Review of agriculture robotics: Practicality and feasibility

Full text

Review of Agriculture Robotics:
Practicality and Feasibility
Sami Salama Hussen Hajjaj
Centre for Advanced Mechatronics and Robotics
Universiti Tenaga Nasional
43000 Kajang, Selangor, Malaysia
Email: ssalama@uniten.edu.my
Khairul Salleh Mohamed Sahari
Centre for Advanced Mechatronics and Robotics
Universiti Tenaga Nasional
43000 Kajang, Selangor, Malaysia
Email: khairuls@uniten.edu.my
Abstract—Concerns over food security have risen sharply in
recent years. The growing human population, coupled with the
shrinking agriculture resources, caused many governments and
international conglomerates around the world to seek new ways
to improve agriculture efficiency. This has lead to increased inter-
est, and spending, in Agriculture Robotics. In Part 1 of this work,
research activities on agriculture robotics were reviewed, with
many showing promising results. However, agriculture robots
remain experimental and far from being implemented on large
operational scales. This paper investigates the possible reasons
for this phenomena, by continuing the review of agriculture
robots, only this time focusing on practicality and feasibility.
Upon extensive review and analysis, the authors concluded that
practical agriculture robots rely not only on advances in robotics,
but also on the presence of a support infrastructure. This
infrastructure encompasses all services and technologies needed
by agriculture robots while in operation, this include a reliable
wireless connection, an effective framework for Human Robot
Interaction (HRI) between robots and agriculture workers, and
a framework for software sharing and re-use. Without such
infrastructure being in place, agriculture robots, no matter how
advanced in design they could be, would remain impractical and
infeasible. However, for many organizations, the technological
and monitory costs of establishing such infrastructure could be
very prohibitive, which renders agriculture robots uneconomical
and enviable. Therefore, the paper concludes that the key to
practical agriculture robotics is to find a novel, cost-effective,
and a reliable approach to develop the support infrastructure
needed for agriculture robots.
Index Terms—Agriculture Robots, Mobile Robots, Precision
Agriculture, Agriculture Automation, Outdoor Robots
I. INTRODUCTION
Concerns over food security have risen in recent years.
Human population and the demand for food is growing ever
rapidly, but agriculture resources are shrinking.
Many factors contributed to this situation; the ageing pop-
ulation of agriculture work force, with younger generations
opting for urban careers, the use of agriculture land for biofuel
and alternative energy, and others. More worryingly, these
factors are expected to increase in the coming years, ringing
socio-economical alarms around the world [1-3]
This led to increased interest in Agriculture Robots, with
global spending on is expected to rise from $817 million in
2013 to more than $16 billion by 2020.
Organizations from around the world are experimenting
on mobile robots, manipulators, humanoids, and drones in
various agriculture tasks. Robots are being tested in harvesting,
picking, herding, and other agriculture activities [4-6].
This continues the effort started in 2014. When the authors
published part I of this work, titled: Review of research in the
area of Agriculture Mobile Robots, [7]. That paper reviewed an
extensive list of 50 publications related to agriculture robots,
highlighting their achievements.
However, despite this apparent success, as well as the recent
developments in robotics in general, agriculture robots remain
experimental at best, and are far from being at applied in any
large scale agriculture operations.
This paper, part II of this effort, investigates this issue
further, by continuing the review of agriculture robots, only
this time focusing on practicality and feasibility. To identify
factors preventing agriculture robots from becoming practical
and feasible for large scale operations.
II. REVIEW OF AGRICULTURE ROBOTS:
PART I: ROBOTI CS DESIGN (SUMMARY)
For the sake of maintaining continuity, this section summa-
rizes part I of this work and highlights its findings.
The paper begins by discussing the importance and need for
agriculture robots, then it goes on to review research activities
on the subject, highlighting key achievements [7].
Agriculture robots, which operate in the outdoors and on
rough terrain, pose a unique set of engineering and tech-
nological challenges that are not usually present for indoor
robots. These agriculture-specific challenges were the focus
these research activities.
Upon further analysis of these publications, it was found
that researchers focused on the following three areas:
•Agriculture specific Navigation
•Agriculture specific Image Processing
•Agriculture specific Handling Rough Terrain
Researchers tackled issues such as navigation on a rough ter-
rain, handling wheel slip, illumination of natural light and its
effects on quality of image processing, the effects of greenery
on localization, stability of tractor-trailer motion, mechanical
design, and more. For specific details and references, the
reader is referred to Part I of this effort [7].

