agronomy

Article
AgROS: A Robot Operating System Based Emulation
Tool for Agricultural Robotics
Naoum Tsolakis 1,2, * , Dimitrios Bechtsis 1,3 and Dionysis Bochtis 1
1 Institute for Bio-Economy and Agri-Technology (iBO), Centre for Research and Technology Hellas (CERTH),
6th km Charilaou-Thermi Rd, GR 57001 Thermi, Thessaloniki, Greece
2 Centre for International Manufacturing, Institute for Manufacturing (IfM), Department of Engineering,
School of Technology, University of Cambridge, Cambridge CB3 0FS, UK
3 Department of Industrial Engineering and Management, International Hellenic University (IHU),
GR 57400 Sindos, Thessaloniki, Greece
* Correspondence: [email protected]; Tel.: +30-2310-498210




Received: 2 May 2019; Accepted: 17 July 2019; Published: 20 July 2019


Abstract: This research aims to develop a farm management emulation tool that enables agrifood
producers to effectively introduce advanced digital technologies, like intelligent and autonomous
unmanned ground vehicles (UGVs), in real-world field operations. To that end, we first provide a
critical taxonomy of studies investigating agricultural robotic systems with regard to: (i) the analysis
approach, i.e., simulation, emulation, real-world implementation; (ii) farming operations; and (iii) the
farming type. Our analysis demonstrates that simulation and emulation modelling have been
extensively applied to study advanced agricultural machinery while the majority of the extant
research efforts focuses on harvesting/picking/mowing and fertilizing/spraying activities; most studies
consider a generic agricultural layout. Thereafter, we developed AgROS, an emulation tool based
on the Robot Operating System, which could be used for assessing the efficiency of real-world
robot systems in customized fields. The AgROS allows farmers to select their actual field from a
map layout, import the landscape of the field, add characteristics of the actual agricultural layout
(e.g., trees, static objects), select an agricultural robot from a predefined list of commercial systems,
import the selected UGV into the emulation environment, and test the robot’s performance in a
quasi-real-world environment. AgROS supports farmers in the ex-ante analysis and performance
evaluation of robotized precision farming operations while lays the foundations for realizing “digital
twins” in agriculture.

Keywords: digital agriculture; agricultural robotics; unmanned ground vehicles; robot operating
system; emulation; precision farming

1. Introduction
The range of human- and environment-related risks that are associated with agricultural operations,
along with the elevated demand for food supplies, highlights the need for intensification and
industrialization in agriculture [1]. Agricultural intensification and industrialization are prerequisites
for harnessing agro-ecological and socio-technological processes’ complexity in order to increase
agri-food production [2], while contemporarily reducing the associated ecological footprint [3] (p. 130).
In this regard, technology-enabled agricultural solutions could support farmers in improving
productivity levels while at the same time ensuring safety and sustainability in agricultural
operations [4,5]. In particular, Industry 4.0 in agriculture is considered a financially, environmentally
and technically viable option to tackle agrifood value chain-related challenges by introducing [6,7]:

• connected equipment and farmers’ interaction with legacy technology;

Agronomy 2019, 9, 0403; doi:10.3390/agronomy9070403 www.mdpi.com/journal/agronomy

Agronomy 2019, 9, 0403 2 of 20

• automation in agricultural operations; and

• scientific assessment methods for real-time monitoring of input requirements and farming outputs.

Intelligent autonomous systems and robotics pose a promising option to respond towards
the constantly growing food demand and the increased resources’ stewardship requirements
in the agricultural sector [8]. The digitalization of agricultural processes and services requires
multi-engineering domains and disciplines [9], thus dictating the adoption of technology-based
intelligent solutions which integrate both software and hardware components. In terms of software
constituents, decision support tools can assist farmers in making evidence-based decisions regarding
agricultural operations to improve productivity and prevent environmental degradation [10].
In particular, the use of information and communication technology can support the bottom-up
transformation of agricultural activities, from monitoring and surveying of plants to tilling, seeding,
watering, fertilizing, weeding and harvesting tasks [11]. To a greater extent, modern information
technology-based systems should create cyber-physical interfaces to demonstrate and communicate
the potential benefits of intelligent machinery, such as unmanned ground vehicles (UGVs), to farmers
so as to create facilitating conditions towards the adoption of advanced technology applications
in real-world operations [12]. Notwithstanding the auxiliary role of decision support systems, the
extant literature clearly demonstrates that the majority of such systems are poorly adapted to actual
farmers’ needs; hence, their uptake and use in agriculture is limited [13]. In addition, most information
technology-based agricultural platforms are limited to software applications, thus failing to efficiently
couple cyber applications to physical machinery. Therefore, existing agricultural toolkits inhibit
farmers from apprehending the connections and linkages within and across the physical and cyber
world [14]. However, to effectively transition towards Agriculture 4.0, i.e., agri-activities enabled by
intelligent applications alike Industry 4.0 in manufacturing, it is significant to realize “digital twins”
and understand the implications of advanced agri-technologies at both software and hardware levels.
In particular, applications at a cyber level have to capture multi-physic and multi-scale operational
aspects that mirror the corresponding physical reality [15].
At a software—cyberspace—level, capturing the plethora of parameters pertinent to agricultural
operations, along with the associated sustainability impact thereof, is a complex task requiring
knowledge over local climate, crop varieties and field geomorphological and hydrological
specifications [16]. Extant tools to evaluate the benefits arising from the adoption of intelligent
autonomous systems have limited capabilities. Simulation, a method extensively used to assess a
system’s impact on an operational setting, requires simplified approximations and assumptions to
deliver a realistic idea about an intelligent system’s operational performance and outputs [17]. In this
regard, emulation techniques are devised to imitate, through cyber-physical interfaces, the behaviour
of real-world equivalent systems under alternative pragmatic conditions in real-time [18]. However,
the application of emulation modelling for a regular farmer is challenging, particularly during the
entire life-cycle of a UGV system, due to time, cost, and capability constrains. In addition, integrating
field characteristics and machinery functionality depends on the complexity of the technology-enabled
farming system.
At a hardware—physical space—level, in today’s digitalization era intelligent technological
applications are pertinent to field operations to tackle uncertainty and variability in rural activities
(e.g., soil dynamics and crops’ water requirements), and ultimately enable self-optimized agricultural
supply networks [19]. At a field level, automation in agriculture provides an appealing alternative
to the repeated heavy-duty tasks that are currently being performed by human labour. In this sense,
automation relates to agriculture by means of highly specialized machinery. Despite the progress
in intelligent automated technologies for agricultural environments, further research is needed to
engage farmers into the intelligent systems’ control loop [8]. Digital technologies in the agrifood
sector have only been studied in isolation with a cross-functional analysis. Related research efforts
focus merely on detailed technical aspects of developed robot systems and do not enable farmers to
perform a predictive analysis and ex-ante assessment of agricultural operations efficiency. Therefore,
Agronomy 2019, 9, 0403 3 of 20

farmers cannot become aware of the sustainability benefits that could be derived from the adoption
and implementation of automated systems (e.g., UGVs) and other digital technologies (e.g., sensors) in
farming operations.
Software- or hardware-based digital solutions to promote sustainable intensification in agriculture
should be able to be tested and assessed prior to their actual deployment [2]. Notwithstanding the
intelligent characteristics and technically advanced capabilities of alternative “high-tech” solutions
in agriculture, research should also focus on attributes that enable or hinder the uptake of digital
interventions like farmers’ personality and objectives, education level, skills, learning style and the
scale of the business activities [13].
The aim of this research is to develop an emulation tool that integrates cyber-physical interfaces
to enable farmers make evidence-based decisions about the implementation of alternative intelligent
farming applications via attempting to answer the following research questions:

• Research Question #1: What are the challenges encountered by current decision support
tools for the predictive analysis performance and ex-ante evaluation of digital applications
to farming operations?
• Research Question #2: Which characteristics should be included in a farm management
emulation-based tool to enable farmers effectively introduce advanced robotic technologies
in real-world field operations?

In order to address the abovementioned research questions, we applied a multi-method approach

with the aim to provide coherent research output [20], with both academic and practice implications in
agriculture. In particular, to address Research Question #1 we performed an inclusive literature review
and a critical taxonomy of studies providing evidence about technology-enabled agricultural systems
analysed either from a cyber or from a physical application’s perspective. Thereafter, to tackle Research
Question #2 we developed AgROS, an integrated farm management tool for emulating the life-cycle of
agricultural activities in a field’s physical space. AgROS acronym stems from “Agriculture” (“Ag”)
and “ROS—Robot Operating System” (“ROS”). AgROS is based on the Robot Operating System and
could be used for emulating operations of real-world intelligent machinery in customized settings for
the ex-ante system performance assessment to enable real-world transformations.
The remainder of the paper is structured as follows: Section 2 sets out the materials and methods
that underpin this research. Section 3 provides a critical taxonomy over robotic technology-enabled
agricultural systems based on the elaborated analysis approach (i.e., simulation, emulation, real-world
implementation), farming operations, and unit of farming activities. A discussion over the analysis
approach is performed to demonstrate associated benefits and shortcomings. Thereafter, the developed
emulation tool is presented in Section 4. AgROS could be used for emulating farming operations
of real-world robot systems along with their operational performance in customized fields. Finally,
conclusions, scientific and practical implications, limitations and recommendations for future research
are discussed in Section 5.

