Computers and Electronics in Agriculture: Zhaoyu Zhai, José Fernán Martínez, Victoria Beltran, Néstor Lucas Martínez T

Computers and Electronics in Agriculture 170 (2020) 105256
Contents lists available at ScienceDirect
Computers and Electronics in Agriculture

journal homepage: www.elsevier.com/locate/compag
Decision support systems for agriculture 4.0: Survey and challenges T

Zhaoyu Zhai , José Fernán Martínez, Victoria Beltran, Néstor Lucas Martínez
⁎
Departamento de Ingeniería Telemática y Electrónica (DTE), Escuela Técnica Superior de Ingeniería y Sistemas de Telecomunicación (ETSIST), Universidad Politécnica de
Madrid (UPM), C/Nikola Tesla, s/n, 28031 Madrid, Spain
ARTICLE INFO ABSTRACT
Keywords: Undoubtedly, high demands for food from the world-wide growing population are impacting the environment
Agriculture and putting many pressures on agricultural productivity. Agriculture 4.0, as the fourth evolution in the farming
Smart farming technology, puts forward four essential requirements: increasing productivity, allocating resources reasonably,
Decision-making adapting to climate change, and avoiding food waste. As advanced information systems and Internet technol-
Decision support systems
ogies are adopted in Agriculture 4.0, enormous farming data, such as meteorological information, soil condi-
tions, marketing demands, and land uses, can be collected, analyzed, and processed for assisting farmers in
making appropriate decisions and obtaining higher profits. Therefore, agricultural decision support systems for
Agriculture 4.0 has become a very attractive topic for the research community. The objective of this paper aims
at exploring the upcoming challenges of employing agricultural decision support systems in Agriculture 4.0.
Future researchers may improve the decision support systems by overcoming these detected challenges. In this
paper, the systematic literature review technique is used to survey thirteen representative decision support
systems, including their applications for agricultural mission planning, water resources management, climate
change adaptation, and food waste control. Each decision support system is analyzed under a systematic manner.
A comprehensive evaluation is conducted from the aspects of interoperability, scalability, accessibility, usability,
etc. Based on the evaluation result, upcoming challenges are detected and summarized, suggesting the devel-
opment trends and demonstrating potential improvements for future research.
1. Introduction waste of natural resources. In the 20th century, Agriculture 3.0

emerged from the rapid development of computing and electronics.
Human beings have cultivated lands and breed animals to obtain Computer programs and robotic techniques allowed agricultural ma-
food for their survival since ancient times. This practice, known as chineries to perform operations efficiently and intelligently. Before the
agriculture, has evolved following a long-term and progressive process problems left in Agriculture 2.0 went too far, strategies were adjusted in
(Tekinerdogan, 2018), going from Agriculture 1.0 to 4.0, as shown in Agriculture 3.0. The reasonable work distribution to agricultural ma-
Fig. 1. chineries reduced the use of chemicals, improved the precision of irri-
In Fig. 1, Agriculture 1.0 refers to the traditional agricultural era, gation and so on. Nowadays, the evolution of agriculture steps into
mainly replying on the manpower and animal forces. In this stage, Agriculture 4.0, thanks to the employment of current technologies like
though simple tools like sickles and shovels were used in agricultural Internet of Things, Big Data, Artificial Intelligence, Cloud Computing,
activities, humans still cannot get rid of heavy manual labor, so pro- Remote Sensing, etc. The applications of these technologies can im-
ductivity remained at a low level. Until the 19th century, steam engines prove the efficiency of agricultural activities significantly. For instance,
were improved and widely used to provide new powers in all walks of Ferrandez-Pastor et al. (2016) took advantages of Internet of Things
life and industries, including agriculture. It came to the era of Agri- and developed a low-cost sensor and actuator network platform. This
culture 2.0 when various agricultural machineries were operated by platform aims at optimizing the production efficiency, increasing
farmers manually and plenty of chemicals were used. Obviously, quality, minimizing environmental impacts, and reducing the use of
Agriculture 2.0 significantly increased the efficiency and productivity resources like energy and water. Wolfert et al. (2017) conducted a
of farm works. Nevertheless, this substantial improvement brought too survey on applying Big Data to smart farming. They have pointed out
harmful consequences: field chemical contaminations, destruction of that Big Data is now used to provide farmers with predictive insights in
the ecological environment, excessive consumption of powers, and farming operations and real-time operational decisions. Liakos et al.
⁎
Corresponding author.
E-mail address: zhaoyu.zhai@upm.es (Z. Zhai).
https://doi.org/10.1016/j.compag.2020.105256
Received 24 August 2019; Received in revised form 15 January 2020; Accepted 29 January 2020
Available online 08 February 2020
0168-1699/ © 2020 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/BY-NC-ND/4.0/).
Z. Zhai, et al. Computers and Electronics in Agriculture 170 (2020) 105256
Fig. 1. A general framework of an ADSS for pest management.
(2018) explored the current state of machine learning techniques in decisions.

agriculture. They have drawn a conclusion that real-time artificial in- An ADSS is not only able to provide a list of options for on-going
telligence enables computer programs to generate rich recommenda- activities, but also may help decision makers to achieve better perfor-
tions and insights for supporting farmers to make proper decisions. mances in future tasks (Alenljung, 2008). Some successful examples
Lopez-Riquelme et al. (2017) developed a precision agriculture appli- have illustrated how Agriculture 4.0 can benefit from ADSSs. For in-
cation on the basis of FIWARE cloud. This application is able to reduce stance, the Watson Decision Platform for Agriculture was released by
the amount of water for irrigation tasks. Thus, their work demonstrates IBM Watson and The Weather Company, combining agriculture with
that using FIWARE cloud services in the agronomic context is highly IBM’s advanced capabilities in Artificial Intelligence, Internet of Things,
beneficial. Bonfante et al. (2019) proposed LCIS DSS, an irrigation and Cloud Computing (Watson Decision Platform for Agriculture,
support system for improving the efficiency of water use in precision https://www.ibm.com/downloads/cas/ONVXEB2A). On the one hand,
agriculture based on three different methodologies: IRRISAT® (remote this platform provides a suite of solutions that spans the farm-to-fork
sensing), W-Mod (simulation modelling), and W-Tens (situ soil sensor). ecosystem and it is able to analyze any factors which have potential
Through their case study in maize, they determined that the first two effects on crops. Farmers can obtain crop pictures by deploying Un-
approaches might represent the best solution in regards to irrigation manned Aerial Vehicles (UAVs). Then, these pictures are uploaded to
water use efficiency. In the stage of Agriculture 4.0, it is worth men- IBM Cloud for further analyses based on computer vision algorithms.
tioning that data from all fields are gathered and processed, providing a The analytic results keep farmers updated with health conditions of
clear view for farmers. crops. Thus, the working efficiency and accuracy of detecting crop
Stakeholders and farmers may encounter difficulties in making diseases are greatly improved. On the other hand, owners of large-scale
proper decisions about agricultural management with the explosive farms can use Watson Decision Platform to estimate the price trending
amount of information (e.g. environmental, crop-related, and economic in trading markets. Under this circumstance, the time for irrigation,
data) (Taechatanasat and Armstrong, 2014). Because it is challenging pollination, phenology, fertilization, harvesting, and selling can be
for them to transfer these data into practical knowledge. Thus, plat- precisely controlled in order to achieve the maximum profits. It is worth
forms like decision support systems (DSSs) are needed in order to assist noting that the inputs to Watson Decision Platform concerns various
them in making evidence-based and precise decisions. sources, such as weather data (provided by the Weather Company), soil
Regarding the definition of a DSS, researchers have described this data (moisture at multiple depths, nutrient content, fertility, and type),
term from various viewpoints. In 1980, Jones (1980) described this equipment data (gathered from sensors in devices), workflow data
term “decision support system” as “a computer-based support system (planting and harvesting dates, fertilizer and pesticide application rates,
for decision makers who deal with semi-structured problems to improve and harvest outputs), and high definition visual imagery (collected by
the quality of decisions”. Sheng and Zhang (2009) defined it as “a satellites, drones, and fixed-wing aircraft). IBM is not the only company
human-computer system which is able to collect, process, and provide who contributes to Agriculture 4.0, another company named Prospera
information based on computers”. Yazdani et al. (2017) considered it as (Digital Farming System, http://prospera.ag/) takes advantages of
“a specific class of computerized information system, enabling to Computer Vision, Artificial Intelligence, and Cloud Computing for de-
manage decision-making activities”. Terribile et al. (2015) explained it veloping a digital farming system that helps farmers to analyze data
as a smart system that provides operational answers and supports de- collected from their fields. This system is capable of suggesting the best
cision-making to specific demands and problems based on collected time for irrigation, fertilization, pollination, and harvesting by mon-
data. Thus, considering the above definitions, an agricultural decision itoring the growth rates of crops. Farmers can also be notified when
support system (ADSS) can be defined as a human-computer system crops are infected by any diseases. According to the statistics from
which utilizes data from various sources, aiming at providing farmers Prospera, the yield production is estimated with 95% accuracy and
with a list of advice for supporting their decision-making under dif- productivity is increased as much as 30%. The limitation of this digital
ferent circumstances. One of the most representative characteristics of farming system is that it only concerns the scenarios of greenhouses and
an ADSS is that it does not give direct instructions or commands to large-scale row crops. As a consequence, it is more interesting that
farmers. Because farmers are in the position of taking the final Prospera can enrich the functionality of the system for providing
2
farmers with adequate suites of solutions. Moreover, Bazzani (2005) be better applied to the domain of agriculture, the requirements of
developed DSIRR, a decision support system for irrigation. DSIRR is Agriculture 4.0 have to be analyzed beforehand.
more than a normative platform to generate the optimal irrigation plan, In 2013, the German government firstly proposed Industry 4.0, known
but an ADSS for exploring the trade-off among conflict objectives and as the fourth industrial evolution (Anderl, 2015). Two years later, Agri-
offering farmers compromising solutions. It considers four categories of culture 4.0 was defined and quickly attracted wide attentions from
data sources, including economic (farm income, profit, and gross do- worldwide researchers (Agriculture 4.0: The future of farming tech-
mestic product), social (public support subsidy and farm employment), nology, https://www.worldgovernmentsummit.org/api/publications).
water (seasonality, consumption, marginal value, and irrigation tech- Four main requirements are put forward and listed as follows.
nology), and environmental indicators (soil cover, nitrogen, pesticide,
and energy). Based on these indicators, a linear model is used to assess • 1. Increasing productivity: The population growth and shortage of
the trade-off among economic-social-environmental objectives. With a food will consequently boost the demand for agricultural produc-
user-friendly graphical interface, farmers can directly control and tions. Meanwhile, people’s diet has been changing as well, mainly
monitor irrigation processes. Though the water usage is significantly reflected in demanding for high-value animal protein. Furthermore,
reduced and farmers can irrigate farming fields more efficiently, DSIRR with the development of urbanization, infrastructures and buildings
requires further developments such as employing modular approaches, would take place of farmlands (Yuan et al., 2018).
permitting to integrate new modules focusing on specific aspects of • 2. Allocating resources reasonably: Natural resources are incredibly
interest. Judging from above successful examples, it is concluded that stressed nowadays. Firstly, unused lands for cultivation are rare and
ADSSs are accelerating the development paces of Agriculture 4.0 from 25% of farmlands are marked as highly degraded due to deforesta-
various perspectives. tion, overcutting vegetation, inadequate fallow periods, etc. (Udias
Though ADSSs are quite helpful in farm management, the un- et al., 2018). Secondly, water resources are overused in an un-
welcome fact is that the use of ADSSs has been limited due to some reasonable way (Dong et al., 2018). Frequent water transfers from
critical issues in Fig. 2 (Tyrychtr and Vostrovsky, 2017). rivers and lakes are causing serious environmental problems.
In Fig. 2, following issues have been pointed out. Thirdly, agricultural machineries are not efficiently deployed due to
improper work distributions. A large amount of energy resources is
• Farmers seldom have experiences or knowledge of using ADSSs. The consequently wasted. (Fountas et al., 2015).
typical graphical interface of ADSSs is sometimes not user-friendly • 3. Adapting to climate change: Climate change has been greatly af-
and it may be confusing for farmers to perform desired operations. fecting the environment. One of the main factors which leads to
• ADSS developers may ignore the requirement analyses from the end climate change is manmade emissions of Greenhouse Gases (GHGs).
users, leading to the fact that inputs and outputs of ADSSs may not The side effects of climate change result in frequent occurrences of
fit farmers’ needs and decision-making styles. droughts, floods, and extreme weather conditions (Czimber and
• The functionalities of current ADSSs are limited and task-specific. Galos, 2016). Additionally, agricultural productions are especially
An ADSS may only focus on a single perspective. As a consequence, vulnerable and sensitive to the impacts of climate change (Kmoch
farmers have to use several ADSSs to manage agricultural activities. et al., 2018). Lack of efforts in adapting to climate change will cause
• When generating the advice, current ADSSs may miss some funda- an increase in uncertainty about food quality, accessibility, and
mental factors, such as climate change, soil spatial variability, crop utilization.
disease, etc. The lack of these considerations may result in imprecise • 4. Avoiding food waste: Food waste comes from each stage of the
outputs from ADSSs. agricultural life cycle, including producing, delivering, marketing,
etc. Firstly, due to the overuse of chemicals, lack of pest manage-
However, the above detected issues are not complete enough. To the ment, and ignorance of climate change adaptations, agricultural
best of our knowledge, most of current surveys mainly focus on com- products may become contaminated and unqualified (van Evert
paring framework differences of ADSSs and analyzing their perfor- et al., 2017), leading to food waste and damage to farmlands. Sec-
mances on specific agricultural tasks or they just explored the current ondly, the world shares a globalized supply and marketing system
state of ADSSs within a small range (e.g. a country) (Tyrychtr and (Borodin et al., 2016). However, the food delivery is a time-sensitive
Vostrovsky, 2017; Hayman, 2004; Nguyen et al., 2007). The critical process. Inappropriate decision-making of deliveries may cause food
issues and upcoming challenges of employing ADSSs in Agriculture 4.0 waste. Thirdly, wasted food is harmful to the environment. Re-
have not been fully investigated. For understanding how ADSSs could cycling and processing wasted food will consume more resources
than producing new ones (Pourmoayed et al., 2016).
Based on the above four requirements, thirteen ADSSs are selected

