and consequently make decisions, which can be very simple (sorting data) or even complex (DSS).

Like the ISO/OSI reference model is the architecture for securely connecting a computer to the Internet, in the same way, to identify the main features of the Internet of Things paradigm, it is necessary to refer to a model that clarifies its elements at various levels. In recent years, given the growing interest in this sector, several reference models have been proposed and, for the objectives of this book, it will be adopted as the one defined by Cisco (IoE, http://internetofeverything.cisco.com) in the context of the World Forum of the Internet of Things.

Figure 1.3 CISCO internet of things (IoT) reference model – IoT world forum.

The model in question identifies seven levels, reported in Fig. 1.3.

Below are provided some details about each level, giving keywords for any further insights that must necessarily be carried out on the Internet documentation source, given the continuous evolution of the topics in question and their speed of obsolescence.

1.1.1.1. Physical devices and controllers

In the first level there are placed the 'Things'. The layer name also includes its controllers, but it should be better specified. An IoT object is not only the measurement/actuation/processing device but also the technological apparatus that allows it to work in the place where it operates. Moreover, an 'object' of the Internet of Things can be the single sensor with its driver (i.e., the hardware and/or software device used to transcribe its signal in digital format) or a complete device where the hardware part can be also a workstation or an industrial apparatus. The IoT object is able to analyze data from the surrounding environment and/or to perform actions on components by means of actuators. Based on this definition, an object in the Internet of Things is therefore an entire system integrated into an Internet of Things platform and designed to interface directly with sensors and actuators.

1.1.1.2. Connectivity

This is the level that can connect IoT objects to the network. The concept of networking underlying this level involves a variety of methodologies and consequently technologies for implementing connectivity between IoT resources. An object can be directly connected to the control system even in 'wired' mode; more often the access point to the IoT platform is implemented as a cloud service theoretically able to hold together a large number of geographically distributed objects. It is important to understand that the decision-making process in an IoT platform is not physically conveyed to one place but is a 'liquid' process allocated to various levels of complexity. To better understand this concept, let us take a case study. Imagine building an actuator node on a solenoid valve (normally closed) when a hygrometric sensor, located in an area adjacent to the solenoid valve, reads a signal below the λ threshold. If the local sensor controller is a board capable of performing a simple control such as 'if, then, else', then the decision whether to open or not the solenoid valve can be taken locally and then the status (open/closed) can be communicated to the IoT platform. However, the decision-making level could be more complex. Let us suppose that the decision-making also includes the weather forecast. In this case, the local node cannot open the solenoid valve because it is not able to know if it will rain or not shortly. To solve the problem, the decision is taken at the platform level, so the sensor node will communicate the overall IoT system, the measurement of the hygrometer. The system will then access to a cloud service to know the weather forecast and compare it, according to a specific model, with the soil moisture taken with the remote sensor. At this point the system will command the solenoid valve to open or remain closed.

As is evident from the example (that will be better developed after), the network topology of the IoT platform is often presented as a mix of solutions and architectures because the decision-making process is a distributed process.

1.1.1.3. Edge computing or cloud computing

In this important level there is the full mapping of cloud services available to implement the requirements of an IoT platform. Technically it is defined 'orchestration' of services, the particular process implementing the policies of combination of services available on the network, to match the IoT system requirements. In the previous example of the solenoid valve to be controlled, the weather service might be a web service that can be invoked in pay-per-use mode to know the weather forecast in any area of the planet, known its geographical coordinates. This modality is defined Software as a Service. With similar mechanisms, entire software platforms (Platform as a Service) can be made available in the cloud. Current technologies to implement this level are Jason and the Rest API (application programming interface).

1.1.1.4. Data accumulation

Hundreds of thousands of objects connected to the network generate a huge amount of data, and these must somehow be stored and then processed as it will be discussed with case study in Chapter 4 of this book; there are well-established methodologies for organizing data, such as the theory of relational databases that can be interrogated using languages created ad hoc as SQL. In recent years, however, the need has arisen to store huge amounts of data in real time efficiently. To meet these needs, NoSQL approaches and distributed file systems for the management of Big Data were born, the current reference technologies are Mongo DB, Hadoop File System and Cassandra DB.