2. Materials and Methods

The scope of this research is to identify the scientific research over software simulation tools and
real-world robotic applications on agricultural environments and develop a tool that encompasses the
merits of both simulation or emulation only focused approaches and field only tests. In this sense,
the two methodological steps that were undertaken to conduct this research are: (i) critical taxonomy
of the extant relevant literature; and (ii) programming and development of the AgROS emulation tool.

2.1. Critical Taxonomy

At a first methodological stage, the object of scrutiny is secondary research. Considering the
aim of this research, the critical taxonomy of the reviewed studies focuses on: (i) the analysis
approach elaborated to assess technology-enabled agricultural systems, i.e., real-world application,
Agronomy 2019, 9, 0403 4 of 20

simulation, emulation; (ii) the farming operations for which the discussed technology-enabled
agricultural system is being evaluated, i.e., monitoring/analysing, harvesting/picking/mowing,
fertilizing/spraying, planting/seeding, watering, weeding, tilling; and (iii) the farming type where the
studied technology-enabled agricultural system is applied, i.e., generic agricultural layout, orchard,
arable field and greenhouse.
In order to ensure scientific integrity, the critical taxonomy of extant studies comprises scientific
articles identified in the Scopus® of Elsevier and Web of Science® of Thomson Reuters databases.
The focus on these databases is attributed to the fact that they record a broad range of peer-reviewed
scientific journals, particularly in the areas of Natural Sciences and Engineering [21].
The following Boolean query was formulated: (“simulation” OR “emulation”) AND (“decision
support systems”) AND (“autonomous vehicles”) AND (“agriculture”). The categories “Article Title,
Abstract, Keywords” were used while the timespan was set from “All years” to “Present”. The reviewed
studies selected to be included in our taxonomy were written in the English language. Pertinent
studies were also examined. By 10 February 2019 we identified 41 studies to be retained for review.
An extensive literature review of autonomous vehicles in agriculture extends the scope of this research.
The first relevant published work is detected in 2009. In addition, a remarkably increasing number
of relevant publications (about 56%) is documented from 2017 onwards, hence highlighting a growing
interest in the field of autonomous intelligent vehicles in agriculture. The distribution of the reviewed
publications per analysis approach, along with the allocation of the studies regarding the analysis
approach used per farming type, is depicted in Figure 1. First of all, 34% of the reviewed publications
report the real-world implementation of autonomous vehicles in field operations. Therefore, the focus of
the current body of literature on physical testing of machinery is evident. Furthermore, around 10%–15%
of the studied works use either simulation or emulation modelling approaches. Approximately 14% of
the reviewed works jointly apply simulation and emulation modelling analysis techniques. Notably,
only four studies apply an integrated analysis approach exploring the cyber-physical interface. Secondly,
regarding the farming type, most of the studies included in our critical taxonomy consider either a
generic agricultural layout (13 articles) or an orchard (11 articles). The generic agricultural layout type
provides a great range of analysis opportunities, with simulation and emulation modelling being greatly
applied possibly owing to the uncontrolled environment of generic fields and the dynamic-changing
conditions that stipulate interesting technical and operational challenges for any tested autonomous
vehicles. On the other hand, the more controlled environments in orchards allow the direct testing of
real-world vehicles due to the lower possibility of unexpected incidents that could lead to equipment
impairment and significant associated costs.
Likewise, the distribution of the reviewed articles per farming operation is illustrated in Figure 2.
In particular, Figure 2 depicts the research approach applied to analyse each farming operation;
a reviewed research article might had applied a particular research analysis approach to multiple
farming operations. Figure 2 clearly demonstrates that monitoring/analysing operations have been
extensively studied in the past (N = 33) from a simulation, emulation, and a real-world application
perspectives. Intelligent vehicles can be easily modified and equipped with multiple sensor devices to
monitor different farming operations. Additionally, simulation and emulation modelling, and real-world
application of intelligent vehicles in harvesting/picking/mowing and fertilizing/spraying activities has
attracted the research interest (N = 13 for both cases) as these farming operations are quite common
in most plantations. However, the extant research with regards to planting/seeding (N = 4), watering
(N = 3), weeding (N = 4), and tilling (N = 4) is rather limited possibly due to the advanced technical
requirements for UGVs.
Agronomy 2019, 9, 0403 5 of 20
Agronomy 2019, 9, x FOR PEER REVIEW 5 of 21

Figure 1. Publications per applied research analysis approach (pie chart on the left); and distribution
of the reviewed publications based on the elaborated research analysis approach per farming type
(bar chart on the right).

Likewise, the distribution of the reviewed articles per farming operation is illustrated in Figure
2. In particular, Figure 2 depicts the research approach applied to analyse each farming operation; a
reviewed research article might had applied a particular research analysis approach to multiple
farming operations. Figure 2 clearly demonstrates that monitoring/analysing operations have been
extensively studied in the past (N = 33) from a simulation, emulation, and a real-world application
perspectives. Intelligent vehicles can be easily modified and equipped with multiple sensor devices
to monitor different farming operations. Additionally, simulation and emulation modelling, and real-
world application of intelligent vehicles in harvesting/picking/mowing and fertilizing/spraying
activities has attracted the research interest (N = 13 for both cases) as these farming operations are
quite common in most plantations. However, the extant research with regards to planting/seeding
Figure1.
Figure 1. Publications
Publications per
per applied
applied research
research analysis
analysis approach
approach (pie
(pie chart
chart on
on the
the left);
left); and
anddistribution
distribution
(N = 4), watering (N = 3), weeding (N = 4), and tilling (N = 4) is rather limited possibly due to the
of the reviewed
of thetechnical publications
reviewed requirements based
publications based on the elaborated research analysis approach per farming
on the elaborated research analysis approach per farming type type
advanced for UGVs.
(barchart
(bar charton
onthe
theright).
right).

Likewise, the distribution of the reviewed articles per farming operation is illustrated in Figure
2. In particular, Figure 2 depicts the research approach applied to analyse each farming operation; a
reviewed research article might had applied a particular research analysis approach to multiple
farming operations. Figure 2 clearly demonstrates that monitoring/analysing operations have been
extensively studied in the past (N = 33) from a simulation, emulation, and a real-world application
perspectives. Intelligent vehicles can be easily modified and equipped with multiple sensor devices
to monitor different farming operations. Additionally, simulation and emulation modelling, and real-
world application of intelligent vehicles in harvesting/picking/mowing and fertilizing/spraying
activities has attracted the research interest (N = 13 for both cases) as these farming operations are
quite common in most plantations. However, the extant research with regards to planting/seeding
(N = 4), watering (N = 3), weeding (N = 4), and tilling (N = 4) is rather limited possibly due to the
advanced technical requirements for UGVs.

Figure 2. Distribution of the reviewed publications based on the elaborated research analysis approach
Figure 2. Distribution
per farming operation. of the reviewed publications based on the elaborated research analysis
approach per farming operation.
2.2. AgROS Development
This research proposes an emulation tool, named AgROS, for the ex-ante assessment of precision
farming activities in near to real-world virtual environments. The tool is developed based on the ROS
platform; ROS, an open source framework created by Willow Garage and Stanford Artificial Intelligence
Laboratory, is considered as a de facto analysis approach and an appropriate tool for analysing “digital
twins” of autonomous robotic systems in the manufacturing industry. The advantages of ROS refer
to the incorporation of multiple compatible libraries combined with third-party hardware support
that contribute to a thorough exploration of a robot system’s functional potential, further providing
opportunities for the application and improvement of vehicle routing and optimization algorithms [22].
AgROS is a wrapper that integrates: (i) the ROS platform; (ii) the Gazebo 3D scenery environment;
(iii) the Open Street Maps tool; and (iv) several 3D models for objects’ representation. Furthermore,
Figure 2. Distribution of the reviewed publications based on the elaborated research analysis
approach per farming operation.
Agronomy 2019, 9, 0403 6 of 20

by introducing a business logic layer AgROS enables the autonomous movement of an intelligent vehicle
in a customizable cyber environment. The AgROS emulation tool enables a range of functionalities
and capabilities to be modelled, including: static agricultural environment field layout/landscape;
dynamic agricultural environment (e.g., cultivation, other on-field activities); intelligent vehicle’s
type (both commercially available and bespoke); vehicle’s equipment (i.e., sensors, actuators);
autonomous environment exploration (i.e., simultaneous localization and mapping—SLAM—planning
and scheduling, navigation); and data processing activities (i.e., image processing, deep learning).