from current literatures. The objective of this paper is to review these
ADSSs for Agriculture 4.0 and detect upcoming challenges by means of
a systematic literature review technique, consisting of the following five
steps.
• 1. Defining a question: What does Agriculture 4.0 require from ADSSs

employments? (presented in Section 1)
• 2. Search for literature: Thirteen representative ADSSs are chosen
from current literatures and projects according to the requirement of
Agriculture 4.0. (presented in Section 2)
• 3. Extracting information from selected works: Each ADSS is in-
troduced and analyzed respectively under a systematic manner.
(presented in Section 2)
• 4. Assessing the quality of selected works: An evaluation between se-
lected thirteen ADSSs is conducted from the aspects of interoper-
ability, scalability, accessibility, usability, etc. (presented in Section
Fig. 2. Some critical issues of employing ADSSs. 3)
3
Fig. 3. A general framework of agricultural decision support systems.
• 5. Drawing a conclusion: Seven upcoming challenges are detected complexity of agricultural problems. When planning an agricultural
according to evaluation results. (presented in Section 4) mission, several factors has to been taken into account: number of in-
volved agricultural machineries, capability of these machineries,
2. Literature review on selected ADSSs number of tasks, etc. The second issue addresses the time window of
agricultural activities. Normally, an agricultural year lasts from 8 to
The selected ADSSs have covered the agricultural applications in: (i) 12 months. Certain activities should be performed at a specific time,
mission planning; (ii) water resources management; (iii) climate change such as when to seed, fertilize, and harvest. A delay of one or more days
adaptation; and (iv) food waste control. Each ADSS is explored from the for mission executions may lead to unexpected economic losses.
systematic view by following the general framework (Fig. 3). The decision-making process in AgriSupport II system is computed
In Fig. 3, agricultural data should be collected in the first place and by a farm planning model algorithm. This algorithm is designed to
treated as inputs to the decision-making tools (modules). Advice about identify which work units are suitable for performing what tasks. For
managing agricultural activities is generated according to the compu- providing this plan, following attributes are considered as inputs to the
tational results. Farmers can then choose the most appropriate option planning model.
and adopt it to solve the problems. It is worth mentioning that con-
straints should be taken into account for guaranteeing the quality of • Mode: It is defined as a possible plan for performing agricultural
provided advice. tasks.
• Technical path: It is defined as a sequence of operations to be per-
2.1. ADSSs for mission planning formed. A mode is composed of several technical paths and their
relationships of precedence.
Current researches on ADSSs for mission planning mainly focus on • Resource: A technical path requires certain resources like machi-
two aspects: task allocation and path planning. On the one hand, neries and human labors. For using resources, the technical path has
agricultural tasks should be allocated to the most appropriate machi- to pay costs, which will be included in the estimated cost of per-
neries for execution, and on the other hand, proper path planning can forming operations.
quickly and precisely guide agricultural machineries to the nearest • Precedence: It is defined as the priority of each operation in the
destinations and then execute tasks. Generally, effective planning ap- technical paths. Those operations with a higher precedence will be
proaches can greatly increase the productivity because agricultural executed in the first place.
tasks are completed within the minimum time. • Time window: It is defined as the starting and completion time of
Four ADSSs for mission planning are reviewed in this manuscript. modes.
The first two (AgriSupport II system and Multi-robot sense-act system)
are related to task allocation and the latter two (ADSS for route plan- After obtaining above inputs, the farm planning model algorithm
ning in soil-sensitive fields and On-board decision-making approach) computes the cost of all feasible modes and compares them with each
are in regards to path planning. other in order to find out the one with the lowest cost, as the optimal
plan for distributing agricultural tasks to work units. This algorithm
2.1.1. AgriSupport II system adopts the CPLEX optimizer as the decision-making tool.
The AgriSupport II system aims at adopting the latest advances in The AgriSupport II system was tested in a farm in Spain where
decision support systems to fulfil the needs of agricultural production different combinations of crops were proposed and experimented with.
processes (Recio et al., 2003). The overall objective of this system is to Twenty-five case studies were considered in the experiments.
provide farmers with sufficient agricultural decision-making sugges- Experimental results suggest that AgriSupport II system is able to pro-
tions like farm operation scheduling, detailed operation cost, resources vide farmers with sufficient advice about distributing agricultural
usage, and profitability analysis. Two main issues are mentioned in works. By adopting those provided advice, farmers can perform agri-
their requirement analyses. The first issue is in regards to the cultural operations with the minimum investments and achieve the
4
greatest working efficiency. Consequently, agricultural productivity the facility unit. Afterwards, tracks are assigned with a route according
will increase, which fulfils the requirement of Agriculture 4.0. to the sequential permutation. The last step is to measure the risk factor
of each route by comparing values of normalized risk indicators and
2.1.2. Multi-robot sense-act system vehicles loads. Finally, the output from this ADSS is the optimal route
Conesa-Munoz et al. (2016) proposed a multi-robot sense-act plan.
system, aiming at performing tasks automatically by using aerial and For verifying the effectiveness of this ADSS, a case study is con-
ground vehicles. The main objective of this system is to improve crop ducted in a field in Denmark. It used Electromagnetic Induction (EMI)
performances and maintain environmental quality by using vehicles to as the risk indicator measurements and GreenStar 3 as the navigation
perform agricultural tasks in large outdoor areas autonomously. In their component for tractors. The experimental result demonstrates that
proposal, aerial units are responsible for gathering environmental data those tractors with heavy loads are dispatched to the areas with low risk
while ground units are considered as mission executors in farmlands. indicator values. As a consequence, damaging impacts on soil stresses
All the vehicles are controlled by a Mission Manager, connected to a and compactions are significantly reduced. Conclusively, the proposed
Base Station computer. Within the Mission Manager, two planners are ADSS for route planning can provide farmers with the optimal plan of
designed for commanding aerial and ground units respectively. work distributions. Meanwhile, soil-sensitive fields can be protected
In regards to the aerial planner, its inputs require command signals from damages caused by tractors with heavy loads. More crops can be
from the Mission Manager, locations of obstacles in farmlands, and grown and agricultural productivity will increase. Thus, this proposal
survey areas. Then, the aerial planner splits fields into small grids ac- fulfils the requirement of Agriculture 4.0 closely.
cording to orientation, overlapping requirements, and image resolution.
A Harmony Search Algorithm (Nabaei et al., 2018) is used to generate 2.1.4. On-board decision-making approach
the optimal plan for aerial units to cover the whole area. In terms of the With the development of advanced robotics, Unmanned Aerial
ground planner, it distributes task sequences to ground vehicles by a Vehicles (UAVs) have been widely used in a board range of applica-
meta-heuristic optimization method. Its required inputs are operation tions, especially in the aspect of agriculture. Alsalam et al. (2017)
areas and commands received from the Mission Manager. It is worth proposed an on-board decision-making approach for UAVs to perform
noting that turning radius and battery capacity are both considered in agricultural operations autonomously. The objective of this work is to
this planner. The output of the ground planner is the best trajectory for detect exact locations of diseased crops and then perform corresponding
each ground vehicle to cover operation areas. operations like spraying herbicides precisely. With the precise use of
The multi-robot sense-act system was tested over 20 times in a farm herbicides, toxic damages to the fields can be greatly reduced. Mean-
in Spain. The experimental result shows that this system is able to while, deploying UAVs to perform spraying missions can improve the
generate the optimal work distribution for a site-specific herbicide working efficiency, which obviously helps farmers to increase agri-
treatment mission. Both aerial and ground units can work co- cultural productivity. The proposed approach is on the basis of the
operatively. Notifications were sent to farmers when vehicles had un- Observation, Orientation, Decision, and Action (OODA) loop shown in
expected failures. Fig. 4.
Overall, this system contributes to assigning agricultural tasks to the In Fig. 4, for collecting data, this approach takes measurements from
most appropriate work units. Farmers can obtain decision supports on ultrasonic sensors, images taken by cameras, and received commands as
agricultural work distributions. Meanwhile, farmers can supervise the inputs during the observation step. After obtaining these inputs, UAVs
entire process and manage the workflow through the multi-robot sense- start the mission and march to the target locations. During the step of
act system. Unexpected failures like internal errors of vehicles, valves mission execution, the on-board computer determines whether UAVs
delay, and work collisions are informed to the farmers. As a con- are following the correct path or not. If a UAV is flying higher than the
sequence, immediate requests on mission re-planning can be proposed. appointed height, the decision-making component will command this
Lastly, thanks to images taken by aerial units and data collected by vehicle to adjust its altitude. Meanwhile, the decision-making compo-
ground units, farmers can know exactly how many herbicides are nent is responsible for checking past waypoints in order to monitor the
needed from crops. Thus, a precise spraying can be performed by re- mission status. The action step includes operations like taking images,
ducing the amount of herbicide usages. In general, the multi-robot approaching to waypoints, and spraying herbicides. After the assigned
sense-act system is a good fit for Agriculture 4.0. mission is completed, UAVs are required to return back to the home
station and convert to the observation mode.
2.1.3. ADSS for route planning in soil-sensitive fields The on-board decision-making approach was verified through sev-
Bochtis et al. (2012) presented an ADSS to help farmers to deploy eral flight missions. The experimental result shows that UAVs can hover
agricultural vehicles in soil-sensitive fields properly. The objective of over target locations upon arrivals. The proposed approach can cor-
their work is to optimize travel paths for minimizing damages to soil- rectly guide UAVs to reach each waypoint. Target locations are ob-
sensitive fields from large-scale vehicles. On the one hand, route opti- tained based on obtained images by running an Object-Based Image
mizations can reduce energy consumptions of vehicles and improve the Analysis (OBIA) algorithm (Peña et al., 2013). Thus, UAVs are able to
working efficiency. On the other hand, it is essential to consider me-
chanical impacts of vehicles on the soil structure, especially the risk of
soil stresses and compactions (Keller et al., 2007). The shorter path a
heavy vehicle travels in the field, the less damage will this vehicle
cause.
In their proposal, the system treats the soil boundary, driving di-
rection, and potential risk indicator measurements as input data. The B-
patterns optimization algorithm is employed as the decision supporting
tool in the system. It takes four steps to generate the optimal plan.
Initially, all tracks are sequentially sorted based on the derived relative
risk map. A threshold is assigned to each track, as an indicator of dis-
tinctions between low and high risks. After the sequential permutation
of tracks is generated, the number of routes is estimated. A single route
is composed of sequential work operations, including filling resources,
forwarding to target locations, performing operations, and returning to Fig. 4. The OODA loop in the proposed on-board decision-making approach.
5
hover over targets and perform precise spraying operations. Con-