1.1.1.5. Data abstraction

In this level methodologies and technologies for assigning a meaning to the data are found. It is well known that the data itself do not represent anything. When the datum answers a question, then it is possible to say that datum becomes information. The right amount of information, needed to make a decision, is defined as 'knowledge'. At this level, methodologies for knowledge organization are implemented. The technological field in which the most widespread knowledge in the world is organized is certainly the semantic web. From the technological point of view, the Resource Description Framework (RDF) standard has been proposed by W3C as a series of declarative languages based on XML8 syntax. RDF is suitable to describe the structure of any identifiable resource in the network with a unique address (IPv4 or Ipv6). RDF is also a tool developed for encoding and exchanging structured metadata allowing interoperability between multiple applications that share information on the web (more details in Section 1.2). Many research groups are working on the semantic description of IoT objects.

1.1.1.6. Application layer

The name of this level is self-explanatory. It is the level in which applications are available, i.e., the set of features obtained, for example, from the combination of services and web platforms useful to achieve a given application goal. In this context, technological development is evolving on the concept of 'mashup', or methods and technologies to obtain software applications via the web through the paradigm of plug and play. According to this paradigm, it is necessary to minimize the development of new software by developing reuse techniques in which recurring software elements (patterns) can be combined even without being specialists in the development of applications. The extreme case is when the end user, using a mashup platform, self-produces his/her own application focussing only on the human–machine interaction component.

1.1.1.7. Collaboration and processes

IoT objects and resources are generally almost always managed within one or more business processes. This means that, together with the other elements of the platform, they contribute to the achievement of well-defined business objectives. Methodologies to define, optimize, monitor and integrate business processes are part of the business process management techniques Therefore, this level includes all the technologies useful for the temporal evolution of the IoT platform status and for collaboration (in time and space) between the resources involved. The digital species of this level are therefore the workflow management system, the DSS, the systems for process simulation (e.g., discrete-event simulation).

1.1.2. Open source internet of things platforms

There are two possible ways for developing or concretely using an IoT system dedicated to smart agriculture: acquiring a technology developed ad hoc or tapping into the countless opportunities that the open source world offers.

While the first way, certainly more professional, involves major investments and therefore it is a prerogative of medium and large companies that can afford them, the second is an option that even small companies with staff with a strong aptitude for Do It Yourself could activate.

In any case, a greater understanding of the potential of more complex technologies, in terms of decision engines, can be supported by the preliminary familiarity that can be gained with open source technologies on small pilot projects.

This paragraph focuses on the second way, more interesting for the objectives that this book wants to reach, which is to apply the potentialities of the IoT technologies in agriculture. It will be shown in which environments it is possible to develop ideas and make them available to the entire community or use them to have a real competitive advantage.

First, let us try to frame the problem. Implementing precision irrigation or smart fertigation means making a series of measurements in the field, adding other useful data and making them available, in real time, to a decision-maker (DSS) that, when certain conditions occur, schedules the spreading of agronomic inputs, i.e., water and/or fertilizer.

What has been described could be outlined with a loop like the one shown in Figs. 1.4 and 1.5.

The system must be able to monitor quantities and make decisions based on what it is measuring. For example, if the monitored parameter is soil moisture, then there will be a moisture sensor plugged in the soil that is periodically checked. The value read by the sensor is sent to the decision-maker system which, on the basis of precise rules, will decide if, where and how much to irrigate. This decision is sent to the actuation system which will proceed to execute the order by activating one or more solenoid valves as indicated by the DSS.

The three blocks such as measurement, decision and actuation can be as simple as the use case just described, but they can also be very complex. The measurement block might monitor multispectral or hyperspectral data. It might work on single points or spatially resolved data. Typically, as reported in the previous paragraph, these data never reach the decision-maker alone, but they are 'conditioned' by a series of other data (metadata) that characterize them (e.g., georeferencing, technical characteristics of measuring instruments, type of calibration, sensor behaviour curves, status information, etc.).