3. Robotic Technology-Enabled Agricultural Systems: Background

In order to assess the benefits emanating from the implementation of robotic technologies to the
agricultural sector, extant research approaches use software simulation modelling, emulation modelling
and/or real-world hardware implementation. Notwithstanding the significant advancements in the
analysis approaches that could benefit farmers and provide academic and practical context, extant
studies mainly focus on the examination of technology solutions per se. In this regard, Table 1 provides
a critical taxonomy of the reviewed studies across three axes, namely: (i) the elaborated analysis
approach; (ii) the investigated farming operation; and (iii) the considered farming type. In the following
sub-sections, we only focus on the elaborated analysis approach as the level of analysis feasible at the
cyber-physical interface is essential for demonstrating expected benefits and thus motivating farmers
to use agricultural robotic systems at customized fields.

Table 1. Critical taxonomy of research studies assessing agricultural robotic systems.

Analysis Approach 1 Farming Operation 2 Farming Type 3

Reference
1 2 3 4 5 6 7 8 9 10 11 12 13 14
[23] ν ν ν ν
[24] ν
[25] ν
[26] ν ν ν
[27] ν ν
[28] ν ν ν ν ν
[29] ν ν ν ν ν
[30] ν ν ν
[31] ν ν ν ν
[32] ν ν
[33] ν ν
[34] ν ν ν
[35] ν ν ν ν ν ν ν ν ν ν
[36] ν ν ν ν ν ν ν ν ν ν
[37] ν ν ν ν
[38] ν ν ν ν
[39] ν ν ν ν
[40] ν ν ν ν ν
[41] ν ν ν ν
[42] ν ν ν ν ν ν ν ν ν ν ν
[43] ν ν ν ν
[44] ν ν ν ν ν
[45] ν ν ν ν ν
[46] ν ν ν ν ν ν ν
[47] ν ν ν ν ν ν ν
[48] ν ν ν ν ν
[49] ν ν ν ν ν ν
[50] ν
[51] ν ν ν ν ν ν ν
[52] ν ν ν ν
[53] ν ν ν
Agronomy 2019, 9, 0403 7 of 20

Table 1. Cont.

Analysis Approach 1 Farming Operation 2 Farming Type 3

Reference
1 2 3 4 5 6 7 8 9 10 11 12 13 14
[54] ν ν ν
[55] ν ν ν
[56] ν ν ν ν
[57] ν ν ν ν ν
[58] ν ν ν ν
[59] ν ν ν ν
[60] ν ν ν ν
[61] ν ν ν
[62] ν ν ν
[63] ν
1 Analysis Approach Coding Reference: 1. Real-world Implementation; 2. Simulation Modelling; 3. Emulation
Modelling. 2 Farming Operation Coding Reference: 4. Monitoring/Analysing; 5. Harvesting/Picking/Mowing;
6. Fertilizing/Spraying; 7. Planting/Seeding; 8. Watering; 9. Weeding; 10. Tilling. 3 Farming Type Coding Reference:
11. Generic Agricultural Layout; 12. Orchard; 13. Arable; 14. Greenhouse.

3.1. Cyber Space: Simulation and Emulation Modelling

The use of simulation and emulation modelling methods can provide valuable feedback for the
study of real-world agricultural environments. The critical taxonomy revealed that the majority of
scientific studies does not consider real-world systems’ implementation basically due to the capital
cost associated to UGVs, both at the prototype and commercial outdoor vehicles’ level. Furthermore,
the use of simulation and emulation modelling approaches minimizes potential incidents that could
impair the robotic agricultural equipment while multiple experiments and scenarios could be assessed.
Emmi et al. [35] presented a simulation environment with 2D and 3D support for examining the
behaviour of a fleet of agricultural robots. The authors implemented their simulation model based on the
Webots platform while the graphical user interface was composed at the Matlab environment. Habibie
et al. [38] used the ROS simulation tool for navigating a UGV in a 3D orchard field implemented in
Gazebo for mapping fruits in an apple farm. The authors concluded that the experimental environment
provides a good accuracy for precision farming activities. Hameed et al. [41] used 2D and 3D simulation
for implementing path planning algorithms for side-to-side field coverage in agriculture. Their work
revealed that the uncovered area could be minimized in case an appropriate driving angle is chosen
during simulation tests. Wang et al. [62] used the Microsoft Robotics Developer Studio and Matlab for
optimizing path planning activities in 3D agricultural fields.
Except for using dedicated robotic platforms, several studies elaborated path planning software
tools, integrated into simulation and emulation modelling efforts, that could provide guidance to
real-world vehicles [39,40,48]. Finally, several studies utilized 2D simulation methods for path planning
of autonomous vehicles in field operations [24,25,29,50,56,63].

3.2. Physical Space: Real-World Implementation

Outdoor agricultural fields provide a challenging environment for autonomous vehicles due to the
demanding and dynamically changing field conditions (e.g., soil moisture) and agricultural landscape
(e.g., random obstacles). Greenhouses are characterized as structured or semi-structured environments
that provide more stable operating conditions for agricultural robotics. Agricultural vehicles use a
plethora of sensors (e.g., GPS—Global Positioning System; RTK-GNSS—Real-Time Kinematic-Global
Navigation Satellite Systems; LiDAR—Light Detection And Ranging; camera and depth camera;
ultrasonic sensors) and actuators (e.g., robotic arm, agricultural equipment) for safely navigating in
the field, gathering data and information, and performing farming activities. According to their size,
two main categories of vehicles for real-world experimentation are identified: (i) large scale vehicles
Agronomy 2019, 9, 0403 8 of 20

(e.g., custom tractors, four-wheel-driving four-wheel-steering electric vehicles); and (ii) small- and
medium-scale vehicles (e.g., differential steering four-wheel-driving small robots).
Bayar et al. [26] used a UGV navigating in fruit orchards based on data retrieved from a
laser scanner, and wheel and steering encoders. The optimal navigation minimized the total costs,
thus demonstrating the potential for the use of UGVs in agriculture. De Preter et al. [31] described the
development of a harvesting robot tested in a greenhouse for strawberries. The investigated harvesting
robot prototype consisted of an electric vehicle and a robotic arm. Blok et al. [27] tested the Husky
robot using the ROS platform and concluded that the Particle Filter algorithm was more suitable for
ensuring navigation accuracy in most trial applications. De Sousa et al. [32] proposed fuzzy behaviours
for navigating the mini-robot Khepera in an agricultural environment. Demim et al. [33] implemented
a navigation algorithm using odometer and laser data from a Pioneer3-AT UGV. Hansen et al. [43] used
real-world vehicles (i.e., the RobuROC4 platform from Robusoft and a UAS Maxi Joker 3 RC-helicopter)
for weed mapping and targeted weed control. Radcliffe et al. [54] implemented machine vision
algorithms for orchard navigation using a custom UGV equipped with multiple sensors. Field tests
were conducted in a peach orchard and the results demonstrated the potential to guide utility vehicles
in orchards. Reina et al. [55] examined terrain assessment for precision agriculture using the Husky
A200 UGV. The authors conducted experiments in a farm for validating the proposed terrain classifier
and identified terrain classes. Wang et al. [61] tested the robustness of deep learning algorithms in
mango orchard trees using real-world ground vehicles equipped with 2D and depth cameras. Durmus
et al. [34] implemented a custom real-world vehicle and tested it in a greenhouse environment using the
ROS platform. The mobile robot was using a depth camera for mapping the greenhouse environment
and was navigating in the greenhouse by using the generated map. Christiansen et al. [30] created a
point cloud model using a LiDAR sensor (Velodyne) in an unmanned aerial vehicle (DJI Matrice 100)
for mapping an agricultural field in the ROS platform. The aerial vehicle used an RTK-GNSS technique
for navigating in the field. Pierzchała et al. [53] implemented a map of a forest environment using ROS
and a real-world custom ground vehicle. The authors tested the use of a LiDAR sensor for creating a
point cloud of the ground surface and accurately identified the exact position of the trees in the forest.
Large-scale agricultural vehicles, like tractors, have been extensively used in field tests.
For example, Hansen et al. [44] investigated the use of LiDAR in localization techniques for tractor
navigation in a semi-structured orchard environment. Pérez-Ruiz et al. [52] implemented an intelligent
sprayer boom on a tractor for pest control applications. Workshop and field experiments provided
evidence that UGVs can be reliably used in performing a range of precision farming activities.