Supporting period
clusively, the proposed approach can provide farmers with decision
Short-term supports on route guidance of UAVs. Also, it enables UAVs to perform
Short-term
Short-term
Short-term
precise spraying operations autonomously, which greatly improves the
working efficiency of agricultural operations and reduces toxic damages
from excessive chemical usages. From this perspective, the proposed
decision-making approach is closely related to the requirement of
Agriculture 4.0.
Guidance for farm operation scheduling
2.1.5. Summary of ADSSs for mission planning

Plans for herbicide treatment missions
Travel paths for agricultural vehicles
Through the review result from Section 2.1.1–2.1.4, the differences

between ADSSs are concluded in Table 1.
In Table 1, these four ADSSs are designed to serve the aspect of
mission planning in Agriculture 4.0, providing farmers with guidance
Flying paths for UAVs
about agricultural operations like chemical treatment and scheduling of

agricultural machineries. It is noted that all ADSSs collect environ-
Decision support
mental and crop-related data as inputs to their decision supporting tools

(models). However, only AgriSupport II system concerns the economic
data. The optimization algorithm is a favourable approach for gen-
erating the solutions. Unfortunately, all ADSSs under this category lack
consideration of mid-term and long-term planning.
2.2. ADSSs for water resources management

B-patterns optimization algorithm
Current researches of ADSSs for water resources management are

generally concerning the irrigation systems. An irrigation system should
Meta-heuristic algorithm
provide farmers with effective decision supports on controlling the

amount of water applied to crops and maintaining landscapes (Alarcon
CPLEX optimizer
et al., 2016; de Wit and Crookes, 2013). It aims at ensuring wettability

of soil fields and basic water needs from crop growths with the
OODA loop
minimum water usages.

Three ADSSs from literatures are reviewed in the next sub-sections
Tool
to summarize contributions from current works.
2.2.1. Smart irrigation decision support system (SIDSS)

The smart irrigation decision support system (SIDSS) was proposed
Environmental, crop-related, and economic data
by Navarro-Hellin et al. (2016). Traditionally, irrigation activities are

planned by an agronomist according to resources like collected me-
teorological data, crop characteristics, and soil measurements. The
objective of the proposed SIDSS is to generate irrigation plans in a more
Environmental and crop-related data
efficient and accurate way with the same resources. With the help of
SIDSS, irrigation activities can achieve better performances with the
minimum water usages.
In their proposal, SIDSS is composed of three components: a col-
lection device, a weather station, and a decision-making component.
The framework of SIDSS is presented in Fig. 5. In regards to the inputs
Data source
to SIDSS, the first two components collect sensing data (volumetric

water content depth, soil water potential, and soil temperature) and
meteorological information (rainfall, wind speed, temperature, relative
humidity, global radiation, dew point, and vapor-pressure deficit). The
decision-making component is in charge of generating decision sup-
ports based on reasoning results. The reasoning process adopts two
machine learning techniques: Partial Least Squares Regression (PLSR)
ADSS for route planning in soil-sensitive fields
(Mehmood et al., 2012) and Adaptive Neuro Fuzzy Inference Systems

Summary of ADSSs for mission planning.
(ANFIS) (Svalina et al., 2013). PLSR is used to deduct unnecessary

variables when soil measurements and meteorological data appear re-
On-board decision-making approach
dundant. ANFIS is employed to minimizing estimated errors under a

given threshold. The output of SIDSS presents the optimal irrigation
Multi-robot sense-act system
plan, indicating the amount of water usages and the time for irrigation
activities.
AgriSupport II system
The smart irrigation decision support system was testified and

evaluated in lemon tree plantations in south-east Spain, where water
resources are very limited. The experimental result demonstrates that
DSS name
SIDSS is able to provide farmers with an irrigation report, which is

