Model predictive control and its application in agriculture: A review

https://doi.org/10.1016/j.compag.2018.06.004Get rights and content

Highlights

  • Survey on the development of MPC during last forty years.

  • The detailed review of MPC in agriculture applications in recent years.

  • Challenges and future perspectives of MPC.

  • The potential future applications of MPC in agriculture.

Abstract

Agriculture plays a decisive role in the survival of humankind. The efficient and precise regulation of agriculture will ensure the welfare of people throughout the world. However, difficult and urgent agricultural challenges persist, and traditional agricultural production regulation methods are time consuming and laborious. Moreover, the application of intelligent algorithms to modern agricultural production requires the support of a database, which can be complex and difficult to use in practice and requires a large amount of computing. Fortunately, model predictive control (MPC) methods can achieve highly accurate control operations with moderate complexity and can also allow for rolling optimization in a limited time domain, which improves precision. Despite being a process control method that originated in industry, MPC is highly suited for application in agriculture because it can effectively address nonlinear and large time-delay systems. The aim of this review is to introduce MPC and describe its current application in agriculture. In this review, the development of MPC and various improved MPC schemes are described. The development of MPC is divided into three stages. In addition, applications and new technologies associated with MPC in irrigation systems, agricultural machinery, agricultural production, product processing and greenhouses are analyzed. Smart agriculture will have a promising future because of the implementation of MPC technologies derived from successful industrial applications. Finally, challenges and future perspectives of MPC technology use in agriculture are summarized and forecasted.

Introduction

Agriculture is the foundation of human existence, it is vital to the survival of humankind. Today, the total global population exceeds 7 billion, and by 2050, the urban population will increase by an additional 2.5 billion through population growth and urbanization, with nearly 90% of the population concentrated in Asia and Africa (Lloyd et al., 2017). However, the amount of available food is limited, especially in Africa, and the food scarcity problem has yet to be solved (Sanchez, 2002); additionally, Asia has a serious shortage of water (Pomeranz, 2009). The earth's water supply is rich, with 1.45 billion cubic kilometers in total, and 72% of the Earth’s surface area is covered with water. However, less than 1% of the world’s fresh water is easy to exploit for direct human use, accounting for approximately 0.007% of the total water on the planet. The total land area of the world exceeds 13 billion hectares; however, the area of potentially arable land throughout the world accounts for 22% of the total land area, at just 3031 million hectares (Lal, 1990). Moreover, traditional agriculture is time-consuming and labor-intensive, with low production efficiency. Considering the increasing population, water shortages, limited land resources, and low production efficiency, it is urgent to efficiently regulate agriculture.

Agricultural systems are complex, multivariate and unpredictable (Kamilaris et al., 2018). Classical control technologies such as those involving on/off, P, proportional integral (PI), and proportion-integration-differentiation (PID) control (Christofides et al., 2013, Afram and Janabi-Sharifi, 2014) are easy to implement but are unable to control moving processes with time delays; in addition, adjusting the controller is cumbersome and time consuming (Wang et al., 2001). Intelligent methods such as fuzzy logic (FL) control and artificial neural network (ANN) control involve not only deterministic mathematical models but also nonmathematical generalized models and mixed models (Afram and Janabi-Sharifi, 2014). However, these methods require learning and reasoning based on data-driven or embedded expert knowledge. Fortunately, the performance of MPC is superior to that of classical control and is easier to implement than intelligent computing algorithms. MPC can achieve high regulation accuracy with moderate complexity. Therefore, this method is highly suited for precision agricultural production.

MPC refers to a class of advanced computer-controlled algorithms that use an explicit process model to predict a plant's future response (Qin and Badgwell, 2003). A series of control inputs are computed at each sampling instant, but only the first computed input is implemented in the process (Bumroongsri and Kheawhom, 2014). The first input in the optimal sequence is then sent to the factory, and the entire calculation is repeated at subsequent control intervals (Mogal and Warke, 2013). The algorithm consists of three parts: prediction model, rolling optimization, and feedback adjustment (Zhang, et al., 2017a, Zhang et al., 2017b, Zhang et al., 2017c). MPC was developed in the early 1960s and has been used widely in process industries (Garriga and Soroush, 2010). The method allows for the introduction of constraints, predictive information, and nonlinear dynamics. Linear MPC (LMPC) is used to solve convex quadratic programming problems (QPs) online. Nonlinear MPC (NMPC) allows for the control of systems with nonlinear dynamics and involves significantly more calculation than does LMPC (Vukov et al., 2015). Early MPC theory incorporated dynamic matrix control (DMC) and model algorithm control (MAC), which are based on the linear quadratic Gaussian (LQG), which is a relatively simple parameter model. Internal model control (IMC) is defined for single-input/single-output, discrete-time systems (Garcia and Morari, 1982). DMC and MAC inspired the development of IMC but lacked stability. Thus, the generalized predictive control (GPC) model was developed; its theoretical basis is more complete than DMC and MAC, and it could also solve the robustness problem to a certain extent. Most industrial processes are highly complex, involve large amounts of interference and are strongly nonlinear; thus, adaptive MPC, robust MPC and NMPC have been developed. However, the calculation time for these MPCs is quite long and inefficient. To solve these problems, the distributed MPC, hybrid MPC, explicit MPC and other new MPC models were established (Lee, 2011). In addition, because actual production processes are random, stochastic MPCs have also been developed (Zhang, et al., 2017a, Zhang et al., 2017b, Zhang et al., 2017c).