3.3. Cyber–Physical Interface: Joint Simulation, Emulation and Real-World Implementations

On the one hand, simulation and emulation modelling approaches provide an inexpensive option
for exploring the proof-of-concept of an algorithm or a robotic vehicle. On the other hand, real-world
applications and experimentations with testbeds are necessary for validating a technology-based
solution and demonstrating pragmatic results. Typical emulation tools enable the software code
transfer to real-world prototypes for validating an algorithm’s efficiency during field tests. Shamshiri
et al. [57] proposed the use of open source emulation platforms for performing tedious agricultural
tasks to foster the adoption of agricultural robots. The authors compared the V-REP, Gazebo and
ARGoS emulation tools while referring to a plethora of other similar platforms (e.g., Webots, CARMEN
RNT, CLARAty, Microsoft Robotics Developer Studio, Orca, Orocos, Player, AMOR, Mobotware).
All of the emulation platforms support ROS and C++ programming and the developed programming
code could be used in real-world vehicles. Mancini et al. [51] used aerial and ground sensor-retrieved
data for improving variable rate treatments. The authors used the ROS platform in order to estimate
the volume and state of plants to identify field management zones. Jensen et al. [46] focused on
tractor guidance systems and used ROS to develop the FroboMind software platform and evaluate the
tractor’s performance in precision agriculture tasks. FroboMind was tested in a real-world tractor with
the methodology demonstrating significant reductions in development time and resources required to
Agronomy 2019, 9, 0403 9 of 20

perform field experiments. In addition, Harik and Korsaeth [45] used ROS and a mobile ground vehicle
equipped with sensors in a greenhouse environment. The authors used emulation modelling and
presented experimental results regarding the robot’s autonomous navigation. Han et al. [42] explored
the use of a tractor in a paddy field for auto-guidance. Their emulation environment was programmed
in C++ while the test location was obtained through Google Maps. The emulation environment
was validated by using a real-world tractor platform. Furthermore, in the GRAPE project a fully
equipped (i.e., GPS, IMU, LiDAR) mobile robot was used in vineyards [23]. Emulation experiments
were conducted using the ROS platform prior to the use of a Husky robot on the field. Vasudevan et
al. [60] used the data gathered via real-world robots (both unmanned aerial and ground vehicles) in
order to create a simulation environment and capture specific indices for precision farming activities.
Gan and Lee [37] applied a Husky UGV in a real-world environment after evaluating its performance
and obstacle avoidance algorithms in the Gazebo emulation environment.
Furthermore, a number of studies jointly explored the use of simulation methods in 2D environments
along with real-world implementations on the field [28,36,47,49,58,59]. Typically, in such studies:
the farming plans and the vehicles’ trajectories are firstly being implemented at the 2D environment;
the real-world vehicle uses the algorithms for navigating in the actual field while simultaneously
providing real-time feedback to the algorithm based on the data gathered through sensors.

4. AgROS: An Emulation Tool for Agriculture

The literature review in this research revealed that current analysis tools for robotic systems in
agriculture are not yet at a mature development phase. A plethora of simulation and emulation tools
exists that use: (i) virtual 2D environments for testing path planning algorithms; (ii) 2D images of
the actual field (captured via Google Earth, Open Street Maps, and satellites); and (iii) 3D virtual
agricultural sceneries. Existing approaches can be useful for testing purposes but fail to capture the
dynamic nature of intelligent systems’ operations in real-world fields. On the other hand, the use of
real-world vehicles can bridge this gap by recreating real-world 3D sceneries captured through the
vehicle’s sensors (LiDARs and depth cameras); however; this approach is capital and time intensive for
actual field tests. In this regard, we propose the combined use of simulation and emulation modelling
along with actual field tests.
The proposed AgROS emulation tool takes a step beyond the simple combination of the
above-mentioned approaches and uses the actual landscape of a field from a digital elevation model
(DEM) file retrieved from Open Street Maps. This practice enables a near to real-world emulation
approach as the DEM file includes information regarding the altitude of a field for better simulating a
vehicle’s navigation in an actual setting. Furthermore, farmers can insert static objects and the utilized
UGV in the 3D scenery from a library of models via an automated procedure. The emulation starts after
selecting the appropriate algorithms. Finally, the programming code can be loaded to the real-world
UGV equivalent for field tests.
At this perpetual beta stage of AgROS development, capturing soil sciences related parameters
(e.g., wet/dry soil conditions, sand/clay composition) and developing respective modules extends the
research focus and the tool’s portfolio of functionalities. In addition, AgROS intends to support the
operational analysis of UGVs in customized fields; therefore, informing about diversified issues like
treatment options for pests, market potential of specific crops or the financial viability of advanced
digital machinery in agriculture, is beyond the scope of this research.

4.1. System Architecture

ROS architecture is based on both the publisher-subscriber and the server-client model.
The publisher-subscriber model is implemented using nodes and topics. A node is a distinct active part,
which gathers, computes and transmits data through channels typically called topics. The server-client
model is also based on nodes in order to manage the transmitted data. The communication between
the two node-services is then implemented, as a service analogous to a web service. ROS uses the
Agronomy 2019, 9, 0403 10 of 20

publisher-subscriber model for asynchronous communication and the server-client model for the
synchronous communication between nodes. In order to create an emulated environment in a
user-friendly manner, the AgROS tool is a wrapper that uses:
• the ROS platform;
• the Gazebo 3D scenery environment;
• the Open Street Maps tool;
• several 3D models to represent objects on the agricultural scenery; and
• the business logic layer that enables a vehicle’s autonomous movement in the virtual world.

4.2. Workflow
AgROS is a decision support tool that enables the ex-ante evaluation of robotic technologies in a
3D emulation environment with the digital equivalent to a real-world agricultural environment
being captured through a geographic information system. In particular, AgROS can emulate
the actual field landscape enriched with real-world objects like: (i) trees; (ii) bushes; (iii) grass;
(iv) static obstacles (e.g., rocks, sheds); and (v) dynamic objects that include other robot models that
simultaneously perform agricultural activities. Furthermore, AgROS provides high customization
possibilities by allowing the user to capture an infinite number of field operational conditions like
geomorphological terrain characteristics, physical conditions and lighting levels to further assess
the performance of sensor devices. Overall, AgROS enables the emulation of a realistic agricultural
environment in a defined geographical location for modelling and assessing day-to-day farming
operations with high accuracy. Agricultural robots could then be simulated to autonomously navigate
within the emulated environment based on programmable deterministic or stochastic algorithms to
perform precision farming activities. The stepwise workflow enabled by AgROS, divided in three
subprocess—landscape creation; model library; routing algorithms—to emulate the environment and
operations in a user-friendly manner, is depicted in Figure 3.
Agronomy 2019, 9, x FOR PEER REVIEW 11 of 21

Figure 3. AgROS
Figure3. workflow.
AgROS workflow.

4.3. Graphical User Interface

The graphical user interface in AgROS is developed in Qt, an open source widget toolkit that
allows the development of graphical user interfaces on various software platforms. The native
language of Qt enables shell script manipulation and uses native C++-based ROS programming. The
Agronomy 2019, 9, 0403 11 of 20

4.3. Graphical User Interface

The graphical user interface in AgROS is developed in Qt, an open source widget toolkit that
allows the development of graphical user interfaces on various software platforms. The native language
of Qt enables shell script manipulation and uses native C++-based ROS programming. The developed
AgROS farm management emulation-based tool is able to execute ROS commands directly from
the graphical user interface to gradually represent all the components of a customized agricultural
environment. In order to develop a robust emulation model of agricultural activities the user has to
follow a stepwise procedure, namely:

• Step 1 – create a ROS compatible executable file (or edit an existing one);
• Step 2 – create a 2D or 3D simulation environment using QGIS, an open source geographic
information system tool (field landscape);
• Step 3 – enrich the field by virtually planting trees and placing UGVs;
• Step 4 – explore the field in Gazebo’s simulation environment using routing and object recognition
algorithms; and
• Step 5 – execute the final simulation based on the ROS backend.

The graphical user interface of AgROS, illustrated in Figure 4, allows: (i) creation of a new model
by typing a preferred AgROS filename; (ii) selection between a 2D or a 3D simulation map; (iii) import
of 3D models of agricultural entities (e.g., trees), obstacles and UGVs for enriching the emulated
environment;
Agronomy 2019, 9,and (iv)
x FOR execution
PEER REVIEW of the emulation model. 12 of 21

Figure 4. AgROS graphical user interface.

Figure 4. AgROS graphical user interface.