Table 1
better than the decisions made by an agronomist. The irrigation report

indicates the precise amount of water usages and the time for irrigation
6
Fig. 5. The framework of SIDSS.
activities. Conclusively, SIDSS meets the requirement of Agriculture 4.0 in irrigated agriculture and improve the irrigation performance, a de-
by allocating natural resources reasonably. cision support system for Middle Rio Grande Conservancy District
(MRGCD DSS) is presented by Oad et al. (2009). The objective of
MRGCD DSS is to analyse water demands in service areas and schedule
2.2.2. Fuzzy decision support system (FDSS)
available water resources to fulfil these demands precisely and effi-
Giusti and Marsili-Libelli (2015) proposed a fuzzy decision support
ciently.
system (FDSS) and it is an improvement over an existing irrigation web
There are three components in MRGCD DSS: a water demand
service based on IRRINET model (Giannerinii and Genovesi, 2011). As a
module, a scheduling module, and a water supply module. Firstly, the
fundamental platform in agriculture planning, FDSS is developed to
water demand module calculates the water shortage capacity by the ET
assist farmers in scheduling daily irrigation activities by combining a
Toolbox, according to input variables like irrigated areas, crop types,
predictive model of soil moisture and an inference system. The objec-
soil types etc. The Integrated Decision Support Consumptive Use
tive of FDSS is to improve irrigation performances and reduce un-
(IDSCU) model is adopted to calculate the amount of water which
necessary water usages.
equals to the water shortage capacity. Secondly, the water supply
In their proposal, FDSS consists of two parts: a predictive model and
module presents the layout of conveyance network, including connec-
an irrigation decision maker. The former component takes meteor-
tions between canals, laterals, and service areas. The flow capacity and
ological information, water resources availability, and crop character-
conveyance losses are computed in this module. Thirdly, the scheduling
istics as inputs. It is worth noting that water balances are considered as
module generates water delivery plans for fulfilling demands from
well. The predictive model generates the variable of soil moisture,
crops. As the output from MRGCD DSS, the delivery plan includes the
which is compared with a pre-defined threshold later. If soil moisture is
number of laterals, irrigation time, irrigation frequency, and the
lower than the threshold, the irrigation decision maker is trigged to
amount of water.
plan the next irrigation activity. The decision maker component con-
MRGCD DSS was verified in a farmland with 28 lateral canals. The
siders three inputs: daily variation of Growing Degree Days (GDD),
irrigation duration, irrigation interval, and flow rates were considered
cumulative rain forecast, and crop evapotranspiration. The inference
as evaluation criteria. Though certain discrepancies do exist between
method used in this component is the Fuzzy C-Means algorithm.
experiments and real practices, MRGCD DSS achieves better perfor-
According to the decision rule set, the generated inference result sug-
mances in most laterals. Conclusively, MRGCD DSS is able to provide
gests the amount of water to irrigate.
sufficient decision supports for farmers on planning irrigation activities.
The proposed FDSS was tested on three crops: corn, kiwi, and po-
Besides helping farmers to save water resources, it also reduces river
tato. Comparing with the previous research work (IRRINET), the ex-
diversions. Therefore, MRGCD DSS fulfils the requirement of
perimental result demonstrates that the performance of FDSS is much
Agriculture 4.0.
better, saving up to 13.55, 18.3, and 72.95 water units for irrigating
three crops respectively. Therefore, FDSS is able to provide farmers
with effective irrigation advice and help them to allocate water re-
2.2.4. Summary of ADSSs for water resources management
sources more reasonably. From this point of view, FDSS definitely fulfils
Table 2 presents the differences between the selected ADSSs for
the requirement of Agriculture 4.0.
water resources management.
In Table 2, SIDSS, FDSS, and MRGCD DSS provides scheduling plans
2.2.3. MRGCD DSS of irrigation. It is concluded that environmental and crop-related data
Due to a decrease of annual rainfall and misuse of water resources, are essential. Predictive models (PLSR, ANFIS, and IDSCU model) and
western United States is suffering from serious drought years and has decision rules (Fuzzy C-Means algorithm) are used to generate the op-
difficulties on irrigating crops. In order to reduce water consumptions tions. The supporting period is limited within a short term.
7
Table 2
Summary of ADSSs for water resources management.
DSS name Data source Tool Decision support Supporting period
SIDSS Environmental and crop-related data PLSR and ANFIS Irrigation reports Short-term and mid-term
FDSS Environmental and crop-related data Fuzzy C-Means algorithm Advice on scheduling irrigation Short-term
MRGCD DSS Environmental and crop-related data IDSCU model Advice on scheduling irrigation Short-term
2.3. ADSSs for climate change adaptation linked: climate, ecology, and socio-economy. The climate component is
used to analyze long-term and seasonal climate data (Franke and
Recently, researchers have been aware of the significance of climate Kostner, 2007), estimating temperature trends, rainfall, precipitation,
change adaptation in agricultural decision support systems (Rickards Ellenberg index, Huglin index, Schwarzel index, etc. The effects of cli-
and Howden, 2012; El-Sharkawy, 2014; Weller et al., 2016). To sum- mate change are then considered as inputs to the ecology component.
marize current contributions, three research works are reviewed in the Within this ecology component, various models are designed, including
following sub-sections. VEGPER (calculating the length of vegetation periods), ONTO (calcu-
lating crop developments in different stages), SVAT-CN (calculating
2.3.1. OCCASION nitrogen and Soil-Vegetation-Atmosphere-Transfer), EROSION (calcu-
In order to maintain agricultural sustainability under climate lating regional water balance), GLPROD (calculating grassland pro-
change, Schutze and Schmitz (2010) proposed a planning system, ductivity and forage quality), etc. In regards to the socio-economy
named OCCASION, for optimizing climate change adaptation strategies component, it has two models: PECG and RAUMIS. The former model is
in irrigated agriculture. The objective of this system is to provide used to evaluate costs and benefits by applying different climate change
farmers with estimated water demands for irrigation according to as- adaptation strategies, while the latter one assesses the impacts of cli-
sessments of climatic variability. mate change over future agricultural activities.
In their proposal, a methodology, named Stochastic Crop-Water LandCaRe DSS takes the following steps to provide end users with
Production Function (SCWPF), is presented, enabling to quantify im- decision supports. Firstly, agricultural problems and scenario simula-
pacts of climate change on irrigation activities. The first step of SCWPF tions are extracted from users’ definitions, including climate, soil, land
is to create climatic data by using LARS-WG stochastic weather gen- use, and so on. Based on the input data, a model is selected to assess the
erators (Semenov et al., 1998). After synthesizing climate scenarios, impacts of climate change over agricultural activities. The output from
SCWPF adopts the One-Dimensional Soil-Vegetation-Atmosphere the employed model is presented by maps, diagrams, tables, and sta-
Transfer (SVAT) model (Mo et al., 2005) to simulate crop productions, tistics.
crop yields, water and nitrogen conditions. The second step of SCWPF is The proposed LandCaRe DSS was tested in two contrasting regions
to construct a complete Crop-Water Production Function (CWPF) by the in Germany. The experimental result shows that this system is able to
Global Evolutional Technique for Optimal Irrigation Scheduling (GET- predict future meteorological information and the length of vegetation
OPTIS). Variables from the weather generators and the SVAT model are periods. Meanwhile, LandCaRe DSS is also capable of analyzing water
treated as inputs to GET-OPTIS. The output of GET-OPTIS is the po- demands for irrigation activities. Conclusively, successful demonstra-
tential CWPFs which represent the potential optimal plan for irrigating tions imply that LandCaRe DSS can provide stakeholders and farmers
the maximum crop yield with the minimum water volumes. In the third with sufficient decision supports on agricultural activities under climate
step, statistical characteristics of all potential CWPFs is computed in a change. Thus, the proposed ADSS fulfils the requirement of Agriculture
non-parametrical way for the purpose of identifying the global optimal 4.0 by adapting to climate change.
irrigation scheduling plan.
The proposed planning tool was tested and evaluated in a field in 2.3.3. GIS-based DSS
France. A basic scenario without rainfall and a complex scenario with For quantifying potential impacts of climate change in Semi-Arid
variability of rainfall are considered. According to evaluation results, Tropical (SAT) regions, Kadiyala et al. (2015) presented an agricultural
OCCASION has achieved the following contributions: (1) farmers can decision support system by integrating a Decision Support System for
obtain adequate information about weather and soil fields. (2) this Agrotechnology Transfer (DSSAT) crop simulation model and a Geo-
planning system allows farmers to assess the potential impacts of cli- graphical Information System (GIS) component. The objective of GIS-
matic variability on farmlands. (3) OCCASION assists farmers in ad- based DSS is to assist farmers in making proper agronomic decisions
justing irrigation scheduling plans, taking climate change into account. under climate change to increase the productivity of ground nuts.
Conclusively, OCCASION fulfils the requirement of Agriculture 4.0. The proposed system consists of four major components: a GIS
component, a DSSAT crop simulation model, a query system, and a
2.3.2. LandCaRe DSS spatial output generating system. Firstly, the GIS component receives
LandCaRe DSS, as an interactive decision support system, was pre- spatial information about position, soil, and weather. Then, these data
sented by Wenkel et al. (2013). The objective of this research is to are considered as inputs to the crop simulation model (DSSAT). Crop
support farmers and stakeholders on adapting farm management to growths and yield can be simulated through the DSSAT model (Jones
climate change as follows: et al., 2003). Based on the spatial information and crop data, the pro-
posed ADSS can provide the following functionalities.
• Providing both historical and predictive climate data for end users
under a clear visualization. • Prediction on future climate characteristics: This functionality mainly
• Providing multi-scenario and multi-model simulations for analyzing concerns the predicting future rainfall and temperature.
uncertainty. • Observation on base yields: This functionality performs productivity
• Providing potential strategies for climate change adaptation. analyses for the selected yields.
• Providing end users with assessments of climate change over agri- • Assessment of climate change over crop yields: This functionality con-
cultural activities. siders how climate change may influence the crop yields.
In the proposed LandCaRe DSS, three components are closely Four adaptation strategies were tested by the proposed ADSS.
8
Table 3
Summary of ADSSs for climate change adaptation.
OCCASION Environmental and crop-related data CWPF and GET-OPTIS Assessments of climate change Short-term
LandCaRe DSS Environmental, crop-related, and economic VEGPER, ONTO, SVAT-CN, Advice on farm management under climate Short-term, mid-term, and long-
data etc. change term
GIS-based DSS Environmental and crop-related data DSSAT model Advice on increasing productivity under Short-term
climate change
Firstly, 10% longer life cycle cultivar was simulated. By applying cli- In order to provide these decision supports and achieve both eco-
mate change adaptation, groundnut productivity increased by 6.8% and nomic and environmental objectives, the supply chain is mathemati-
2.1% in southern regions and highland regions respectively. Secondly, cally formulated as a multi-objective linear programming model (Ji
the proposed ADSS took drought tolerance into account. The simulation et al., 2018). The required inputs to the economic objective function are
result showed significant improvements in all test fields. Thirdly, heat warehouse costs for storing beef and transportation costs for export
tolerance was considered in simulations and the result demonstrated departures and import arrivals. The input to the environmental objec-
positive responses. Fourthly, adaptation strategy supplemental irriga- tive function is the total amount of CO2 emissions released during the
tion was tested. All regions can benefit from precision irrigation advice transportation. Meanwhile, several constraints are defined, such as a
and base yields increased by 23.6%, 21.9%, 19.3% and 16.1% respec- balanced beef inventory in all warehouses, the capability of meeting
tively. Conclusively, the proposed ADSS is able to predict future climate market demands, the minimum travel paths, etc. The MOLP model is
characteristics and monitor conditions of groundnut fields. Famers can computed by the ILOG-OPL development studio and CPLEX 12.2 opti-
obtain proper decision supports on performing agricultural activities mization solver. The computed Pareto frontier represents the optimal
under climate change. Therefore, it fulfils the requirement of distribution strategy, as the output of the MOLP model. The optimiza-
Agriculture 4.0 by considering climate change adaptation. tion solver employs the Ɛ-constraint method (Dehghan et al., 2014),
meaning that the economic objective function is selected for optimi-
2.3.4. Summary of ADSSs for climate change adaptation zation, while the environmental objective function is treated as an
Table 3 summarizes the research work of ADSSs for climate change additional constraint.
adaptation. The proposed supply chain was implemented and tested in Brazil.
In Table 3, LandCaRe DSS concerns the widest range of data sources The idea is to export beef from Brazil to Europe. The experimental re-
when adapting to climate change, while economic data are not involved sult shows that the MOLP model is able to generate the optimal dis-
in OCCASION and GIS-based DSS. It seems that these three ADSSs fa- tribution strategy within a reasonable time. The distribution strategy
vour model-based approaches and various models are adopted for as- indicates the number of tracks, type of tracks, and route for each track.
sessing the impacts of climate change. By following this strategy, the minimum total transportation cost and
the minimum amount of released CO2 emissions are achieved at the
2.4. ADSSs for food waste control same time. Furthermore, the optimized logistics network not only en-
ables a quick delivery with the minimum cost, but also ensures the food
Optimizing the supply chain is widely acknowledged as one of the quality and safety for consumers, avoiding food waste during the
most effective approaches for avoiding food waste. On the one hand, transportation. Therefore, the proposed supply chain fulfils the re-
the optimized supply chain enables to deliver agricultural products to quirement of Agriculture 4.0.
the nearest destinations within the minimum time (Hamprecht et al.,
2005). On the other hand, consumers, as the end of supply chains, can 2.4.2. Quality sustainability decision support system (QSDSS)
reflect needs of markets. Responses from consumers are essential be- Increased demands for food quality and safety have been challen-
cause they can provide adequate information, assisting farmers in ad- ging the global supply chain seriously. Logistics managers prefer to use
justing plans of agricultural activities (Muller et al., 2009). In this decision support systems to optimize delivery strategy for ensuring the
section, three proposals are reviewed for analyzing current ADSSs for food quality and safety. Ting et al. (2013) presented a quality sustain-
food waste control. ability decision support system (QSDSS) based on the association rule
mining and the Dempster’s rule of combination. The main objective of
2.4.1. MOLP-based beef supply chain QSDSS is to discover the association measures between logistics flows
The globalization of supply chains enables cross-border deliveries and provide logistics managers with decision supports for red wine
for agricultural products. Due to increased distances between partners, deliveries, including transportation modes, types of delivered goods,
ADSSs can be used to determine suppliers, distribution channels, delivery routes, etc.
transportation modes, inventories at each warehouse, and so on The workflow of the proposed QSDSS is presented in Fig. 6. Firstly,
(Cordeau et al., 2006; Harris et al., 2011). Soysal et al. (2014) applied a the knowledge base contains pre-processed logistics flow data. These
Multi-Objective Linear Programming model to a beef supply chain in data are then extracted by the association rule mining component in
order to demonstrate how farmers can benefit from a well-organized order to detect interesting association rules based on support and
logistics network. The economic and environmental objective functions confidence measures (Le and Lo, 2015). The Apriori algorithm (Li et al.,
are considered in the proposed beef supply chain. The objective of this 2016) is used in the association rule mining component for identifying
ADSS is to minimize the total transportation costs and the total amount the potential associations and assigning a weight to each association.
of released greenhouse gas emissions. Logistics managers are allowed to input a new delivery request with
In the proposed supply chain, following advice have to be offered information about product types, quantities, and transportation modes.
during the transportation processes: After receiving the new case, the Dempster’s rule of combination
component can aggregate related associations between cases and gen-
• Inventory amounts of beef in each warehouse. erate the most appropriate logistics route on the basis of assigned
• Number of tracks used during transportation. weights. The output of QSDSS is the route with the highest weight, as
• Type of tracks used during transportation. the optimal delivery strategy.
• Routes for each track. The proposed QSDSS was verified and tested through a red wine
9
• Selections of supply points.

• Products from selected supply points.
• Selections of delivery vehicles.
• Scheduling of routes for delivery vehicles.
• Scheduling of delivery time.
In their proposal, each component of e-grocery deliveries is treated
as an agent with specific behaviors. For example, supply points are
treated as location agents, while delivered items are regarded as pro-
duct agents. With all the agents, an agent-based simulation is formed
and used to randomly generate demands from consumers and un-
certainty in food decay. The required inputs to the proposed DSS are the
number of supply points, number of delivery vehicles, quantities and
the quality of delivered items, etc. An optimization component is used
to schedule pick-up and delivery options. The scheduling plan is com-
puted by a heuristic optimization algorithm (Tarantilis and Kiranoudis,
2001), aiming at minimizing delivery distances and maximizing food
Fig. 6. The workflow of QSDSS. quality. After responding to all the requests from consumers, a relocate
operator is used to evaluate current pick-up options and delivery stra-
tegies. The output of the proposed DSS is a complete delivery plan,
industry in Collazoni. A set of potential associations was successfully
displayed through a graphical interface. This delivery plan demon-
extracted from the knowledge base, 145 rules in total. Meanwhile, a
strates delivery routes, delivery time, inventory distributions, etc. Ad-
complete logistics route is aggregated based on the most interesting
ditionally, the quality of delivered products and remaining items, as
rules. Four members are selected to evaluate QSDSS, including senior
well as the amount of wasted food are all presented through the in-
managers, project consultants, and logistics coordinators. The experi-
terface.
mental result shows that QSDSS is able to improve the delivery per-
The proposed DSS was tested in Austria with 255 local stores, 24
formance in the following aspects.
vehicles, and one depot. The experimental result shows that more than
• Securing quality level. After adopting QSDSS, the product return rate
one thousand items are successfully delivered or picked up by con-
sumers. By employing the optimal delivery plan, food waste is sig-
from consumers shows a decrease by 60%.
• Reducing logistics costs. QSDSS enables to reduce costs of re-shipping
nificantly reduced and food quality is assured during the whole process.
From this perspective, the proposed DSS for e-grocery deliveries fulfils
and continual shipping by 45%.
• Improving satisfaction of consumers. Due to quality assurance, the
the requirement of Agriculture 4.0.
number of damaged red wine decreases from 2134 to 530, avoiding

returns from consumers. 2.4.4. Summary of ADSSs for food waste control
• Enhancing logistics visibility. Hidden information (interesting asso- Table 4 demonstrates the analytic result from Section 2.4.1–2.4.3.
In Table 4, the selected ADSSs are in regards to controlling food
ciations) is mined by QSDSS and displayed to the logistics managers
through a user-friendly interface, providing them with real-time waste from the point of view of optimizing the supply chain. As a
decision supports. consequence, the category of economic data is the main factor con-
sidered in the research work. Similar to ADSSs for mission planning,
Conclusively, by employing QSDSS, food quality and safety can be ADSSs for food waste control generate the delivery plans by employing
greatly assured. Therefore, QSDSS fulfils the requirement of Agriculture the optimization algorithms. Unfortunately, none of these works pay
4.0 in the aspect of avoiding food waste. attention to the long-term planning.
3. Evaluation and upcoming challenges