As one of the most promising control strategies, MPC has been widely studied (Froisy, 1994, Morari and Lee, 1999, Qin and Badgwell, 2003, Rawlings and Mayne, 2009) and applied in industry (Qin, 1997, Seki et al., 2001, Han and Qiao, 2014). The application of MPC to agriculture can yield significant productivity and efficiency benefits. However, no review of MPC use in agricultural applications has been reported. MPC has been applied to agriculture, but it has not yet been applied in all respects. In this review, MPC development and current application in agriculture are described. The development of MPC can be divided into three stages: classical MPC, improved MPC, and the latest MPC. These MPCs are described in Chapter 2. Currently, MPC is used mainly in agricultural applications such as irrigation systems, machinery, production, product processing, and greenhouses. These applications are described in Chapter 3. The challenges and future perspectives of MPC are discussed in Chapter 4. Finally, conclusions are discussed.

Section snippets

The development of MPC

In the 1960s and 1970s, the concept of MPC appeared in the literature. However, MPC was not introduced into process industries until the 1980s (Froisy, 1994). In general, the evolution of this scheme can be divided into three stages according to the degree of technological development. The theoretical principle of MPC is shown in Fig. 1. The input is r(k), the initial output is ydk+j, the output after optimization is ymk+j, and the output after on-line correction is ypk+j; after a number of

Application of MPC in agriculture

Agricultural production objects and industrial control objects have many similarities, such as a large time delay, nonlinearity, and constraints. This chapter summarizes the current application of MPC in agriculture, such as in irrigation systems, agricultural machinery, agricultural production, product processing, and greenhouses.

Challenges and future perspectives

MPC cannot describe unstable systems and originally had difficulty recognizing online models, but it now reduces calculation time while maintaining high performance. The biggest challenge in the future may be the trade-off between computing time, performance level, and economic cost. It is very difficult to reduce calculation time and cost without affecting performance. In addition, for some complex algorithms, it is very difficult to guarantee stability and feasibility. Thus, some of those

Conclusions

This review has briefly discussed the current development of MPC and its application in agriculture. Despite the many reviews of MPC and its development in industry, no summary of MPC use in agricultural applications has yet been published. In this article, we have roughly divided the development of MPC into three stages according to the different problems solved. Classical MPC mainly solves multivariable constraint control problems. However, this method is not suitable for nonlinear systems.

Acknowledgments

This work was supported by the EU FP7 Framework Program “Innovative model and demonstration based on water management for resource efficiency in integrated multitrophic agriculture and aquaculture systems” [FP7-ENV-2013-WATER-INNO-DEMO, 619137]. The authors would like to thank American Journal Experts for providing an English-language edit of this article.

References (156)

  • T. Coen

    Cruise control using model predictive control with constraints

    Comput. Electron. Agric.

    (2008)
  • D. Delgoda

    Irrigation control based on model predictive control (MPC): formulation of theory and validation using weather forecast data and AQUACROP model

    Environ. Modell. Software

    (2016)
  • M.Y. El Ghoumari

    Non-linear constrained MPC: real-time implementation of greenhouse air temperature control

    Comput. Electron. Agric.

    (2005)
  • F. Fele

    Coalitional model predictive control of an irrigation canal

    J. Process Control

    (2014)
  • J. Figueiredo

    SCADA system with predictive controller applied to irrigation canals

    Control Eng. Pract.

    (2013)
  • H. Fukushima

    Adaptive model predictive control for a class of constrained linear systems based on the comparison model

    Automatica

    (2007)
  • M.S. Ghasemi et al.

    Robust tube-based MPC of constrained piecewise affine systems with bounded additive disturbances

    Nonlinear Anal. Hybrid Syst

    (2017)
  • G. Goebel et al.