AgROS enables the emulation of agricultural activities in either a simple landscape (flat layout)
AgROS enables the emulation of agricultural activities in either a simple landscape (flat layout)
or in a 3D (real-world layout) landscape with geomorphological characteristics. The user, optionally,
or in a 3D (real-world layout) landscape with geomorphological characteristics. The user, optionally,
can insert models (e.g., plant trees and dynamic objects like vehicles) and finally import the simulated
can insert models (e.g., plant trees and dynamic objects like vehicles) and finally import the simulated
environment to the Gazebo software for performing emulations. The graphical user interface of the
environment to the Gazebo software for performing emulations. The graphical user interface of the
AgROS enables the user to: (i) insert a flat landscape (Figure 5a); (ii) insert a 3D landscape (Figure 5b);
AgROS enables the user to: (i) insert a flat landscape (Figure 5a); (ii) insert a 3D landscape (Figure
(iii) import agricultural models (e.g., trees) (Figure 5c); and (d) emulate the integrated field environment
5b); (iii) import agricultural models (e.g., trees) (Figure 5c); and (d) emulate the integrated field
(Figure 5d). In particular, the AgROS graphical user interface provides the “Orchard Tool” option that
environment (Figure 5d). In particular, the AgROS graphical user interface provides the “Orchard
allows a user to develop a customized field by importing different emulated models of orchard trees.
Tool” option that allows a user to develop a customized field by importing different emulated models
of orchard trees. In this regard, AgROS allows users to: (i) select the “Orchard Tool”; (ii) define the
position of each tree; (iii) use computer input devices to import trees or use the embedded automatic
plantation mode enabled through specifying orchard row width, row number, trees’ spacing and
initial point at the map; and (iv) save the final world file. As a last step, the user selects to “Execute
or in a 3D (real-world layout) landscape with geomorphological characteristics. The user, optionally,
can insert models (e.g., plant trees and dynamic objects like vehicles) and finally import the simulated
environment to the Gazebo software for performing emulations. The graphical user interface of the
AgROS enables the user to: (i) insert a flat landscape (Figure 5a); (ii) insert a 3D landscape (Figure
5b); (iii) 2019,
Agronomy import agricultural models (e.g., trees) (Figure 5c); and (d) emulate the integrated12field
9, 0403 of 20
environment (Figure 5d). In particular, the AgROS graphical user interface provides the “Orchard
Tool” option that allows a user to develop a customized field by importing different emulated models
In this regard, AgROS allows users to: (i) select the “Orchard Tool”; (ii) define the position of each
of orchard trees. In this regard, AgROS allows users to: (i) select the “Orchard Tool”; (ii) define the
tree; (iii) use computer input devices to import trees or use the embedded automatic plantation mode
position of each tree; (iii) use computer input devices to import trees or use the embedded automatic
enabled through specifying orchard row width, row number, trees’ spacing and initial point at the
plantation mode enabled through specifying orchard row width, row number, trees’ spacing and
map; and (iv) save the final world file. As a last step, the user selects to “Execute AgROS environment”
initial point at the map; and (iv) save the final world file. As a last step, the user selects to “Execute
to run the emulation.
AgROS environment” to run the emulation.

Figure 5. Functionalities of the AgROS graphical user interface: (a) insert a flat landscape; (b) insert a 3D
landscape; (c) import agricultural models (e.g., trees); and (d) emulate the integrated field environment.

4.3.1. Field Layout Environment

The realistic 3D landscape is directly selected from QGIS. QGIS is able to create, edit and visualize
geospatial information on several operating systems and can be used for reading and editing Open
Street Maps data that include information regarding GPS data, aerial photography or even local
knowledge. Another useful format is DEM, a raster file that allows the representation of the actual
terrain including hills hade, slope and aspect information in the QGIS tool. Figure 6 illustrates the
process of creating a 3D landscape in AgROS through using the tool’s graphical user interface and the
QGIS tool, up to the point of importing the .png format grayscale depth file to the Gazebo software.
Following the creation of the 3D landscape of a targeted field, the user can import emulation
models of agricultural objects (e.g., trees) at selected positions. The technical challenge is to plant a tree
model at the proper z scale according to the 3D landscape. The AgROS tool automatically corrects the
planting position along the z-axis by using the geometry of the 3D landscape (Figure 7). The correction
process starts from processing the values of the .png format grayscale depth file to explore the z-values
considering that the grayscale value of each pixel at the .png format file corresponds to the z-value of
the pixels. After identifying the highest (h1, max z-value) and the lowest (h2, min z-value) z-values, the
total height of the landscape can be calculated (h, model height) along with the field’s typical slope (λ).
The field slope is empirically recovered using the field’s dimensions in the Gazebo model. This process
enables the calculation of the z-value of a given pixel, from the grayscale value of the corresponding
pixel at the depth file, and is used for the plantation of tree models to the final planting position (x,y,z).
editing Open Street Maps data that include information regarding GPS data, aerial photography or
even local knowledge. Another useful format is DEM, a raster file that allows the representation of
the actual terrain including hills hade, slope and aspect information in the QGIS tool. Figure 6
illustrates the process of creating a 3D landscape in AgROS through using the tool’s graphical user
interface2019,
Agronomy and9,the
0403QGIS tool, up to the point of importing the .png format grayscale depth file 13
toofthe
20
Gazebo software.

Figure 6. Field
Field layout
layout environment
environment in AgROS comprises of: (a) Standard
Standard Open Street Maps; (b) digital
Agronomy 2019, 9,
elevation
elevation x FORfor
model
model PEER
forthe REVIEW
the region;
region; (c) (c) exact
exact geographical
geographical region;
region; and
and (d) 3D(d) 3D landscape
landscape to be
to be used used 14
in of 21
in Gazebo.
Gazebo.

Following the creation of the 3D landscape of a targeted field, the user can import emulation
models of agricultural objects (e.g., trees) at selected positions. The technical challenge is to plant a
tree model at the proper z scale according to the 3D landscape. The AgROS tool automatically corrects
the planting position along the z-axis by using the geometry of the 3D landscape (Figure 7). The
correction process starts from processing the values of the .png format grayscale depth file to explore
A (m ax)
the z-values considering that the grayscale value of each pixel at the .png format filee Vcorresponds
G rayscal alue:m ax to
z
G azebo
the z-value of the pixels. After identifying the highest (h1, max z-value) and the lowest (h2, min z- M odel H eight:h1
h1
value) z-values, the total height of the landscape can be calculated (h, model height) along with the
field’s typical slope ε: z=λz'
(λ). The field slope is empirically recovered using +β field’s dimensions in the
the
B(m in) h=h1–h2
G rayscale V alue:m in
Gazebo model. This
G azebo M odelH eight:h2
process enables the calculation of the z-value of a given pixel, from the grayscale
value of the corresponding pixel at the depth file, and is used for the plantation of tree models to the
final planting position (x,y,z).
x

Figure 7. The technique used by AgROS to automatically adjust the planting position at the z-axis.
Figure 7. The technique used by AgROS to automatically adjust the planting position at the z-axis.
4.3.2. Static and Dynamic Algorithms
4.3.2. Static and Dynamic Algorithms
Among the main objectives of the developed AgROS tool is the implementation of a path planning
Among the
and navigation main objectives
package of algorithmsof that
the could
developed AgROS
be used for thetool is the implementation
autonomous of a path
navigation of agricultural
planninginand
vehicles navigation
fields. The pathpackage
planningof package
algorithmswasthat could beatused
developed for the
the C++ autonomouslanguage
programming navigation of
as an
agricultural vehicles in fields. The path planning package was developed at the
independent module. Alternative static and dynamic path planning algorithms could be implementedC++ programming
language
and testedasusing
an independent
the AgROS module.
tool. Alternative static and dynamic path planning algorithms could
be implemented anda tested
At a first step, using the
static routing AgROS tool.
algorithm was implemented that uses feedback from the user in
order to capture the number of tree rows, thewas
At a first step, a static routing algorithm implemented
starting and endingthatposition
uses feedback fromrow
of the first theand
userthe
in
order to capture the number of tree rows, the starting and ending position of the first
distance among the rows. This static solution could be used either in orchard fields or even in arable row and the
distance
fields. among
The outputtheofrows. This static
the algorithm is solution could be
the navigation of used either in orchard
the autonomous fields
vehicle acrossor all
eventheinrows
arable
of
fields. The output of the algorithm
the orchard as demonstrated in Figure 8. is the navigation of the autonomous vehicle across all the rows of
the orchard as demonstrated in Figure 8.
be implemented and tested using the AgROS tool.
At a first step, a static routing algorithm was implemented that uses feedback from the user in
order to capture the number of tree rows, the starting and ending position of the first row and the
distance among the rows. This static solution could be used either in orchard fields or even in arable
fields.
AgronomyThe output
2019, 9, 0403 of the algorithm is the navigation of the autonomous vehicle across all the rows
14 of of
20
the orchard as demonstrated in Figure 8.

Figure 8. Deterministic area coverage implemented in AgROS.

Figure 8. Deterministic area coverage implemented in AgROS.
A dynamic routing algorithm is also incorporated that uses the 3D landscape as input from the
AgROSA dynamic
emulation routing algorithm
tool to navigateisthealsoUGV
incorporated
across thethat uses the
orchard 3Dand
field landscape
identifyasininput from the
real-time the
AgROS emulation tool to navigate the UGV across the orchard field and identify
surrounding environment with the LiDAR. The unique functionality of AgROS is that the entry position in real-time the
surrounding
of the vehicleenvironment
is random and with
thethe LiDAR.UGV
emulated The has
unique functionality
no prior of AgROS
information is that
about the thelayout.
field’s entry
Agronomy
position 2019,
of 9, xvehicle
the FOR PEERis REVIEW
random and the emulated UGV has no prior information about the 15 of 21
field’s
A depth camera is used to identify trees and map the field. These processes run in parallel in order to
layout. A depth camera
simultaneously identifyisthe
used to identify
exact location trees andatmap the field. These processes run in parallel in
order to simultaneously identify the exact of trees
location ofthe orchard
trees at theand map
orchard the
andfield.
map Thethe geospatial
field. The
coordinatescoordinates
geospatial of the orchard trees
of the and their
orchard treeslocalization on the mapon
and their localization are recorded
the map areinrecorded
order to inbeorder
used to
in
third
be party
used services
in third party(Figure 9). (Figure 9).
services

Figure 9. AgROS’s geospatial mapping process: (a) real-time localization of trees; and (b) recording of
Figure 9. AgROS’s geospatial mapping process: (a) real-time localization of trees; and (b) recording
every tree on the field.
of every tree on the field.