2.4.3. Decision support system for e-grocery deliveries
Unnecessary costs and food waste are usually resulted from in- After presenting the thirteen ADSSs, we evaluate each one of them
appropriate delivery options and unreasonable inventory distributions. from eight aspects, including their accessibility, scalability, interoper-
For avoiding food losses in e-grocery deliveries, Fikar (2018) proposed ability, etc. According to the evaluation result, future trends and up-
a decision support system on the basis of agent-based simulations and coming challenges are summarized when developing new ADSSs. It is
dynamic routing procedures. The overall objective of this system is to promising that future ADSSs can better serve Agriculture 4.0 by over-
optimize the inventory distributions and generate the optimal delivery coming detected challenges.
strategy for logistics managers.
It is assumed that e-grocery providers are fully aware of the in- 3.1. Evaluation of selected ADSSs
ventory distributions and quality of products at all times, while con-
sumers prefer to receive those products with longer shelf lives. The Table 5 and 6 presents the evaluation criteria and scores: (i) if the
proposed DSS for e-grocery deliveries are expected to provide logistics aspect is fully considered and described with technical details, it
managers with following decision supports. achieves three stars (best); (ii) if the aspect is partially mentioned, but
Table 4
Summary of ADSSs for food waste control.
MOLP-based beef supply chain Economic data MOLP and Ɛ-constraint method Delivery plans of transporting beef Short-term
QSDSS Economic data Apriori algorithm and Dempster’s rule of combination Delivery plans of transporting wine Short-term
DSS for e-grocery deliveries Economic data Heuristic optimization algorithm Delivery plans of e-grocery Short-term and mid-term
10
Table 5
Evaluation results of selected ADSSs (I).
ADSS name Accessibility Scalability Interoperability Uncertainty and dynamic factors
AgriSupport II system ★★★ ★★★ ★★★ ★★

Multi-robot sense-act system ★★★ ★★★ ★★★ ★★
ADSS for route planning in soil-sensitive fields ★★★ ★★★ ★★★ ★
On-board decision-making approach ★★ ★★★ ★★★ ★★
SIDSS ★ ★★★ ★★★ ★★★
FDSS ★ ★★★ ★★★ ★★
MRGCD DSS ★★★ ★★★ ★★★ ★★
OCCASION ★ ★★★ ★★★ ★★★
LandCaRe DSS ★★★ ★★★ ★★★ ★★★
GIS-based DSS ★ ★★★ ★★★ ★★
MOLP-based beef supply chain ★ ★★★ ★★★ ★
QSDSS ★ ★★★ ★★★ ★
DSS for e-grocery deliveries ★★★ ★★★ ★★★ ★★★
Total Stars of evaluation criteria 26/39 39/39 39/39 27/39
without further explanations, it achieves two stars (medium); and (iii) if fields, etc. Some good examples of GUIs are presented in Fig. 7.
the aspect is not addressed at all, it achieves one star (worst). The as- In Fig. 7, the GUI of AgriSupport II system enables to enter data and
pects with more stars indicate that they had been thoughtfully con- present generated decision supports for agricultural activities. More-
sidered in the selected ADSSs, while, those with fewer stars drew less over, operators can perform machinery analysis, task programming,
attention and are possible to raise potential challenges in the future. and economic analysis. They can also obtain recommendations (a list of
Lastly, Table 7 displays the overall remarks achieved by each ADSS. The feasible options for operations) through this GUI. The GUI of the multi-
remark is measured through achieved stars divided by total 24 stars. robot sense-act system displays data generated by the Mission Manager
The evaluation criteria are selected from the Software Quality component, such as plans, execution states, alarms, and so on, guiding
Requirements and Evaluation (SQuaRE) standard, including accessi- operators through different workflow steps.
bility, interoperability, scalability, and functionality completeness Generally, the graphical visualization can hide complexity of ADSSs,
(ISO/IEC 25010:2011, BSI Standards Publication, https://www.iso.org/ enabling farmers to manage agricultural activities more easily and ef-
standard/35733.html). ficiently. Thus, as an essential component, a GUI can improve the ac-
In Table 7, it is concluded that the average remark of all thirteen cessibility of ADSSs. However, nearly half of the selected ADSSs have
ADSSs achieves 16.31 stars (67.95%), which means current ADSSs not addressed this issue. Furthermore, the GUIs provided by current
cannot serve Agriculture 4.0 perfectly and they still have room for ADSSs sometimes display the computation processes and require com-
improvement. The best one among selected ADSSs is OCCASION, which plex text inputs, leading to noises and confusions for farmers.
achieves 20 stars out of 24, because this ADSS covers most of the
evaluation criteria. While the MOLP-based beef supply chain achieves
3.1.2. Scalability
12 stars out of 24 and it takes the last position due to its inadequate
This aspect addresses the capability of ADSSs to process the growing
consideration on accessibility, uncertainty, re-planning, etc. For further
amount of missions (Chu et al., 2016). Meanwhile, the scalability in-
explanation, evaluation details are given in the following sub-sections.
dicates the extendibility of an ADSS. For example, extra components
can be added into an ADSS for enriching its functionality.
3.1.1. Accessibility According to Table 5, it is satisfied that all thirteen ADSSs have
This aspect mainly refers to the graphical user interface (GUI) of an concerned the aspect of scalability (achieving all 39 stars). For example,
ADSS (Shirogane et al., 2008). A GUI is necessary because it provides QSDSS is composed of several components. It is possible to add new
operators with the possibility of establishing new missions, monitoring components in its architecture. Meanwhile, operators can define new
mission status, checking available information, etc. Meanwhile, it cases when the number of red wine orders increases, which means
should visualize the generate decision supports for users. QSDSS can deal with the growing amount of missions. LandCaRe also
In Table 5, we note that seven of selected ADSSs provide GUIs for pays attention to this aspect. It employs multiple decision support
operators (stakeholders and farmers), including AgriSupport II system, models to generate strategies for agricultural activities. It is promising
multi-robot sense-act system, ADSS for route planning in soil-sensitive to introduce more models for enriching its functionalities.
Table 6
Evaluation results of selected ADSSs (II).
ADSS name Re-planning Expert knowledge Prediction and forecast Analysis on historical information
AgriSupport II system ★ ★★ ★ ★
Multi-robot sense-act system ★ ★★ ★ ★
ADSS for route planning in soil-sensitive fields ★ ★ ★★ ★★
On-board decision-making approach ★★ ★ ★ ★
SIDSS ★ ★★★ ★★★ ★
FDSS ★ ★ ★★★ ★
MRGCD DSS ★ ★★ ★★ ★★★
OCCASION ★★★ ★ ★★★ ★★★
LandCaRe DSS ★ ★★ ★★★ ★
GIS-based DSS ★ ★ ★★★ ★
MOLP-based beef supply chain ★ ★ ★ ★
QSDSS ★ ★★ ★ ★
DSS for e-grocery deliveries ★ ★ ★ ★★
Total stars of evaluation criteria 16/39 20/39 25/39 19/39
11
Table 7
Overall remarks of selected ADSSs.
ADSS name Overall remark ADSS name Overall remark
AgriSupport II system 16/24 (66.67%) OCCASION 20/24 (83.33%)