The DSS might make simple decisions (e.g., activate the solenoid valve if the humidity is below a threshold), might implement complex decision algorithms, might use real-time data and historical data, or might access to local data or data provided by web services (e.g., weather data, historical data, statistical data, etc.).

Even the actuation system might be more complex than a simple command system, including procedures to optimize the use of energy, distributed intelligence, wireless systems, ad hoc networks, etc.

From this point onwards, a smart irrigation system will be considered, starting from a real and very simple use case. Consequently, even the underlying decision-maker system will reflect this simplicity. At the end it will be added elements to the decision-making process which, to be implemented, will necessarily have to draw on complex services and perhaps geographically distributed. Once this basic mechanism is understood, it will be more understandable to place the technologies and methodologies that will be discussed in detail in the following chapters, which will reflect the current state of the art of the sector.

Figure 1.4 The decision support system process.

Figure 1.5 The Arduino UNO device.

1.1.2.1. Arduino

An IoT system based on open source and low cost technologies must provide specific modules for each of the logical blocks indicated in the previous loop (Fig. 1.4). A technological tool of this family is certainly the card developed in Italy called Arduino.

Arduino (Fig. 1.5) is an open source hardware platform composed of a series of electronic boards equipped with a microcontroller. It was conceived and developed by some members of the Interaction Design Institute of Ivrea (Italy) as a tool for rapid prototyping and for hobby, educational and professional purposes.

With Arduino it is possible to realize, in a relatively quick and simple way, small devices such as light controllers, speed controllers for motors, light sensors, automatisms for temperature and humidity control and many other projects that use sensors, actuators and communication devices. It is combined with a simple integrated development environment (IDE) for programming the microcontroller (Arduino IDE). All the software supplied is free, and the circuit diagrams are distributed as free hardware.

If the aim is to build a simple IoT system with Arduino and apply it to smart fertigation, for instance, it is proposed the following loop scheme:

The Arduino board in the previous loop (Fig. 1.6) could be unique for small projects because it is perfectly able to handle both input and output signals (analog and digital). The moisture sensors (one or more of them) are connected to the Arduino board which, after data preprocessing, delivers the soil water content values to the DSS. This is a software that is run on a PC/server/workstation locally or remotely (in this case Arduino must be equipped with a WiFi shield). The DSS returns Arduino the decision of which solenoid valves to activate. The Arduino will actuate on the solenoid valves mounted on a manifold using relays.

The connection not only between Arduino and the workstation but also between Arduino and the sensor can be wired but also wireless. In the latter the resulting IoT system is certainly more complex, as each sensor node becomes a node of a real WSN, but at the same time, the result is a flexible system able to adapt to the many agronomic activities (Fig. 1.7).

Figure 1.6 Arduino UNO integration schema for internet of things irrigation/fertigation

To give concreteness to what has been reported up to now, let us consider a more complex IoT system for irrigation. Following it will be showed how to progress by adding ingredients to get closer to the smart agriculture concept. The individual elements of the system and how they can be integrated with each other will be described below.

1.1.2.2. The solenoid valves

A solenoid valve could be assimilated to a tap where the action of closure and/or opening is not left to a human being but to an implementation of electromechanical type. The most common solenoid valves are equipped with a chamber divided into two compartments by a diaphragm. The diaphragm guarantees the tightness between compartments. A solenoid (electromagnet) is connected to the diaphragm which, when supplied, e.g., with a voltage of 24   V dc, raises it and allows water to flow. A solenoid valve is defined as 'normally closed' (NC) when, in the absence of voltage at the ends of the solenoid, the valve is at rest (closed) and does not allow the flow of liquid (e.g., water) to pass between the inlet and outlet; conversely, it is called 'normally open' (NO).

Figure 1.7 Arduino-based wireless sensor node architecture.

Figure 1.8 Relay module for Arduino.

1.1.2.3. Relay

The relay is a switch operated by an electromagnet consisting of a coil of wire wound around a core of ferromagnetic material (Fig. 1.8). As the current passes, the tabs are attracted to each other and the circuit becomes closed.