Finally, a dynamic area coverage plan is calculated through the implementation of the generated
Finally, a dynamic area coverage plan is calculated through the implementation of the generated
map and the corresponding recorded coordinates. The routing algorithm explores the rows of the
map and the corresponding recorded coordinates. The routing algorithm explores the rows of the
orchard field in order to cover the entire free area of the field by traversing the inter-tree rows (Figure 10).
orchard field in order to cover the entire free area of the field by traversing the inter-tree rows (Figure
10).
 of every tree on the field.

Finally, a dynamic area coverage plan is calculated through the implementation of the generated
map and the corresponding recorded coordinates. The routing algorithm explores the rows of the
orchard field
Agronomy in 9,order
2019, 0403 to cover the entire free area of the field by traversing the inter-tree rows (Figure
15 of 20
10).

Figure 10. Dynamic area coverage by the UGV in AgROS.

Figure 10. Dynamic area coverage by the UGV in AgROS.
4.4. Real-World Implementation
4.4. Real-World Implementation
In the AgROS emulation environment the model of an experimental four-driving-wheels
Agronomy 2019, 9, x FOR PEER REVIEW 16 of 21
In the AgROS emulation
four-steering-wheels environmentfield
electric-powered the model of anthe
robot and experimental
source codefour-driving-wheels four-
were tested on a specified
steering-wheels
orchard. electric-powered
The emulation field
modelthe robot and
enables thetransfer
source code were tested
of theon a specified orchard.
The emulation model enables directthe direct
transfer and
and testing of testing
the developed developed modules
modules with with
the robot
the robot in real-world agricultural fields. Figure 11 presents the robot that was used for testing
in real-world agricultural fields. Figure 11 presents the robot that was used for testing the developed the
developed
modules onmodules
the field.on the field.

Figure 11. Actual robot used for testing the AgROS emulation model on an actual orchard.
Figure 11. Actual robot used for testing the AgROS emulation model on an actual orchard.
5. Conclusions
5. Conclusions
The future of agriculture heavily lays on technology and this is also supported by labour shortages,
The future
especially of agricultural
at harsh agriculture operations.
heavily laysAdvanced
on technology and this technologies
and intelligent is also supported bybecome
have to labour
shortages, especially at harsh agricultural operations. Advanced and intelligent technologies have to
become part of a farmer’s toolbox and support every-day farming operations, potentially in a 24/7
basis. Agricultural environments are usually complex and unstructured, while they are characterized
by challenging landscapes under various weather conditions. The direct introduction and application
of autonomous vehicles in physical fields could be challenging in case such systems are not
Agronomy 2019, 9, 0403 16 of 20

part of a farmer’s toolbox and support every-day farming operations, potentially in a 24/7 basis.
Agricultural environments are usually complex and unstructured, while they are characterized by
challenging landscapes under various weather conditions. The direct introduction and application of
autonomous vehicles in physical fields could be challenging in case such systems are not adequately
tested at a cyber level with regard to undertaking heavy-duty farming activities. The proposed AgROS
tool is targeted to address these challenges by providing a platform for emulating the functional
capabilities and performance of real-world intelligent machinery in physical fields. The ex-ante system
performance assessment could foster the upscaled deployment of robotized operations.
In order to support farmers in the transition towards the digital era in agriculture, this research
grounded two research questions and applied alternative methodologies in an attempt to effectively
address them. In particular, to tackle Research Question #1, we discussed farmers’ characteristics
that hinder the adoption of intelligent systems in farming operations. In addition, we referred to the
myopic focus of existing software tools on technical and mechanical aspects of agricultural machinery.
Thereafter, we provided a critical taxonomy of extant studies along three axes, namely: (i) analysis
approach; (ii) farming operation; and (iii) farming type. The taxonomy illustrates the limited scope
of existing research efforts; studies consider the application of software tools on generic field layouts
and tend to neglect the customized needs of farmers. Moreover, in response to Research Question #2,
we devised and analytically described the subprocess that underpins AgROS—landscape creation;
model library; routing algorithms—along with the functionalities of a recommended graphical user
interface. The aforementioned processes are necessary to emulate an agricultural environment and
enable farmers to effectively introduce advanced robotic technologies in real-world field operations.

5.1. Scientific Implications

AgROS enables scenery analysis enabled by the employment of place and object recognition
techniques and algorithms. At a greater extent, the capability of AgROS to generate topological maps of
customized fields and organize the emulated agricultural environment in an abstract manner, to enable
a UGV’s autonomous navigation and sensor-driven task-planning capabilities, methodologically
contributes to the field of robotics [64]. In the digitalization discourse, AgROS allows context-aware
robot navigation enabled by sensors, thus allowing the realization of “digital twins” in agriculture.
The backbone of AgROS integrates into a user-friendly emulation tool: (i) a real-world elevation model
via an automated process; (ii) an agri-environment (e.g., plants, orchard trees, etc.) to an elevation
model; (iii) a range of UGVs in the emulated agricultural environment; and (iv) a collection of routing
algorithms for the autonomous navigation of UGVs in the emulated operational setting.

5.2. Practical Implications

AgROS is a farm management emulation tool, based on ROS, that aims to enable farmers to
effectively introduce advanced technologies in real-world field operations. The developed AgROS
emulation tool could be used by farmers in order to explore and assess the use of agricultural robotic
systems on customized fields. More specifically, the tool allows farmers to: (i) select their actual field
from a map layout; (ii) import the landscape of the field to create an emulation equivalent; (iii) add
characteristics of the actual agricultural layout (e.g., trees, static objects); (iv) select an agricultural robot
(e.g., from a predefined list of commercially available systems) to import at the emulated environment;
and (iv) test the robot’s performance in a quasi-real-world environment.
All the implemented algorithms and modules of the emulated robot systems are compatible with
commercial real-world hardware equivalents and can even be deployed to custom real-world robots.
Any modifications, both at the field layout and at the selected routing algorithms, directly impact
the output of the system and provide valuable feedback to the farmer, prior to proceeding to any
cost-intensive actual field tests and applications.
Agronomy 2019, 9, 0403 17 of 20

5.3. Limitations
In conducting this research, some limitations exist which provide interesting grounds for exploring
further research horizons and expanding the practical contributions of AgROS. Firstly, emulation
modules of AgROS have been introduced to an experimental four-driving-wheels four-steering-wheels
electric-powered field robot for initial testing, as described in Section 4.4. However, a detailed
experimental protocol to compare the resulting efficiency and performance of robot navigation or field
mapping activities, obtained by both the emulation tool and a real-world equivalent implementation,
is still in design. Secondly, AgROS is in a perpetual beta stage of development and inherently
incorporates limited technical features from a soil science perspective. For example, along with the
field slope, field soil conditions and composition could also significantly affect the traversability of a
real-world ground vehicle.

5.4. Future Research

Regarding future research directions, we are planning to deploy an extensive experimentation
of AgROS and engage farmers to both familiarize them with digital technologies and receive actual
feedback and constructive remarks about the effectiveness and efficiency of the tool. Incorporating
feedback from farmers could guide the technical development of AgROS and enrich the tool with
further modules. We envision to position the tool as a “digital twin” paradigm in agriculture for
enabling predictive emulation of UGVs in field operations to assist in: (i) deriving operational
scheduling and planning programs for agricultural activities performed by robots [65]; (ii) informing
evidence-based implementation scenarios for agricultural operations with economic, environmental
and social sustainability benefits for agrifood supply chain stakeholders [66]; and (iii) empowering
farmers in the forthcoming landscape of digitalization in the agri-food sector [67].