Multi-robot sense-act system 16/24 (66.67%) LandCaRe DSS 19/24 (79.17%)
ADSS for route planning in soil-sensitive fields 16/24 (66.67%) GIS-based DSS 16/24 (66.67%)
On-board decision-making approach 15/24 (62.50%) MOLP-based beef supply chain 12/24 (50.00%)
SIDSS 18/24 (75.00%) QSDSS 13/24 (54.17%)
FDSS 15/24 (62.50%) DSS for e-grocery deliveries 17/24 (70.83%)
MRGCD DSS 19/24 (79.17%)
Fig. 7. Examples of presented GUIs: (a) AgriSupport II system; (b) Multi-robot sense-act system.
Conclusively, current research works have achieved great con- For those ADSSs which have not concerned uncertainty and dy-
tributions to the aspect of scalability. namic factors, such as the MOLP-based beef supply chain, it is sug-
gested that authors should pay attention to this issue. Because un-
certainty like meteorological conditions may cause a delay during
3.1.3. Interoperability
international deliveries.
On the one hand, interoperability emphasizes on integrating func-
Conclusively, uncertainty and dynamic factors should not be ig-
tions and knowledge from heterogeneous components in a single ADSS,
nored in ADSSs and further improvements on this issue are expected in
on the other hand, it represents that an ADSS can work with other
the future.
external components or systems (Teixeira et al., 2018). For instance, an
ADSS for climate change adaptation can link with an external weather
station to obtain the meteorological information. 3.1.5. Re-planning
According to Table 5, all selected ADSSs have taken interoperability Unexpected failures may occur when performing agricultural ac-
into account (achieving all 39 stars). For instance, components of tivities. Therefore, integrating re-planning mechanisms (Zhou et al.,
QSDSS can work with each other cooperatively. After pre-processing, 2018) into ADSSs seems to be a promising approach. The re-planning
data formats of product types, quantity, delivery routes, and transpor- mechanism is supposed to enhance the robustness of decision supports
tation modes are unified. Thus, heterogeneous components can share a by adjusting current strategies or generating new ones.
common understanding for the collected information. Another example Unfortunately, in Table 6, the aspect of re-planning draws the least
can refer to SIDSS. This proposed decision support system can work attention among all criteria (19 stars out of 39). Only two ADSSs have
with an external weather station for collecting the meteorological in- concerned this aspect: multi-robot sense-act system and OCCASION. In
formation. Sensing data from collection devices can also be transmitted the former ADSS, authors mentioned that robot teams can re-plan the
to the decision support component in SIDSS. overall tasks in case of failure of one unit. Because an unexpected
Overall, the developers of ADSSs have made great efforts on the failure may stop the entire work until the machine is repaired. This
aspect of interoperability. feature of fault tolerance can greatly enhance the robustness of this
ADSS. While in the latter ADSS, OCCASION focuses on scheduling ir-
rigation activities under climate change. As the environment is dyna-
3.1.4. Uncertainty and dynamic factors
mically changing with times, a re-planning process is a necessity to
Uncertainty and dynamic factors may cause unexpected results.
adjust current adaptation strategies.
Thus, ADSSs should consider these changes during runtime (Verbeke,
It is disappointing that the rest of ADSSs has not covered the aspect
2005).
of re-planning. Therefore, this is a serious challenge for future ADSSs.
In Table 5, it is glad to see that more than half of selected ADSSs
have addressed the issue of uncertainty and dynamic factors (achieving
27 stars out of 39), especially for those ADSSs for climate change 3.1.6. Expert knowledge
adaptation (achieving 8 stars out of 9). For instance, SIDSS considers Knowledge from experienced experts is highly valuable for ADSSs
soil temperature, soil water potential, rainfall, wind speed, tempera- when generating the feasible decision supports (Poch et al., 2004).
ture, humidity, and global radiation when generating the decision Moreover, experts can adjust inappropriate strategies.
supports. These dynamic variables absolutely have tremendous impacts In Table 6, we detect that six of selected ADSSs have employed the
on irrigation activities. Adequate rainfall will surely reduce the amount expert knowledge (achieving 20 stars out of 39). For example, QSDSS is
of water usages and the frequency of irrigation activities. on the basis of the Dempster’s rule of combination. The decision rule set
12
is pre-defined by experienced users or domain experts. Similarly, the 2016). Simplified and user-friendly GUIs enable farmers to get
supply network module of MRGCD DSS takes opinions of experienced started with ADSSs more quickly. Data visualizations like showing
users as additional constraints. Under this circumstance, the generated results in formats of map, table, list, line chart, pie chart, and flow
decision supports will better fit users’ needs and their decision-making chart are especially welcomed. Operations like dragging, clicking,
styles. and drawing on portable devices are also acceptable for farmers.
Due to the limitations of computation time and complexity of Unnecessary text inputs and displays of computation processes
agricultural problems, an ADSS may provide users with inaccurate should be avoided in GUIs because such information may cause
decision supports, sometimes even wrong advice. Therefore, it is worth tremendous confusions from farmers’ point of view. After all,
considering to adopt knowledge from experienced users and domain farmers care more about obtaining decision supports on how to
experts. However, current ADSSs remain to be improved in this aspect. perform agricultural activities in the most efficient way, not how the
strategies and solutions are computed.
3.1.7. Prediction and forecast • Enriching decision supports for the whole life cycle of Agriculture 4.0: In
Predictions of productivity, market fluctuations, and costs (Patel Agriculture 4.0, an ADSS is supposed to provide farmers with ade-
et al., 2018) may enable ADSSs to generate more accurate decision quate advice during the whole life cycle. According to the duration
supports, while forecasts of meteorological information (Pham and of agricultural activities, short-term, mid-term, and long-term
Kamei, 2012) are especially helpful when planning agricultural activ- planning is defined (Francis et al., 2008). A short-term planning
ities. covers tactical day-to-day decision-making activities, such as as-
In Table 6, seven of selected ADSSs support the function of pre- signing agricultural tasks to the most appropriate machineries,
diction and forecast, such as SIDSS, FDSS, OCCASION, and so on. For generating the optimal travel paths for each machinery, scheduling
example, a predictive model of soil moisture is designed in FDSS. This daily and weekly irrigation activities, etc. A mid-term planning
model concerns input variables of growing degree days, crop evapo- should offer seasonal decision supports for farmers. For instance,
transpiration, and rainfall. The output is in regards to the prediction of fertilization is usually performed by farmers based on their own
soil moisture, which can help the irrigation decision maker component observations and experiences in the past, leading to imprecise che-
to generate more accurate advice. For OCCASION, the impacts of the mical usages and causing seriously damages to soil fields and crops.
predicted climate variability on the maize growth and irrigation are However, with the help of mid-term planning, ADSSs can provide
considered during experiments. farmers with detailed advice about the perfect time to fertilize, the
However, current contributions to the prediction and forecast in amount of chemical applications, and position of crops. Regarding
ADSSs are not enough. Thus, future improvements on this aspect re- the long-term planning, it generally refers to the yearly decision-
mains to be achieved. making activities. For example, agricultural machineries surely
suffer from equipment losses. After serving for several months, old
3.1.8. Analysis on historical information and damaged components have to be replaced by new ones. By
Historical data and strategies contain valuable information which monitoring the status of each machinery, ADSSs can notify farmers
can improve the quality of future decision supports (Poczeta et al., about which machineries are non-operational anymore and what
2018). For example, considering historical strategies as a training set, components should be bought for replacement. Unfortunately, cur-
machine learning techniques can be adopted for learning successful rent ADSSs mainly focus on short-term planning, lacking con-
experiences from the training set. siderations on mid-term and long-term planning. Therefore, it is
In Table 6, around one third of selected ADSSs have performed urgent to integrate more functionalities of ADSSs and enrich deci-
analysis on historical information. For example, DSS for e-grocery de- sion supports throughout the whole life cycle of Agriculture 4.0.
liveries takes consumer’s historical pick-up preferences into account • Adapting to uncertainty and dynamic factors: Uncertainty and dynamic
when generating the optimal delivery plan. MRGCD DSS concerns his- factors do exist in agriculture, but the fact is that few ADSSs take
torical environmental data for comparing with current situations. them into account. Generally, uncertainty and dynamic factors come
However, ignorance of analyzing historical information indeed from the following aspects. Firstly, meteorological conditions have
worries us. Because historical information not only includes successful great influences on crop growths. For example, rising temperature
experiences, but also failure cases. Current agricultural activities can be may shorten the growth circle of crops. Consequently, fertilization,
performed by referring to the solutions for past cases, which had been weeding, and harvesting periods should change correspondingly as
successfully dealt with before. Thus, it is suggested that future ADSSs well (Asseng et al., 2004). ADSSs have to take uncertainty and dy-
can cover historical information. namic factors of climate change into account for providing farmers
with accurate decision supports. Secondly, conditions of farmlands
3.2. Upcoming challenges are dynamically changing as well, especially soil moisture and re-
maining nutrition in the fields (Banger et al., 2017). A low value of
According to the summary of thirteen ADSSs in Table 1–4 and the soil moisture requires farmers to perform irrigation activities more
evaluation result in Table 5–7, several upcoming challenges are de- frequently, while a high value of nutrition remaining in the fields
tected. These challenges demonstrate the potential improvements and requires farmers to fertilize less amount of manures. Monitoring on
developing trends of ADSSs for researchers in the future. By overcoming environmental changes is vital because decision supports are gen-
the detected challenges, future ADSSs can better serve Agriculture 4.0. erated based on these dynamic data. Thirdly, farmers have to handle
uncertainty and dynamic factors of economic effects from markets
• Simplifying GUIs to enhance accessibility of ADSSs: Though more than (Lin et al., 2013). The price of an agricultural product may be af-
half of selected ADSSs have provided farmers with GUIs for visua- fected by several factors like total production, logistics, inventory in
lizing gathered data, establishing agricultural missions, and mon- local warehouses, consumers’ demands, etc. Little changes in a
itoring the status of on-going missions, it is reported that farmers single factor may lead to a chain reaction. Thus, it is suggested that
sometimes have difficulties on performing desired operations ADSSs should pay attention to uncertainty and dynamic factors.
through provided GUIs. Undoubtedly, most of farmers are not fa- • Considering re-planning components: Re-planning is a challenging
miliar with computer knowledge and optimization algorithms. topic for ADSSs. On the one hand, unexpected failures and issues
Meanwhile, farmers prefer not to spend too much time in learning may arise from time to time, such as mechanical failures of an
how to use decision support systems. When designing an ADSS, it is agricultural machinery and sudden changes in weather. These fail-
suggested that the GUIs should be as simple as possible (Rose et al., ures and issues may lead to the impossibility of following original
13
strategies to complete assigned missions (Evers et al., 2014). complex environment, an ADSS is very helpful for assisting farmers in
Therefore, ADSSs should adjust current strategies or generating a performing various agricultural activities.
new solution for providing further decision supports for farmers to Based on the requirement of Agriculture 4.0, thirteen ADSSs are
continue agricultural missions. On the other hand, when an agri- selected from current literatures and projects. These ADSSs are sur-
cultural mission is being executed, ADSSs detect a better strategy for veyed through their data sources, planning tools, generated decision
carrying on the rest of the mission (Zhou et al., 2018). Consequently, supports, solved problems, and supporting periods. Eight aspects (ac-
ADSSs should inform farmers of the latest suggestions. By adopting cessibility, scalability, interoperability, etc.) are selected from the
the newly generated strategy, farmers can complete the rest of the SQuaRE standard and treated as criteria for evaluating these ADSSs and
mission more efficiently and smoothly. detecting their shortcomings. Based on the evaluation results, it is de-
• Adopting knowledge from experienced experts: Some researchers intend tected that the selected thirteen ADSSs only achieved an average re-
to develop ADSSs which resolve agricultural problems autono- mark at 16.31 stars (full remark at 24 stars). Therefore, the following
mously without any human interventions. Unfortunately, current challenges are summarized: (i) simplifying graphical user interfaces to
ADSSs have not reached such intelligent level yet. Due to the lim- improve accessibility and usability; (ii) enriching functionalities to
itation of computation time and complexity of agricultural pro- provide more adequate decision supports during the whole life cycle of
blems, ADSSs may provide farmers with inaccurate decision sup- Agriculture 4.0; (iii) adapting to uncertainty and dynamic factors to
ports, sometimes even wrong suggestions. Therefore, agricultural provide accurate decision supports; iv) considering re-planning me-
knowledge from experienced experts is needed for the purpose of chanisms to strengthen the robustness of ADSSs; (v) adopting knowl-
validating the feasibility of generated strategies and correcting the edge from experienced experts in case of adjusting inappropriate de-
mistakes in provided decision supports (Kamali et al., 2017). An cision supports; (vi) enabling prediction and forecast to prepare farmers
interactive interface should be designed in ADSSs, allowing experts for future decision-making activities; and (vii) performing analysis on
to express their knowledge and opinions. By checking generated historical information to enhance the quality of decision supports.
strategies before executing, ADSSs are able to lower the possibility Conclusively, these challenges demonstrate future development
of making mistakes. trends of employing ADSSs in Agriculture 4.0 and potential improve-
• Enabling prediction and forecast: Though predictions and forecasts are ments of ADSSs for researchers. It is promising to see that future ADSSs
especially helpful for farmers to get prepared in advance, few ADSSs can better serve Agriculture 4.0 by overcoming these challenges.
take this issue into consideration. Generally, the following four
types of predictions and forecasts are recommended. Firstly, crop Declaration of Competing Interest
growths depend on multiple factors like weather, soil, irrigation,
and fertilization. An early estimation on agricultural production is The authors declare that they have no known competing financial
helpful for farmers to detect whether certain operations should be interests or personal relationships that could have appeared to influ-
performed to improve product quality (Chlingaryan et al., 2018). ence the work reported in this paper.
Secondly, forecasts of climate change enable farmers to adjust crop
management and avoid unnecessary climatic risks (Han et al., Acknowledgements
2017). Thirdly, by detecting potential symptoms and early signs,
ADSSs are able to warn farmers about possible occurrences of pests The research leading to the presented results has been undertaken
and diseases, helping them to take certain precautions to avoid within the AFARCLOUD European Project (Aggregate Farming in the
further losses (Chougule et al., 2016). Fourthly, by analyzing market Cloud), under Grant Agreement No. 783221-AFarCloud-H2020- ECSEL-
fluctuations, ADSSs can predict consumers’ demands and the price 2017-2, partially by the Electronic Components and Systems for
trend of agricultural products. As a consequence, farmers will then European Leadership Joint Undertaking (ECSEL JU) and in part by the
produce more market-oriented products in order to gain higher Spanish Ministry of Economy, Industry and Competitiveness (Ref:
profits (MacFarlane, 1996). PCI2018-092965), and partially supported by the China Scholarship
• Performing analysis on historical information: Strategies of historical Council (CSC).
missions usually contain valuable information, including not only
successful experiences, but also failure cases. However, current References
ADSSs seldom analyze historical information. A historical mission
strategy is applicable under the circumstance of a corresponding Tekinerdogan, B., 2018. Strategies for technological innovation in Agriculture 4.0.
historical data set. It is promising to compare the real-time data set Reports. Wageningen University, Wageningen, Netherlands.
Ferrandez-Pastor, F.J., Garcia-Chamizo, J.M., Nieto-Hidalgo, M., Mora-Pascual, J., Mora-
with historical ones to generate feasible strategies for current mis- Martinez, J., 2016. Developing ubiquitous sensor network platform using Internet of
sions within a shorter computation time by using intelligent algo- Things: Application in precision agriculture. Sensors 16 (7). https://doi.org/10.
rithms like machine learning, deep learning, bio-inspired algorithms 3390/s16071141.
Wolfert, S., Ge, L., Verdouw, C., Bogaardt, M.J., 2017. Big data in smart farming – a
(Ali et al., 2018). Because similar patterns between historical and review. Agric. Syst. 153, 69–80. https://doi.org/10.1016/j.agsy.2017.01.023.
current data sets may be recognized and matched. Successful ex- Liakos, K.G., Busato, P., Moshou, D., Pearson, S., Bochtis, D., 2018. Machine learning in
periences in past cases can be used as references in performing agriculture: a review. Sensors 18 (8). https://doi.org/10.3390/s18082674.
Lopez-Riquelme, J.A., Pavon-Pulido, N., Navarro-Hellin, H., Soto-Valles, F., Torres-