Fig. 1.8 shows a relay module that can be controlled by the Arduino board with the corresponding connection diagram. As the solenoid valve must be supplied with 24V DC, it is necessary to draw this voltage from the mains using a transformer. In the real world, of course, it is rare to have to deal with a single solenoid valve to be controlled (the reasons will be clear in the following chapters) but with an array of solenoid valves. In this case, the open source world provides an array of relays like the one shown in Fig. 1.9, which allows to control several solenoid valves simultaneously.

Figure 1.9 Eight-relay array.

The Vcc: 5V, Gnd and In1 connections are made directly on the Arduino board; the first two will power the relay while In1 will have to be connected to a digital PIN (e.g., pin number 13) and will allow to drive the opening or closing of the relay circuit.

1.1.2.4. Moisture sensor

The open source and low-cost world offers different sensors for a few euros. Their accuracy is not extremely high but for real applications the margin of error they provide is (in a wide range of cases) irrelevant. One of the most important sensors is the one that measures the water content of the substrate, the moisture sensor. Figs. 1.10–1.12 shows the complete device.

It consists of a component, which has to be plugged in the soil, and a driver that allows to read the value of moisture and transform it into an analog signal or a digital signal (wet/not wet) on the basis of a threshold adjustable on the driver board through an on-board trimmer (blue component on the board).

Figure 1.10 Moisture sensor with its driver.

Figure 1.11 Rain sensor.

Figure 1.12 Hardware connection schema for smart irrigation fertigation with Arduino UNO board.

A cheap sensor of this type cannot obviously provide a calibrated signal. This operation should be done empirically by associating a certain level of moisture induced on a sample to the corresponding voltage read on the analog pin. After that, all sensors of the same type can be used with the same calibration. For this type of sensor it is important to respect the polarity in the connection between the actual sensor and the driver board.

1.1.2.5. Rainfall sensor

It is a low cost sensor that has the same components of the moisture sensor seen above.

The actual sensor (Fig. 1.11) is a passive component that changes its resistance depending on the contact with water. The driver board allows to vary the sensitivity in this case through a trimmer on it. The driver board output can be either a digital signal (rain/no rain), whose threshold depends on the position of the trimmer, or an analog signal that can vary from 1024 (no rain) to 0 (completely submerged in water). As far as the calibration of the sensor is concerned, the same considerations as for the moisture sensor apply. Also, for this type of sensor, it is important to respect the polarity in the co-location between the sensor and the driver card.

The complete (Fig. 1.12) then consists of

- solenoid valve

- relay

- hygrometric sensor

- rain sensor

- transformer (220V AC, 24V DC)

- Arduino board (UNO).

However, it is necessary to focus on a particular aspect. Arduino board can be seen as a minicomputer that has the ability to send and receive analog and digital signals and also to perform some processing. To do this, the card must run a programme. This software must be written using an IDE, which is a programming environment containing all software libraries (API). The IDE for developing software for Arduino is free and can be downloaded from the Internet network. The software written for Arduino is called SKETCH. Once the sketch has been written, it must be downloaded to the board via a PC-Arduino connection, which is done with a simple serial cable with USB connector. After downloading the sketch to the board, the board will run the programme as soon as it is powered.

As for rainfall sensor, let us assume that the sketch programme has a very simple decision-making process (DSS):

• Step 1: Arduino interrogates the rainfall sensor with a given frequency (e.g., one sample every 5s).

• Step 2: If the values of the parameters indicate no rain and the conditions of irrigation planning are met, then Arduino asks the relay to open the solenoid valve because it is time to irrigate.

• Step 2bis: If the values of the parameters indicate rain, then Arduino must instruct the relay to close the solenoid valve anyway.

• Step 4: Back to Step 1.

The UML activity diagram notation is used to represent the process just described.

Figure 1.13 Simple decision support system UML activity diagram.

As it is showed in Fig. 1.13, the flow reports the work of each resource involved. At the same time, the decision engine within the red line dotted box is shown. Adding also the logic of the ground sensor, the flow diagram must be modified, but in any case the DSS will consist of a composition of assertions of the type: IF  THEN     ELSE   .

To use the full potential of an IoT infrastructure, the device must be part of a network of cooperating objects.