Author Contributions: Conceptualization: N.T., D.B. (Dimitrios Bechtsis) and D.B. (Dionysis Bochtis);
methodology: N.T., D.B. (Dimitrios Bechtsis) and D.B. (Dionysis Bochtis); resources: D.B. (Dimitrios Bechtsis) and
N.T.; writing—original draft preparation: N.T. and D.B. (Dimitrios Bechtsis); writing—review and editing: D.B.
(Dionysis Bochtis); project administration: N.T. and D.B. (Dionysis Bochtis); funding acquisition: N.T. and D.B.
(Dionysis Bochtis).
Funding: This research has received funding from the General Secretariat for Research and Technology (GSRT)
under reference no. 2386, Project Title: “Human-Robot Synergetic Logistics for High Value Crops” (project
acronym: SYNERGIE).
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Tzounis, A.; Katsoulas, N.; Bartzanas, T.; Kittas, C. Internet of Things in agriculture, recent advances and
future challenges. Biosyst. Eng. 2017, 164, 31–48. [CrossRef]
2. Garnett, T.; Appleby, M.C.; Balmford, A.; Bateman, I.J.; Benton, T.G.; Bloomer, P.; Burlingame, B.; Dawkins, M.;
Dolan, L.; Fraser, D.; et al. Sustainable intensification in agriculture: Premises and policies. Science 2013, 341,
33–34. [CrossRef] [PubMed]
3. Sundmaeker, H.; Verdouw, C.; Wolfert, S.; Pérez Freire, L. Internet of food and farm 2020. In Digitising the
Industry—Internet of Things Connecting Physical, Digital and Virtual Worlds, 1st ed.; Vermesan, O., Friess, P.,
Eds.; River Publishers: Gistrup, Denmark; Delft, The Netherlands, 2016; pp. 129–151.
4. Vasconez, J.P.; Kantor, G.A.; Auat Cheein, F.A. Human–robot interaction in agriculture: A survey and current
challenges. Biosyst. Eng. 2019, 179, 35–48. [CrossRef]
5. Kunz, C.; Weber, J.F.; Gerhards, R. Benefits of precision farming technologies for mechanical weed control
in soybean and sugar beet—Comparison of precision hoeing with conventional mechanical weed control.
Agronomy 2015, 5, 130–142. [CrossRef]
Agronomy 2019, 9, 0403 18 of 20

6. Bonneau, V.; Copigneaux, B. Industry 4.0 in Agriculture: Focus on IoT Aspects; Directorate-General Internal
Market, Industry, Entrepreneurship and SMEs; Directorate F: Innovation and Advanced Manufacturing;
Unit F/3 KETs, Digital Manufacturing and Interoperability by the consortium; European Commission:
Brussels, Belgium, 2017.
7. Lampridi, M.G.; Kateris, D.; Vasileiadis, G.; Marinoudi, V.; Pearson, S.; Sørensen, C.G.; Balafoutis, A.;
Bochtis, D. A case-based economic assessment of robotics employment in precision arable farming. Agronomy
2019, 9, 175. [CrossRef]
8. Bechar, A.; Vigneault, C. Agricultural robots for field operations: Concepts and components. Biosyst. Eng.
2016, 149, 94–111. [CrossRef]
9. Miranda, J.; Ponce, P.; Molina, A.; Wright, P. Sensing, smart and sustainable technologies for Agri-Food 4.0.
Comput. Ind. 2019, 108, 21–36. [CrossRef]
10. Rose, D.C.; Morris, C.; Lobley, M.; Winter, M.; Sutherland, W.J.; Dicks, L.V. Exploring the spatialities of
technological and user re-scripting: The case of decision support tools in UK agriculture. Geoforum 2018, 89,
11–18. [CrossRef]
11. El Bilali, H.; Allahyari, M.S. Transition towards sustainability in agriculture and food systems: Role of
information and communication technologies. Inf. Process. Agric. 2018, 5, 456–464. [CrossRef]
12. Tsolakis, N.; Bechtsis, D.; Srai, J.S. Intelligent autonomous vehicles in digital supply chains: From
conceptualisation, to simulation modelling, to real-world operations. Bus. Process Manag. J. 2019, 25,
414–437. [CrossRef]
13. Rose, D.C.; Sutherland, W.J.; Parker, C.; Lobley, M.; Winter, M.; Morris, C.; Twining, S.; Ffoulkes, C.; Amano, T.;
Dicks, L.V. Decision support tools for agriculture: Towards effective design and delivery. Agric. Syst. 2016,
149, 165–174. [CrossRef]
14. Tao, F.; Zhang, M.; Nee, A.Y.C. Chapter 12—Digital Twin, Cyber–Physical System, and Internet of Things.
In Digital Twin Driven Smart Manufacturing, 1st ed.; Tao, F., Zhang, M., Nee, A.Y.C., Eds.; Academic Press:
New York, NY, USA, 2019; pp. 243–256.
15. Rosen, R.; von Wichert, G.; Lo, G.; Bettenhausen, K.D. About the importance of autonomy and digital twins
for the future of manufacturing. IFAC-PapersOnLine 2015, 48, 567–572. [CrossRef]
16. Leeuwis, C. Communication for Rural Innovation. Rethinking Agricultural Extension; Blackwell Science: Oxford,
UK, 2004.
17. Ayani, M.; Ganebäck, M.; Ng, A.H.C. Digital Twin: Applying emulation for machine reconditioning.
Procedia CIRP 2018, 72, 243–248. [CrossRef]
18. McGregor, I. The relationship between simulation and emulation. In Proceedings of the 2002 Winter
Simulation Conference, San Diego, CA, USA, 8–11 December 2002; Volume 2, pp. 1683–1688. [CrossRef]
19. Zambon, I.; Cecchini, M.; Egidi, G.; Saporito, M.G.; Colantoni, A. Revolution 4.0: Industry vs. agriculture in
a future development for SMEs. Processes 2019, 7, 36. [CrossRef]
20. Espinasse, B.; Ferrarini, A.; Lapeyre, R. A multi-agent system for modelisation and simulation-emulation of
supply-chains. IFAC Proc. Vol. 2000, 33, 413–418. [CrossRef]
21. Mongeon, P.; Paul-Hus, A. The journal coverage of Web of Science and Scopus: A comparative analysis.
Scientometrics 2016, 106, 213–228. [CrossRef]
22. Bechtsis, D.; Moisiadis, V.; Tsolakis, N.; Vlachos, D.; Bochtis, D. Unmanned Ground Vehicles in precision
farming services: An integrated emulation modelling approach. In Information and Communication Technologies
in Modern Agricultural Development; HAICTA 2017, Communications in Computer and Information Science;
Salampasis, M., Bournaris, T., Eds.; Springer: Cham, Switzerland, 2019; Volume 953, pp. 177–190. [CrossRef]
23. Astolfi, P.; Gabrielli, A.; Bascetta, L.; Matteucci, M. Vineyard autonomous navigation in the Echord++ GRAPE
experiment. IFAC-PapersOnLine 2018, 51, 704–709. [CrossRef]
24. Ayadi, N.; Maalej, B.; Derbel, N. Optimal path planning of mobile robots: A comparison study. In Proceedings
of the 15th International Multi-Conference on Systems, Signals & Devices (SSD), Hammamet, Tunisia,
19–22 March 2018; pp. 988–994. [CrossRef]
25. Backman, J.; Kaivosoja, J.; Oksanen, T.; Visala, A. Simulation environment for testing guidance algorithms
with realistic GPS noise model. IFAC-PapersOnLine 2010, 3 Pt 1, 139–144. [CrossRef]
26. Bayar, G.; Bergerman, M.; Koku, A.B.; Konukseven, E.I. Localization and control of an autonomous orchard
vehicle. Comput. Electron. Agric. 2015, 115, 118–128. [CrossRef]
Agronomy 2019, 9, 0403 19 of 20