current agricultural activities. Meanwhile, ADSSs can abandon Sanchez, R., 2017. A software architecture based on FIWARE cloud for precision
useless strategies by judging from failures in past cases. Further- agriculture. Agric. Water Manag. 183, 123–135. https://doi.org/10.1016/j.agwat.
more, by learning from historical data sets, regular patterns can be 2016.10.020.
Bonfante, A., Monaco, E., Manna, P., De Mascellis, R., Basile, A., Buonanno, M.,
drawn and used to predict future circumstances (Ghorbani et al.,
Gantilena, G., Esposito, A., Tedeschi, A., De Michele, C., Belfiore, O., Catapano, I.,
2019). Conclusively, the efficiency of decision-making and quality Ludeno, G., Salinas, K., Brook, A., 2019. LCIS DSS – An irrigation supporting system
of generated decision supports can be significantly improved by for water use efficiency improvement in precision agriculture: A maize case study.
performing analysis on historical information. Agric. Syst. 176. https://doi.org/10.1016/j.agsy.2019.102646.
Taechatanasat, P., Armstrong, L., 2014. Decision support system data for farmer decision
making. In: Proceedings of Asian Federation for Information Technology in
4. Conclusions Agriculture, pp. 472–486.
Jones, J.W., 1980. Decision support systems – An organizational perspective – KEEN, PG,
MORTON. MSS. Adm. Sci. Q. 25 (2), 376–382. https://doi.org/10.2307/2392463.
This paper has presented a comprehensive survey of current agri- Sheng, Y.K., Zhang, S., 2009. Analysis of problems and trends of decision support systems
cultural decision support systems for Agriculture 4.0. Due to the cap- development. In: In: 2009 International Conference on E-Business and Information
ability of processing a large amount of agricultural data and handling System Security, pp. 1216–1218.
14
Yazdani, M., Zarate, P., Coulibaly, A., Zavadskas, E.K., 2017. A group decision making Mehmood, T., Liland, K.H., Snipen, L., Saebo, S., 2012. A review of variable selection
support system in logistics and supply chain management. Expert Syst. Appl. 88, methods in Partial Least Squares Regression. Chemometr. Intell. Lab. Syst. 118,
376–392. https://doi.org/10.1016/j.eswa.2017.07.014. 62–69. https://doi.org/10.1016/j.chemolab.2012.07.010.
Terribile, F., Agrillo, A., Bonfante, A., Buscemi, G., Colandrea, M., D’Antonio, A., De Svalina, I., Galzina, V., Lujic, R., Simunovic, G., 2013. An adaptive network-based fuzzy
Mascellis, R., De Michele, C., Langella, G., Manna, P., Marotta, L., Mileti, F.A., inference system (ANFIS) for the forecasting: The case of close price indices. Expert
Minieri, L., Orefice, N., Valentini, S., Vingiani, S., Basile, A., 2015. A web-based Syst. Appl. 40 (15), 6055–6063. https://doi.org/10.1016/j.eswa.2013.05.029.
spatial decision supporting system for land management and soil conservation. Solid Giusti, E., Marsili-Libelli, S., 2015. A fuzzy decision support system for irrigation and
Earth. 6, 903–928. https://doi.org/10.5194/se-6-903-2015. water conservation in agriculture. Environ. Model. Softw. 63, 73–86. https://doi.org/
Alenljung, B., 2008. Envisioning a future decision support system for requirements en- 10.1016/j.envsoft.2014.09.020.
gineering: a holistic and human-centered perspective. Ph.D. Dissertation. Dept. Giannerinii, G., Genovesi, R., 2011. Irrinet: IT services for farm water management, a
Comput. Info. Sci., Linköping Univ., Linköping, Sweden. large scale implementation in Italy. In: 8th European-Federation-for-Information-
Bazzani, G.M., 2005. An integrated decision support system for irrigation and water Technology-in-Agriculture, Food and the Environment/8th World Congress on
policy design: DSIRR. Environ. Model Softw. 20 (2), 153–163. https://doi.org/10. Computers in Agriculture, pp. 554–559.
1016/j.envsoft.2003.12.017. Oad, Ramchand, Garcia, Luis, Kinzli, Kristoph-Dietrich, Patterson, Dave, Shafike, Nabil,
Tyrychtr, J., Vostrovsky, V., 2017. The current state of the issue of information needs and 2009. Decision support systems for efficient irrigation in the middle Rio Grande
dispositions among small Czech farms. Agr. Econ-Czech. 63 (4), 164–174. https:// Valley. J. Irrig. Drain Eng. 135 (2), 177–185. https://doi.org/10.1061/(ASCE)0733-
doi.org/10.17221/321/2015-AGRICECON. 9437(2009)135:2(177).
Hayman, P., 2004. Decision support systems in Australian dryland farming: A promising Rickards, L., Howden, S.M., 2012. Transformational adaption: agriculture and climate
past, a disappointing present and uncertain future. Science Congress. change. Crop Pasture Sci. 63, 240–250. https://doi.org/10.1071/CP11172.
Nguyen, N.C., Wegener, M.K., Russell, I.W., 2007. Decision support systems in Australian El-Sharkawy, M.A., 2014. Global warming: Causes and impacts on agroecosystems pro-
agriculture: State of the art and future development. AFBM J. 4 (1-2), 1–7. https:// ductivity and food security with emphasis on cassava comparative advantage in the
doi.org/10.22004/ag.econ.122229. tropics/subtropics. Photosynthetica 52 (2), 161–178. https://doi.org/10.1007/
Anderl, R., 2015. Industrie 4.0 – technological approaches, use cases, and implementa- s11099-014-0028-7.
tion. AT Autom. 63 (10), 753–765. https://doi.org/10.1515/auto-2015-0025. Weller, S., Janz, B., Jorg, L., Kraus, D., Racela, H.S.U., Wassmann, R., Butterbach-Bahl, K.,
Yuan, J.J., Lu, Y.L., Ferrier, R.C., Liu, Z.Y., Su, H.Q., Meng, J., Song, S., Jenkins, A., 2018. Kiese, R., 2016. Greenhouse gas emissions and global warming potential of tradi-
Urbanization, rural development and environmental health in China. Environ. Dev. tional and diversified tropical rice rotation systems. Glob. Chang. Biol. 22 (1),
28, 101–110. https://doi.org/10.1016/j.envdev.2018.10.002. 432–448. https://doi.org/10.1111/gcb.13099.
Udias, A., Pastori, M., Dondeynaz, C., Moreno, C.C., Ali, A., Cattaneo, L., Cano, J., 2018. A Schutze, N., Schmitz, G.H., 2010. OCCASION: New planning tool for optimal climate
decision support tool to enhance agricultural growth in the Mekrou river basin (West change adaption strategies in irrigation. J. Irrig. Drain. Eng. 136 (12), 836–846.
Africa). Comput. Electron. Agric. 154, 467–481. https://doi.org/10.1016/j.compag. https://doi.org/10.1061/(ASCE)IR.1943-4774.0000266.
2018.09.037. Semenov, M.A., Brooks, R.J., Barrow, E.M., Richardson, C.W., 1998. Comparison of the
Dong, C., Huang, G., Cheng, G.H., Zhao, S., 2018. Water resources and farmland man- WGEN and LARS-WG stochastic weather generators for diverse climates. Clim. Res.
agement in the Songhua River watershed under interval and fuzzy uncertainties. 10 (2), 95–107. https://doi.org/10.3354/cr010095.
Water Resour. Manag. 32, 4177–4200. https://doi.org/10.1007/s11269-018-2035-0. Mo, X.G., Liu, S.X., Lin, Z.H., Xu, Y.Q., Xiang, Y., McVicar, T.R., 2005. Prediction of crop
Fountas, S., Sorensen, C.G., Tsiropoulos, Z., Cavalaris, C., Liakos, V., Gemtos, T., 2015. yield, water consumption and water use efficiency with a SVAT-model growth model
Farm machinery management information system. Comput. Electron. Agric. 110, using remotely sensed data on the North China Plain. Ecol. Modell. 183 (2), 301–322.
131–138. https://doi.org/10.1016/j.compag.2014.11.011. https://doi.org/10.1016/j.ecolmodel.2004.07.032.
Czimber, K., Galos, B., 2016. A new decision support system to analyse the impacts of Wenkel, K.O., Berg, M., Mirschel, W., Wieland, R., Nendel, C., Kostner, B., 2013.
climate change on the Hungarian forestry and agricultural sectors. Scand. J. For. Res. LandCaRe DSS – An interactive decision support system for climate change impact
31 (7), 664–673. https://doi.org/10.1080/02827581.2016.1212088. assessment and the analysis of potential agricultural land use adaptation strategies. J.
Kmoch, L., Pagella, T., Palm, M., Sinclair, F., 2018. Using local agroecological knowledge Environ. Manag. 127, 168–183. https://doi.org/10.1016/j.jenvman.2013.02.051.
in climate change adaptation: A study of tree-based options in Northern Morocco. Franke, J., Kostner, B., 2007. Effects of recent climate trends on the distribution of po-
Sustainability. 10 (10). https://doi.org/10.3390/su10103719. tential natural vegetation in Central Germany. Int. J. Biometeorol. 52 (2), 139–147.
van Evert, F.K., Fountas, S., Jakovetic, D., Crnojevic, V., Travlos, I., Kempenaar, C., 2017. https://doi.org/10.1007/s00484-007-0096-5.
Big data for weed control and crop protection. Weed Res. 57 (4), 218–233. https:// Kadiyala, M.D.M., Nedumaran, S., Singh, P., Chukka, S., Irshad, M.A., Bantilan, M.C.S.,
doi.org/10.1111/wre.12255. 2015. An integrated crop model and GIS decision support system for assisting agro-
Borodin, V., Bourtembourg, J., Hnaien, F., Labadie, N., 2016. Handling uncertainty in nomic decision making under climate change. Sci. Total Environ. 521, 123–134.
agricultural supply chain management: a state of the art. Eur. J. Oper. Res. 254, https://doi.org/10.1016/j.scitotenv.2015.03.097.
348–359. https://doi.org/10.1016/j.ejor.2016.03.057. Jones, J.W., Hoogenboom, G., Porter, C.H., Boote, K.J., Batchelor, W.D., Hunt, L.A.,
Pourmoayed, R., Nielsen, L.R., Kristensen, A.R., 2016. A hierarchical Markov decision Wilkens, P.W., Singh, U., Gijsman, A.J., Ritchie, J.T., 2003. The DSSAT cropping
process modeling feeding and marketing decisions of growing pigs. Eur. J. Oper. Res. system model. In, 2nd. Cropping Systems.
250 (3), 925–938. https://doi.org/10.1016/j.ejor.2015.09.038. Hamprecht, E., Corsten, D., Noll, M., Meier, E., 2005. Controlling the sustainability of
Recio, B., Rubio, F., Criado, J.A., 2003. A decision support system for farm planning using food supply chains. Supply Chain Manag. 10 (1), 7–10. https://doi.org/10.1108/
AgriSupport II. Decis. Support Syst. 36, 189–203. https://doi.org/10.1016/S0167- 13598540510578315.
9236(02)00134-3. Muller, C., Vermeulen, W.J.V., Glasbergen, P., 2009. Perceptions on the demand side and
Conesa-Munoz, J., Valente, J., del Cerro, J., Barrientos, A., Ribeiro, A., 2016. A multi- realities on the supply side: A study of the South African table grape export industry.
robot sense-act approach to lead to a proper acting in environmental incidents. Sustain. Dev. 17 (5), 295–310. https://doi.org/10.1002/sd.425.
Sensors 16 (8). https://doi.org/10.3390/s16081269. Cordeau, J.F., Pasin, F., Solomon, M.M., 2006. An integrated model for logistic network
Nabaei, A., Hamian, M., Parsaei, M.R., Safdari, R., Samad-Soltani, T., Zarrabi, H., design. Ann. Oper. Res. 144 (1), 59–82. https://doi.org/10.1007/s10479-006-
Ghassemi, A., 2018. Topologies and performance of intelligent algorithms: a com- 0001-3.
prehensive review. Arti. Intell. Rev. 49, 79–103. https://doi.org/10.1007/s10462- Harris, I., Naim, M., Palmer, A., Potter, A., Mumford, C., 2011. Assessing the impact of
016-9517-3. cost optimization based on infrastructure modelling on CO2 emissions. Int. J. Prod.
Bochtis, D.D., Sorensen, C.G., Green, O., 2012. A DSS for planning of soil-sensitive field Econ. 131 (1), 313–321. https://doi.org/10.1016/j.ijpe.2010.03.005.
operations. Decis. Support Syst. 53 (1), 66–75. https://doi.org/10.1016/j.dss.2011. Soysal, M., Bloemhof-Ruwaard, J.M., van der Vorst, J.G.A.J., 2014. Modelling food lo-
12.005. gistics networks with emission considerations: The case of an international beef
Keller, T., Defossez, P., Weisskopf, P., Arvidsson, J., Richard, G., 2007. SoilFlex: A model supply chain. Int. J. Prod. Econ. 152, 57–70. https://doi.org/10.1016/j.ijpe.2013.12.
for prediction of soil stresses and soil compaction due to agricultural field traffic 012.
including a synthesis of analytical approaches. Soil Tillage Res. 93 (2), 391–411. Ji, Y., Li, M., Qu, S.J., 2018. Multi-objective linear programming games and applications
https://doi.org/10.1016/j.still.2006.05.012. in supply chain competition. Future Gener. Comput. Syst. 86, 591–597. https://doi.
Alsalam, B.H.Y., Morton, K., Campbell, D., Gonzalez, F., 2017. Autonomous UAV with org/10.1016/j.future.2018.04.041.
vision based on-board decision making for remote sensing and precision agriculture. Dehghan, S., Kazemi, A., Amjady, N., 2014. Multi-objective robust transmission expan-
IEEE Aerospace Conf. sion planning using information-gap decision theory and augmented e-constraint
Pena, J.M., Torres-Sanchez, J., de Castro, A., Kelly, M., Lopez-Granados, F., 2013. Weed method. IET Gener. Trans. Dist. 8 (5), 828–840. https://doi.org/10.1049/iet-gtd.
mapping in early-season maize fields using object-based analysis of unmanned aerial 2013.0427.
vehicle (UAV) images. PLoS one 8 (10). https://doi.org/10.1371/journal.pone. Ting, S.L., Tse, Y.K., Ho, G.T.S., Chung, S.H., Pang, G., 2013. Mining logistics data to
0077151. assure the quality in a sustainable food supply chain: a case in the red wine industry.
Alarcon, J., Garrido, A., Juana, L., 2016. Modernization of irrigation systems in Spain: Int. J. Prod. Econ. 152, 200–209. https://doi.org/10.1016/j.ijpe.2013.12.010.
Review and analysis for decision making. Int. J. Water Resour. Dev. 32 (3), 442–458. Le, T.D.B., Lo, D., 2015. Beyond support and confidence: Exploring interestingness
https://doi.org/10.1080/07900627.2015.1123142. measures for rule-based specification mining. Reengineering (SANER).
de Wit, M., Crookes, D.J., 2013. Improved decision-making on irrigation farming in arid Li, J.F., Tang, Y., Xiao, X., Gao, Y., 2016. Applied research of an improved Apriori al-
zones using a system dynamics model. S. Afr. J. Sci. 109, 11–12. https://doi.org/10. gorithm in the logistics industry. Automation (ICMIA).
1590/sajs.2013/20130191. Fikar, C., 2018. A decision support system to investigate food losses in e-grocery de-
Navarro-Hellin, H., Martinez-del-Rincon, J., Domingo-Miguel, R., Soto-Valles, F., Torres- liveries. Comput. Ind. Eng. 117, 282–290. https://doi.org/10.1016/j.cie.2018.02.
Sanchez, R., 2016. A decision support system for managing irrigation in agriculture. 014.
Comput. Electron. Agric. 124, 121–131. https://doi.org/10.1016/j.compag.2016.04. Tarantilis, C.D., Kiranoudis, C.T., 2001. A meta-heuristic algorithm for the efficient dis-
003. tribution of perishable foods. J. Food Eng. 50 (1), 1–9. https://doi.org/10.1016/
15
S0260-8774(00)00187-4. 1016/S0378-4290(03)00154-0.
Shirogane, J., Mori, T., Iwata, H., Fukazawa, Y., 2008. Accessibility evaluation for GUI Banger, K., Yuan, M.W., Wang, J.M., Nafziger, E.D., Pittelkow, C.M., 2017. A vision for
software using source programs. Software Engineering (JCKBSE). incorporating environmental effects into Nitrogen management decision support
Chu, H., Yang, Y.J., Li, Q.S., Xu, Y.F., Wei, H.P., 2016. A scalable clinical intelligent tools for U.S. maize production. Front. Plant Sci. 8. https://doi.org/10.3389/fpls.
decision support system. Telematics (ICOST). 2017.01270.
Teixeira, B., Pinto, T., Silva, F., Santos, G., Praca, I., Vale, Z., 2018. Multi-agent decision Lin, Y.Z., Deng, X.Z., Jin, Q., 2013. Economic effects of drought on agriculture in North
support tool to enable interoperability among heterogeneous energy systems. Appl. China. Int. J. Disaster Risk Sci. 4 (2), 59–67. https://doi.org/10.1007/s13753-013-
Sci. 8 (3). https://doi.org/10.3390/app8030328. 0007-9.
Verbeke, W., 2005. Agricultural and the food industry in the information age. 11th Evers, L., Barros, A.I., Monsuur, H., Wagelmans, A., 2014. Online stochastic UAV mission
International Congress of the European-Association-of-Agricultural-economists planning with time windows and time-sensitive targets. Eur. J. Oper. Res. 238 (1),
(EAAE). 348–362. https://doi.org/10.1016/j.ejor.2014.03.014.
Zhou, H.X., Zeng, Z., Lian, L., 2018. Adaptive re-planning of AUVs for environmental Kamali, F.P., Borges, J.A.R., Meuwissen, M.P.M., de Boer, I.J.M., Lansink, A.G.J.M.O.,
sampling missions: a fuzzy decision support system based on multi-objective particle 2017. Sustainability assessment of agricultural systems: the validity of expert opinion
swarm optimization. Int. J. Fuzzy Syst. 20 (2), 650–671. https://doi.org/10.1007/ and robustness of a multi-criteria analysis. Agric. Syst. 157, 118–128. https://doi.
s40815-017-0398-7. org/10.1016/j.agsy.2017.07.013.
Poch, M., Comas, J., Rodriguez-Roda, I., Sanchez-Marre, M., Cortes, U., 2004. Designing Chlingaryan, A., Sukkarieh, S., Whelan, B., 2018. Machine learning approaches for crop
and building real environmental decision support systems. Environ. Modell. Softw. 19 yield prediction and nitrogen status estimation in precision agriculture: a review.
(9), 857–873. https://doi.org/10.1016/j.envsoft.2003.03.007. Comput. Electron. Agric. 151, 61–69. https://doi.org/10.1016/j.compag.2018.05.
Patel, N., Rajput, T.B.S., Dinkar, D.K., Ram, S., Singla, S.K., 2018. DOMIS: a decision 012.
support system for design and cost estimation of micro-irrigation systems. Curr. Sci. Han, E.J., Ines, A.V.M., Baethgen, W.E., 2017. Climate-agriculture-modeling and decision
115 (12), 2240–2248. https://doi.org/10.18520/cs/v115/i12/2240-2248. tool (CAMDT): a software framework for climate risk management in agriculture.
Pham, H.V., Kamei, K., 2012. Hybrid intelligent decision support system model for si- Environ. Model. Softw. 95, 102–114. https://doi.org/10.1016/j.envsoft.2017.06.
mulation of disaster and weather forecasting. 4th Asian Conference on Intelligent 024.
Games and Simulation/4th. Asian Simulation Technology. Chougule, A., Jha, V.K., Mukhopadhyay, D., 2016. AgroKanti: Location-aware decision
Poczeta, K., Kubus, L., Yastrebov, A., Papageorgiou, E.I., 2018. Application of fuzzy support system for forecasting of pests and diseases in grapes. 3rd International
cognitive maps with evolutionary learning algorithm to model decision support Conference on Information System Design and Intelligent Applications.
systems based on real-life and historical data. 9th International Workshop on MacFarlane, R., 1996. Modelling the interaction of economic and socio-behavioral factors
Computational Optimization (WCO). in the prediction of farm adjustment. J. Rural Stud. 12 (4), 365–374. https://doi.org/
Rose, D.C., Sutherland, W.J., Parker, C., Lobley, M., Winter, M., Morris, C., Twining, S., 10.1016/S0743-0167(96)00034-4.
Ffoulkes, C., Amano, T., Dicks, L.V., 2016. Decision support tools for agriculture: Ali, M., Deo, R.C., Downs, N.J., Maraseni, T., 2018. Multi-stage committee based extreme
towards effective design and delivery. Agric. Syst. 149, 165–174. https://doi.org/10. learning machine model incorporating the influence of climate parameters and sea-
1016/j.agsy.2016.09.009. sonality on drought forecasting. Comput. Electron. Agric. 152, 149–165. https://doi.
Francis, C.A., Leiblein, G., Breland, T.A., Salomonsson, L., Geber, U., Sriskandarajah, N., org/10.1016/j.compag.2018.07.013.
Langer, V., 2008. Transdisciplinary research for a sustainable agriculture and food Ghorbani, M.A., Deo, R.C., Karimi, V., Kashani, M.H., Ghorbani, S., 2019. Design and
sector. Agron. J. 100 (3), 771–776. https://doi.org/10.2134/agronj2007.0073. implementation of a hybrid MLP-GSA model with multi-layer perceptron-gravita-
Asseng, S., Jamieson, P.D., Kimball, B., Pinter, P., Sayre, K., Bowden, J.W., Howden, S.M., tional search algorithm for monthly lake water level forecasting. Stoch. Environ. Res.
2004. Simulated wheat growth affected by rising temperature, increased water deficit Risk Assess. 33 (1), 125–147. https://doi.org/10.1007/s00477-018-1630-1.
and elevated atmospheric CO2. Field Corp Res. 85 (2–3), 85–102. https://doi.org/10.
16