As it will be pointed out in the following chapters of the book, highly advanced sensors are used to make very complex measurements. These measures could be combined to build the appropriate knowledge and then make 'reasonable' decisions. Examples might be multi- and hyperspectral measurements, thermographic measurements, georadar measurements, etc. In the same way, it will be possible to build DSS systems able to use complex methodologies based on artificial intelligence, geostatistics, heuristic-based optimization systems, etc.

But to make these elements fit into a real IoT-type system for smart agriculture, aimed at understanding the needs of the complex real world, the concept of DSS becomes distributed and can no longer be physically confined to a single place.

By adding a key element, which is a shield (i.e., a card that extends the potential of Arduino) able to connect an Arduino card to the WiFi network, the resulting device could be defined smart object (SO) or an object that is able to cooperate with other systems and objects to achieve a common goal.

In fact, by opening a WiFi communication window with a workstation connected to the network, it will be possible for the Arduino board to leave the decision-making to other subjects more specialized in doing this work, sending the raw or preprocessed data coming directly from the field. The workstation can in turn process and decide or request the advanced services on the network using SOA (service-oriented architecture) calls or REST calls with an XML or similar data format (e.g., Jason).

The activity flow is therefore modified to implement a distribute decision logic. In Fig. 1.14 there is a generalization of the activity flow, where the level of request for remote services can be theoretically grafted onto infinity.

Once equipped with a WiFi module, the system becomes a network of nodes and can communicate not only with PC, server and workstation but also with other similar nodes to synchronize their activities in complex processes.

The WiFi shield, to be integrated in the Arduino board, looks like the one in Fig. 1.15, known by the name of ESP8266. Fig. 1.15 also shows a way to connect the ESP8266 to the Arduino board. Here, the connections are made through a matrix board (white component in Fig. 1.15), while in the real case the connections can be made directly between the components.

So far the location of the essential elements of an intelligent agriculture-oriented IoT platform has been described. Anyone can build a control module for an array of solenoid valves that can be activated appropriately by using an open source board. In the following chapters, some examples of actual platforms developed with these logics will be described, integrated in complete information systems for smart farming domain.

There are other controller boards, more sophisticated than Arduino, that allow one to get real smart IoT objects. These boards enable to manage input/output channels with a real operating system. One of the most known technologies in this area is certainly the Raspberry Pi, a minicomputer which will be described in more details in the next paragraph.

Figure 1.14 Distributed decision support system UML activity diagram.

Figure 1.15 ESP8266 WiFi module for Arduino and connections schema.

1.1.3. From an object to a smart object

As previously mentioned, the use of cognitive technologies can help understand, learn reason, interact and thus increase efficiency. Cognitive IoT technologies allow many types of correlations of large amount of structured and unstructured data from multiple sources, such as historic weather data, social media posts, research notes, soil information, market place information, images, etc., to extract knowledge and provide organizations with richer insights and recommendations to take action and improve yields (Irimia, 2016).

In general, the term IoT devices or cognitive IoT refers to smart object (‘SO’) and it is not well defined. In this section the concept of object will be more fully investigated, as well as its relationship with the SO.

In computer science (object oriented programming and database), an object can be a variable, a data structure, a function or a method and as such a value in memory referenced by an identifier (Wiki, 2018).

Thus, an object is a well-defined passive entity to be manipulated by a programme. Usually, the main object feature is its interface showing its available properties and its possibilities to communicate by messages and to interact with other entities in a synchronous or asynchronous manner. These messages can trigger the execution of object operations. The communicating objects have been defined as structures to encapsulate the complexity and the technical details of their implementation and their behaviour. From the outside, an object can be viewed as a black box providing different services (their operations). Over years, different formal techniques have been used to model and implement communicating systems, including embedded systems, using an object-oriented paradigm (Gherbi and Khendek, 2006). The goal of some communicating object techniques had been the one to make possible the communication through the web, such as CORBA (Pope, 1998), Remote Method Invocation (Jaworski, 1999), etc. In these technologies, object request brokers can be used to make the communication between objects possible in a distributed environment for the remote object service discovery and call. Nowadays, object is defined in different computer science fields and recently object may be referred as small form factor networking embedded system: IoT.