27. Blok, P.M.; van Boheemen, K.; van Evert, F.K.; Kim, G.-H.; IJsselmuiden, J. Robot navigation in orchards with
localization based on Particle filter and Kalman filter. Comput. Electron. Agric. 2019, 157, 261–269. [CrossRef]
28. Bochtis, D.; Griepentrog, H.W.; Vougioukas, S.; Busato, P.; Berruto, R.; Zhou, K. Route planning for orchard
operations. Comput. Electron. Agric. 2015, 113, 51–60. [CrossRef]
29. Cariou, C.; Gobor, Z. Trajectory planning for robotic maintenance of pasture based on approximation
algorithms. Biosyst. Eng. 2018, 174, 219–230. [CrossRef]
30. Christiansen, M.P.; Laursen, M.S.; Jørgensen, R.N.; Skovsen, S.; Gislum, R. Designing and testing a UAV
mapping system for agricultural field surveying. Sensors 2017, 17, 2703. [CrossRef] [PubMed]
31. De Preter, A.; Anthonis, J.; De Baerdemaeker, J. Development of a robot for harvesting strawberries.
IFAC-PapersOnLine 2018, 51, 14–19. [CrossRef]
32. De Sousa, R.V.; Tabile, R.A.; Inamasu, R.Y.; Porto, A.J.V. A row crop following behavior based on primitive
fuzzy behaviors for navigation system of agricultural robots. IFAC Proc. Vol. 2013, 46, 91–96. [CrossRef]
33. Demim, F.; Nemra, A.; Louadj, K. Robust SVSF-SLAM for unmanned vehicle in unknown environment.
IFAC-PapersOnLine 2016, 49, 386–394. [CrossRef]
34. Durmus, H.; Gunes, E.O.; Kirci, M. Data acquisition from greenhouses by using autonomous mobile robot.
In Proceedings of the 5th International Conference on Agro-Geoinformatics, Tianjin, China, 18–20 July 2016;
pp. 5–9. [CrossRef]
35. Emmi, L.; Paredes-Madrid, L.; Ribeiro, A.; Pajares, G.; Gonzalez-De-Santos, P. Fleets of robots for precision
agriculture: A simulation environment. Ind. Robot 2013, 40, 41–58. [CrossRef]
36. Ericson, S.K.; Åstrand, B.S. Analysis of two visual odometry systems for use in an agricultural field
environment. Biosyst. Eng. 2018, 166, 116–125. [CrossRef]
37. Gan, H.; Lee, W.S. Development of a navigation system for a smart farm. IFAC-PapersOnLine 2018, 51, 1–4.
[CrossRef]
38. Habibie, N.; Nugraha, A.M.; Anshori, A.Z.; Masum, M.A.; Jatmiko, W. Fruit mapping mobile robot on
simulated agricultural area in Gazebo simulator using simultaneous localization and mapping (SLAM). In
Proceedings of the 2017 International Symposium on Micro-NanoMechatronics and Human Science (MHS),
Nagoya, Japan, 3–6 December 2018; pp. 1–7. [CrossRef]
39. Hameed, I.A.; Member, S. Coverage path planning software for autonomous robotic lawn mower using
Dubins’ curve. In Proceedings of the 2017 IEEE International Conference on Real-time Computing and
Robotics (RCAR), Okinawa, Japan, 14–18 July 2017; pp. 517–522. [CrossRef]
40. Hameed, I.A.; Bochtis, D.D.; Sørensen, C.G.; Jensen, A.L.; Larsen, R. Optimized driving direction based on a
three-dimensional field representation. Comput. Electron. Agric. 2013, 91, 145–153. [CrossRef]
41. Hameed, I.A.; La Cour-Harbo, A.; Osen, O.L. Side-to-side 3D coverage path planning approach for agricultural
robots to minimize skip/overlap areas between swaths. Robot. Auton. Syst. 2016, 76, 36–45. [CrossRef]
42. Han, X.Z.; Kim, H.J.; Kim, J.Y.; Yi, S.Y.; Moon, H.C.; Kim, J.H.; Kim, Y.J. Path-tracking simulation and field
tests for an auto-guidance tillage tractor for a paddy field. Comput. Electron. Agric. 2015, 112, 161–171.
[CrossRef]
43. Hansen, K.D.; Garcia-Ruiz, F.; Kazmi, W.; Bisgaard, M.; la Cour-Harbo, A.; Rasmussen, J.; Andersen, H.J.
An autonomous robotic system for mapping weeds in fields. IFAC Proc. Vol. 2013, 46, 217–224. [CrossRef]
44. Hansen, S.; Blanke, M.; Andersen, J.C. Autonomous tractor navigation in orchard—Diagnosis and supervision
for enhanced availability. IFAC Proc. Vol. 2009, 42, 360–365. [CrossRef]
45. Harik, E.H.; Korsaeth, A. Combining hector SLAM and artificial potential field for autonomous navigation
inside a greenhouse. Robotics 2018, 7, 22. [CrossRef]
46. Jensen, K.; Larsen, M.; Jørgensen, R.; Olsen, K.; Larsen, L.; Nielsen, S. Towards an open software platform for
field robots in precision agriculture. Robotics 2014, 3, 207–234. [CrossRef]
47. Khan, N.; Medlock, G.; Graves, S.; Anwar, S. GPS guided autonomous navigation of a small agricultural
robot with automated fertilizing system. SAE Tech. Pap. Ser. 2018, 1, 1–7. [CrossRef]
48. Kim, D.H.; Choi, C.H.; Kim, Y.J. Analysis of driving performance evaluation for an unmanned tractor.
IFAC-PapersOnLine 2018, 51, 227–231. [CrossRef]
49. Kurashiki, K.; Fukao, T.; Nagata, J.; Ishiyama, K.; Kamiya, T.; Murakami, N. Laser-based vehicle control in
orchard. IFAC Proc. Vol. 2010, 43, 127–132. [CrossRef]
50. Liu, C.; Zhao, X.; Du, Y.; Cao, C.; Zhu, Z.; Mao, E. Research on static path planning method of small obstacles
for automatic navigation of agricultural machinery. IFAC-PapersOnLine 2018, 51, 673–677. [CrossRef]
Agronomy 2019, 9, 0403 20 of 20

51. Mancini, A.; Frontoni, E.; Zingaretti, P. Improving variable rate treatments by integrating aerial and ground
remotely sensed data. In Proceedings of the 2018 International Conference on Unmanned Aircraft Systems
(ICUAS), Dallas, TX, USA, 12–15 June 2018; pp. 856–863. [CrossRef]
52. Pérez-Ruiz, M.; Gonzalez-de-Santos, P.; Ribeiro, A.; Fernandez-Quintanilla, C.; Peruzzi, A.; Vieri, M.; Tomic, S.;
Agüera, J. Highlights and preliminary results for autonomous crop protection. Comput. Electron. Agric. 2015,
110, 150–161. [CrossRef]
53. Pierzchała, M.; Giguère, P.; Astrup, R. Mapping forests using an unmanned ground vehicle with 3D LiDAR
and graph-SLAM. Comput. Electron. Agric. 2018, 145, 217–225. [CrossRef]
54. Radcliffe, J.; Cox, J.; Bulanon, D.M. Machine vision for orchard navigation. Comput. Ind. 2018, 98, 165–171.
[CrossRef]
55. Reina, G.; Milella, A.; Galati, R. Terrain assessment for precision agriculture using vehicle dynamic modelling.
Biosyst. Eng. 2017, 162, 124–139. [CrossRef]
56. Santoro, E.; Soler, E.M.; Cherri, A.C. Route optimization in mechanized sugarcane harvesting.
Comput. Electron. Agric. 2017, 141, 140–146. [CrossRef]
57. Shamshiri, R.R.; Hameed, I.A.; Pitonakova, L.; Weltzien, C.; Balasundram, S.K.; Yule, I.J.; Grift, T.E.;
Chowdharyet, G. Simulation software and virtual environments for acceleration of agricultural robotics:
Features highlights and performance comparison. Int. J. Agric. Biol. Eng. 2018, 11, 15–31. [CrossRef]
58. Thanpattranon, P.; Ahamed, T.; Takigawa, T. Navigation of autonomous tractor for orchards and plantations
using a laser range finder: Automatic control of trailer position with tractor. Biosyst. Eng. 2016, 147, 90–103.
[CrossRef]
59. Vaeljaots, E.; Lehiste, H.; Kiik, M.; Leemet, T. Soil sampling automation case-study using unmanned ground
vehicle. Eng. Rural Dev. 2018, 17, 982–987. [CrossRef]
60. Vasudevan, A.; Kumar, D.A.; Bhuvaneswari, N.S. Precision farming using unmanned aerial and ground
vehicles. In Proceedings of the International Conference on Technological Innovations in ICT for Agriculture
and Rural Development (TIAR), Chennai, India, 15–16 July 2016; pp. 146–150. [CrossRef]
61. Wang, Z.; Underwood, J.; Walsh, K.B. Machine vision assessment of mango orchard flowering.
Comput. Electron. Agric. 2018, 151, 501–511. [CrossRef]
62. Wang, B.; Li, S.; Guo, J.; Chen, Q. Car-like mobile robot path planning in rough terrain using multi-objective
particle swarm optimization algorithm. Neurocomputing 2018, 282, 42–51. [CrossRef]
63. Yakoubi, M.A.; Laskri, M.T. The path planning of cleaner robot for coverage region using genetic algorithms.
J. Innov. Digit. Ecosyst. 2016, 3, 37–43. [CrossRef]
64. Kostavelis, I.; Gasteratos, A. Semantic mapping for mobile robotics tasks: A survey. Robot. Auton. Syst. 2015,
66, 86–103. [CrossRef]
65. D’Urso, G.; Smith, S.L.; Mettu, R.; Oksanen, T.; Fitch, R. Multi-vehicle refill scheduling with queueing.
Comput. Electron. Agric. 2018, 144, 44–57. [CrossRef]
66. Bechtsis, D.; Tsolakis, N.; Vlachos, D.; Iakovou, E. Sustainable supply chain management in the digitalisation
era: The impact of Automated Guided Vehicles. J. Clean. Prod. 2017, 142 Pt 4, 3970–3984. [CrossRef]
67. Rotz, S.; Gravely, E.; Mosby, I.; Duncan, E.; Finnis, E.; Horgan, M.; LeBlanc, J.; Martin, R.; Neufeld, H.T.;
Nixon, A.; et al. Automated pastures and the digital divide: How agricultural technologies are shaping
labour and rural communities. J. Rural Stud. 2019, 112–122. [CrossRef]

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).