Computers and Electronics in Agriculture: Zhaoyu Zhai, José Fernán Martínez, Victoria Beltran, Néstor Lucas Martínez T

Uploaded by

Copyright:

Available Formats

Computers and Electronics in Agriculture: Zhaoyu Zhai, José Fernán Martínez, Victoria Beltran, Néstor Lucas Martínez T

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Computers and Electronics in Agriculture: Zhaoyu Zhai, José Fernán Martínez, Victoria Beltran, Néstor Lucas Martínez T

Uploaded by

Copyright:

Available Formats

Computers and Electronics in Agriculture 170 (2020) 105256

Contents lists available at ScienceDirect

Computers and Electronics in Agriculture

Decision support systems for agriculture 4.0: Survey and challenges T

ARTICLE INFO ABSTRACT

1. Introduction waste of natural resources. In the 20th century, Agriculture 3.0

Fig. 1. A general framework of an ADSS for pest management.

(2018) explored the current state of machine learning techniques in decisions.

Based on the above four requirements, thirteen ADSSs are selected

• 1. Defining a question: What does Agriculture 4.0 require from ADSSs

Fig. 3. A general framework of agricultural decision support systems.

hover over targets and perform precise spraying operations. Con-

2.1.5. Summary of ADSSs for mission planning

Through the review result from Section 2.1.1–2.1.4, the differences

about agricultural operations like chemical treatment and scheduling of

mental and crop-related data as inputs to their decision supporting tools

2.2. ADSSs for water resources management

Current researches of ADSSs for water resources management are

provide farmers with effective decision supports on controlling the

et al., 2016; de Wit and Crookes, 2013). It aims at ensuring wettability

minimum water usages.

to summarize contributions from current works.

2.2.1. Smart irrigation decision support system (SIDSS)

by Navarro-Hellin et al. (2016). Traditionally, irrigation activities are

to SIDSS, the first two components collect sensing data (volumetric

(Mehmood et al., 2012) and Adaptive Neuro Fuzzy Inference Systems

(ANFIS) (Svalina et al., 2013). PLSR is used to deduct unnecessary

dundant. ANFIS is employed to minimizing estimated errors under a

The smart irrigation decision support system was testified and

SIDSS is able to provide farmers with an irrigation report, which is

better than the decisions made by an agronomist. The irrigation report

Fig. 5. The framework of SIDSS.

• Selections of supply points.

number of damaged red wine decreases from 2134 to 530, avoiding

3. Evaluation and upcoming challenges

AgriSupport II system ★★★ ★★★ ★★★ ★★

AgriSupport II system 16/24 (66.67%) OCCASION 20/24 (83.33%)

You might also like