What is smart and nonsmart object?

Different definitions of SO were given related to the background and perspectives of the researchers (Want et al., 1999; Beigl et al., 2001; Mattern, 2003; Streitz et al., 2005; Fortino et al., 2018). Giancarlo Fortino et al. define an SO as an autonomous, physical digital object augmented with sensing/actuating, processing, storing and networking capabilities. SOs are able to sense/actuate, store and interpret information created within themselves and around the neighbouring external world where they are situated, act on their own, cooperate with each other and exchange information with other kinds of electronic devices and human users (Fortino et al., 2018). From the point of view of artificial intelligence and distributed computing, an intelligent agent has the same characteristics as a SO (Fortino et al., 2018).

Notice that an IoT device is not an SO, if it lacks cognitive abilities to cope with environmental changes and interact with humans for acting. Therefore, a basic SO must have the following components:

- A nonempty set of sensors to sense environment status;

- A nonempty set of actuators or actuator interfaces to control environment status;

- A processing unit running a cognitive or decision-making programme;

- A network interface for message exchange with other SOs in the vicinity or with remote server or user.

A deployment of platform of SOs for a smart farming application including local or/and remote servers is illustrated in Fig. 1.16.

1.1.3.1. State of the art of smart objects

Due to resource constraints, SO is designed to fulfil a specific task such as smart irrigation. Therefore, it is important to investigate the cognitive techniques that can be embedded into an SO platform (resource constraints).

The question regards the type of intelligence of an SO or the kind of task that an SO can fulfil for smart farming application.

1.1.3.2. State of the art of cognitive techniques

The research in the field of ‘Artificial Intelligent’ (AI) was started officially in 1956, and Minsky, McCarthy, Newell and Simon are considered to be the 'fathers' (Nilsson, 2009). At its beginning, AI investigated the propositional logic and the representation of knowledge (i.e., expert system). The developed methods and tools were related to reasoning on knowledge bases, represented by facts and rules. Operational research algorithms and implementations for logic clauses have been proposed (Robinson, 1965). Over decades, the field of AI was extended to become a multidisciplinary science containing reinforcement learning, adaptive control theory, information theory, theory of computation, game theory, etc. (Russell and Norvig, 2009; Hutter, 2005, 2018; Boedecker et al., 2017). The success stories of Deep Blue (IBM) and AlphaGo (DeepMind, Google) open a new age of AI (Fabbri, 2018). Deep Blue programme defeated the world chess champion Garry Kasparov on 11 May 1997 and AlphaGo programme defeated the world Go champion Lee Sedol on 9–15 March 2016. Big Blue machine was a supercomputer having 30 CPU and 480 chess CPU, where AlphaGo supercomputer has 1202 CPU and 176 GPU (Fabbri, 2018). Notice that intelligent machine based on supercomputer is designed to solve specific problem and it can outperform human capability such as playing chess or Go. However, human being is more agile, adaptive and creative and uses the five senses to perceive the environment before making decisions. This is not the case with today's intelligent machine.

Figure 1.16 Basic smart object platform.

1.1.3.3. State of the art of smart objects platform

Nowadays, it is possible to integrate ‘microcontroller unit’, environment sensors (i.e., NEMS/MEMS sensors), low energy and long-range wireless communication medium (e.g., NB-IoT, LoRa) into a single very-large-scale integration chip to implement a SO for smart farming application (Conti et al., 2018). However, the size of a SO hardware platform based on the ‘Commercial Off-The-Shelf’ (COTS) components is smaller than a credit card (Beigl and Gellersen, 2018).

The basic hardware architecture of an SO is illustrated in Fig. 1.17.

It is important to highlight that a low-cost SO can implement only simple cognitive algorithms (decision-making) due to resource constraints. The question is: can the SOs meet the requirements of smart farming? According to the state of the art, it is now certain that SOs cannot replace and fully assume the tasks of farmers. However, the SOs can help the farmer to monitor and care for farm animals and grow crops such as through automatic irrigation of a cultivated field. Different classifications of SO are investigated in Hutter, 2005; Hutter, 2018; Kortuem et al., 2010; Yachir et al. (2016).