    Semi-explicit MPC based on subspace clustering

    Automatic

    (2017)
  • J.K. Gruber

    Nonlinear MPC based on a Volterra series model for greenhouse temperature control using natural ventilation

    Control Eng. Pract.

    (2011)
  • K. Horváth

    New offset-free method for model predictive control of open channels

    Control Eng. Pract.

    (2015)
  • B. Houska

    An auto-generated real-time iteration algorithm for nonlinear MPC in the microsecond range

    Automatica

    (2011)
  • J. Kalmari

    Nonlinear model predictive control of hydraulic forestry crane with automatic sway damping

    Comput. Electron. Agric.

    (2014)
  • J. Kalmari

    Coordinated motion of a hydraulic forestry crane and a vehicle using nonlinear model predictive control

    Comput. Electron. Agric.

    (2017)
  • A. Kamilaris

    Deep learning in agriculture: a survey

    Comput. Electron. Agric.

    (2018)
  • E. Kayacan

    Distributed nonlinear model predictive control of an autonomous tractor–trailer system

    Mechatronics

    (2014)
  • E. Kayacan

    Towards agrobots: Identification of the yaw dynamics and trajectory tracking of an autonomous tractor

    Comput. Electron. Agric.

    (2015)
  • B. Kersbergen

    Distributed model predictive control for railway traffic management

    Transp. Res. Part C Emerging Technol.

    (2016)
  • B. Kouvaritakis

    Brief paper: explicit use of probabilistic distributions in linear predictive control

    Automatica

    (2010)
  • T. Kraus

    Moving horizon estimation and nonlinear model predictive control for autonomous agricultural vehicles

    Comput. Electron. Agric.

    (2013)
  • J.H. Lee

    From robust model predictive control to stochastic optimal control and approximate dynamic programming: a perspective gained from a personal journey

    Comput. Chem. Eng.

    (2014)
  • D. Li

    FPGA based QDMC control for reverse-osmosis water desalination system

    Desalination

    (2012)
  • D. Limon

    Robust tube-based MPC for tracking of constrained linear systems with additive disturbances

    J. Process Control

    (2010)
  • X. Liu

    Process control based on principal component analysis for maize drying

    Food Control

    (2006)
  • A.C. McCarthy

    Simulation of irrigation control strategies for cotton using Model Predictive Control within the VARIwise simulation framework

    Comput. Electron. Agric.

    (2014)
  • P.R.C. Mendes

    A practical approach for hybrid distributed MPC

    J. Process Control

    (2017)
  • M. Morari et al.

    Model predictive control: past, present and future

    Comput. Chem. Eng.

    (1999)
  • A. Pawlowski

    Improving feedforward disturbance compensation capabilities in Generalized Predictive Control

    J. Process Control

    (2012)
  • V. Adetola et al.

    Robust adaptive MPC for constrained uncertain nonlinear systems

    Int. J. Adapt Control Signal Process.

    (2011)
  • J.V. Aguilar

    Predictive control of irrigation canals – robust design and real-time implementation

    Water Resour. Manage.

    (2016)
  • H. Akçakaya et al.

    An application of robust model predictive control with integral action

    Instrum Sci. Technol.

    (2009)
  • A. Alessio et al.

    A survey on explicit model predictive control

    Nonlinear Model Predictive Control

    (2009)
  • Alrifaee, B. et al., 2016. Coordinated Non-Cooperative Distributed Model Predictive Control for Decoupled Systems Using...
  • A. Álvarez

    Constrained predictive control of an irrigation canal

    J. Irrig. Drain. Eng.

    (2013)
  • Bayer, F.A. et al., 2016. A tube-based approach to nonlinear explicit MPC. In: 55th IEEE Conference on Decision and...
  • M. Bounkhel et al.

    Optimal harvesting effort for nonlinear predictive control model for a single species fishery

    Math. Problems Eng.

    (2015)
  • M. Breckpot

    Flood control with model predictive control for river systems with water reservoirs

    J. Irrig. Drain. Eng.

    (2013)
  • P. Bumroongsri et al.

    Off-line robust constrained mpc for linear time-varying systems with persistent disturbances

    Math. Problems Eng.

    (2014)
  • E.F. Camacho et al.

    Robust adaptive model predictive control of a solar plant with bounded uncertainties

    Int. J. Adapt Control Signal Process.

    (2015)
  • M. Cannon

    Stochastic tubes in model predictive control with probabilistic constraints

    IEEE Trans. Autom. Control

    (2011)
  • Cariou, C. et al., 2010. Autonomous Maneuvers of a Farm Vehicle with a Trailed Implement in Headland. In: 7th...
  • Cited by (160)

    View all citing articles on Scopus
    View full text