IoT-Based hierarchical control system
In the Province of Almería (South-East of Spain) is situated the largest concentration of greenhouses in the world (more than 32000 hectares of surface) . This agriculture competes in the European market because of lower costs and ability to offer out-of-season products . To maintain the competitiveness of the sector against competition from countries with lower labor costs, a large effort is carried out nowadays directed to the introduction of technology in each one of the phases of the crop growth, connecting systems so as to allow an integrated, multidimensional view of farming activities, data sharing and enabling deeper understanding on how the whole ecosystem works. This use of ICT involves a large amount of data, physical and virtual sensors, control loops, communication networks, models and optimisation techniques, forecast, vegetable production, energy costs, etc…
In this situation, one of the initiatives is the development of a decision support based on IoT technology for greenhouse tomato production control involving a large amount of data, models and algorithms, focused on important aspects such as water and energy use efficiency for tomato cultivation in greenhouses . In this IoT solution (Figure 1), the “fog”-based hierarchical control system is used to accommodate greenhouse crop-growth system complexity and to provide a multiobjective optimization solution to satisfy the many conflicting objectives inherent in control system design . It rises as a solution to improve the system efficiency, obtaining economic benefits for the farmer.
Therefore, in farming systems, the introduction of new technologies is an essential part of this process and improving productivity and quality are key to maintain profitability. Currently, the greenhouse growth control is realised with local supervisory systems. The introduction of IoT technology is generating a new paradigm in greenhouse control. For this task, a FIWARE-based IoT platform (Figure 1) was designed to integrate information from different sources, helping to optimize crop growth (production), costs reduction and assisting in decision-making at the right time. Hence, the grower has to choose what the optimal conditions for plant growth are, and to convert them into control specifications. In general, to control the greenhouse environment and fertigation, a grower has to manage temperature, air humidity, CO2 concentration, radiation, and water and nutrients supply in a system with strong physical relationships (e.g. radiation and the water consumption), which influence the greenhouses control decisions .
In this platform (Figure 1), the first part of the system is centered in the integration of the different data sources. The system receives the data necessary for decision-making (cloud computing) from the local systems (climate and fertigation sensors and actuators), field-notebooks (labours), lab analyses and from other agents (mainly public agencies) like weather forecast (National Agency of Meteorology, Spanish Ministry of Agriculture, Food and Environment), market price fluctuations (Agricultural Prices Observatory-Andalusian Government) and energy prices (Spanish energy market). Prediction models are used in two layers: (i) Lower layer, the weather forecast is used to predict the greenhouse climate and fertigation conditions through the use of climate and water balance models . These models allow the user to know about the crop surrounding conditions, including actuators and the setpoint tracking and disturbance rejection controllers. With this architecture, supervision and control of each greenhouse facility is carried out by the local supervisor and the data storage is established using ‘fog computing’ techniques and (ii) on the upper layer, growth models  and diseases early warning prediction systems  are established for different crops, such as tomato or sweet peppers. Such models can be used to predict crop growth and the emergence of cryptogamic diseases. Both models are based on greenhouse climatic variables and data from field-notebook information to provide support tools for decision-making systems in production management. The models provide results as a function of variations in climatic conditions in the greenhouse environment. The availability of this information will help optimizing crop growth (production), diseases control, reducing costs and assisting in decision-making at the right time.
In this framework, multiobjective optimization is used to calculate the climate and fertigation setpoints along the control horizon considered. This hierarchical IoT-based system, which combines public information and different farmer data, is used as a multi-objective optimization solution to meet the many contradictory objectives inherent in designing the control system and exchanging data (interoperability) available to different users: farmers, technical engineers, production planners and public administrations. In this approach, control decisions are modified by market price fluctuations and environmental regulations to improve a large number of possible objectives. Therefore, optimizing the production process in a greenhouse agrosystem may be summarized as tackling the following objectives: optimal crop growth (greater production along with better quality), reduction in associated costs (mainly fuel, electricity, and fertilizers), reduction in residue use (mainly pesticides and ions in the soil), avoid the diseases presence and development, and improvement in water-use efficiency . For an objective function, using climate, fertigation, early warning and growing models, along with optimization techniques, the optimal trajectories are determined to during the crop cycle .
Scheme of the IoT platform for decision support in tomato crops in greenhouses
Benefits and improvements to the different actors and activities along the supply chain will be delivered in terms of: improved use of resources, better access to data and exchange, synchronization, storage reduction and cost. Technology and data exchange can be essential tools in the search for solutions through the introduction of technology at each stage of the supply chain, creating relationships between the different stages based on transparency and information about products and processes. In addition, through the use of publicly available data, the value of such data can be harnessed. An IoT-based DSS was designed for greenhouse crops, using both farmer information, publicly available information, pseudophysical models and multiobjective optimization.
 F. Rodríguez, M. Berenguel, J. L. Guzmán, A. Ramírez-Arias, Modeling and control of greenhouse crop growth, Springer, 2015.
 J. Sánchez-Molina, M. Li, F. Rodríguez, J. Guzmán, H. Wang, X. Yang, Development and test verification of air temperature model for Chinese solar and Spanish Almeria-type greenhouses, International Journal of Agricultural and Biological Engineering 10 (4) (2017) 66–76.
 J.A. Sánchez-Molina, N. Pérez, F. Rodríguez, J.L. Guzmán, J.C. López, Support system for decision making in the management of the greenhouse environment based on growth model for sweet pepper, Agricultural Systems 139 (2015) 144-152.
 A. Ramírez-Arias, F. Rodríguez, J.L. Guzmán, M. Berenguel, Multiobjective hierarchical control architecture for greenhouse crop growth, Automatica 48 (3) (2012) 490-498.
 Ming Li, Sining Chen, Fang Liu, Li Zhao, Qingyu Xue, Hui Wang, Meixiang Chen, Peng Lei, Dong, Mei Wen, J. A. Sánchez-Molina, J. F. Bienvenido, Zhenfa Li, Xinting Yang, A risk management system for meteorological disasters of solar greenhouse vegetables, Precision Agriculture 18 (6) (2017) 997–1010.
Jorge Antonio Sánchez-Molina and Manuel Muñoz Rodríguez
Automatic Control, Robotics and Mechatronics Research Group, University of Almería, CeiA3, CIESOL, Spain