III. REVIEW OF AGRICULTURE ROBOTS,
PART II : PRACTICALITY AND FEASIBILITY
When considered individually, the publications and achieve-
ments reviewed in Part I are promising and exciting. Re-
searchers tackled the challenges mentioned above and reported
success in tackling them.
However, and despite this success, as well as the recent
advances in robotics, agriculture robotics remain experimental.
There is no reported application of agriculture robots on
industrial level, or in large scale agriculture operations, [5].
For that to happen, agriculture robots need to become
practical and feasible, and the following researchers recog-
nized this fact, and attempted to enhance their robots with
support systems, this section reviews these works, highlighting
strengths and weakness in each.
A. Durmus et al. (2015)
This team developed Agrobot; a multi-purpose agriculture
robot. From the start, they divided their work into two main
parts; designing the robot, and developing a cloud-based
service to link that robot to farmers via their mobile devices.
They built their robot in-house under a 2000$ budget, using
available tooling and equipment, and were able to control it
using a wireless controller, [8].
The interesting thing about their work, is that they identified
what makes a successful agriculture robot; not just a complete
mobile robot, but also a robot that is connected to the farmers
wirelessly, through a cloud service, and through the farmers’
mobile devices, [8].
Unfortunately however, their implementation of this idea
was not yet complete. Their robot’s software focuses only
hardware abstraction, and it is not autonomous. Secondly,
although they outlined their plans for the their cloud service,
they are yet to publish or report about it.
Aside from the vague RF Link needed to connect the robot
to the cloud, there was no mention of how this link-up would
take place, and how the service would be compatible with the
farmer’s mobile devices.
B. Kashiwazaki et al. (2010)
This team proposed the Greenhouse Partner Robot System,
which is an agricultural support system designed to facilitate
cooperation between humans and robots. Focusing on two
agriculture activities; harvesting and pest control, [9].
They introduced a four-wheeled mobile carts, that run
autonomously on a guidance line in the greenhouse using its
tracker sensors. For controlling the harvesting and pest control
activities, the carts also contain a control area that includes a
joystick, control panel, and an RFID tag reader, [9].
The robot travel within the space of the greenhouse au-
tonomously, while the human Partner uses its control panel to
perform the required agriculture tasks, [9].
This team, as did the first team, recognized the importance
of establishing an interaction between robots and the agri-
culture robots, hence the emphasise that was placed on the
support system and not just the actual robot.
Another interesting thing about this work, is how they
separated the design and programming of the robot into two
parts; the design and programming of the autonomous robot,
and that of the agriculture task. In essence, compartmentalizing
the task of programming the agriculture robot, which allowed
for more focus and complexity to be achieved.
However, and although the robot in this set-up did automate
the target agriculture tasks (harvesting and pest control), it is
still limited to the controlled environment of the greenhouse.
The robot is not suitable for the open field as there is not
protocol for a long distance interaction.
C. Amer et al. (2015)
This team developed a prototype for a multi-purpose agri-
culture robot, named Agribot, to perform multiple agriculture
tasks. Through its hexpad body design and walking mecha-
nism, the robot could travel in any direction with ease, [10].
Their robot is linked to the world via WiFi, its on-board
laptop connects to a WiFi signal. Through this connection,
the robot is linked to its operator. Also, the WiFi connection
allows the robot to connect to other robots and coordinate
tasks, [10].
While this robot demonstrated great agriculture versatility,
through its ability to perform various agricultural tasks, such as
harvesting, and weeding. The obvious limitation of this robot
is its dependence on WiFi.
As reported by the team themselves, the robot is limited to
a small area where the WiFi signal is present, and although
the use of WiFi signal booster was discussed in the report, it
was not actually tested, [10].
This is an example of how lack of infrastructure and support
would render an agriculture robot impractical, no matter how
good its design and performance are. In this work, it would be
impractical to establish WiFi connections in the open fields,
as WiFi routers and cables would be exposed to the elements,
power and maintenance costs would be too high.
D. Tardaguila et al. (2014)
This group developed the VineRobot, an autonomous agri-
culture mobile robot designed to help in the production and
agriculture of wine by the Seventh Framework Program, which
is sponsored by the European Union (EU), figure 1, [11]
Figure 1. The VineRobot, an agriculture mobile robot, [12]

The VineRobot is designed to roam the field autonomously,
collecting data on the state of the vineyard, such as the
vegetative development, water status, grape composition, and
other important data.
The acquired images and data generated by the robot is then
transferred wirelessly to the grape growers and technicians
monitoring the robot in real time.
This data is then processed in real time and used to focus
efforts and resources, such as manpower and equipment, where
they are needed in the field, hence fulfilling the meaning of
the term Precision Agriculture, [11], [12].
In this setup, this robot is an example of a non-intrusive
agriculture robot; the task of this robot, roaming the field and
collecting data, does not interrupt the activities of the farmers
or growers, nor does it require direct interaction with them.
Instead, the robot can perform its task independently from
the farmers, who would be busy picking grapes, working the
lands, and others tasks, perhaps oblivious the very existence
of the robot and its activites.
Being non-intrusive may be possible for this task, but is not
the case for other agriculture tasks. Also, an effective human
robot interaction is still needed for monitoring and diagnostics
purposes. Also, and as discussed earlier, an effective wireless
communication is needed between the robot and its operators.
E. Eaton et al. (2008)
This team identified the need to establish a dedicated infras-
tructure for agriculture robotics. They proposed the Precision
Farming System to manage the various automated agriculture
activities, as shown in figure 2, [13].
Figure 2. The farming system architecture, proposed by [13]
The proposed system is built on two primary subsystems;
the Precision Farming Data Set (PFDS), and the Precision
Agriculture Data Set (PADS). Together, these two system
would enable the functionality of agriculture robots.
However, and as reported by the authors themselves, the
proposed system is a complex system of systems, each with its
own subsystems as well. Also, these systems require redefining
and pre-planning of the layouts of the fields prior to operations,
which is obviously very disruptive.
Although their idea may have been too complex to achieve,
the concept itself is commendable. Agriculture robots do need
a support system that outlines field information, establishes
wireless contact with human operators, and helps robots gather
information about the field.
F. Emmi et al. (2014)
This group of researchers argued that incorporating many
electronic systems in agriculture robots would boost their
functions, but it would also increase its size, weight, costs,
and reliability. Therefore, they proposed that for an agriculture
robot to be successful, it must strike a balance between
complexity and functionality.
They developed their own system architecture, named the
RHEA project. The RHEA caters for robot design and control,
WiFi communication, software interface for human users.
Basically, all the important points discussed so far about
agriculture robots, [14]
They reported a successful implementation of their project,
their prototype comprises of three ground mobile units based
on a commercial tractor chassis, fitted with various equipment,
as shown in figure 3, [14].
Figure 3. General hardware architecture for the RHEA project, [14]
Although they identified the need to reduce size, complexity,
and costs of agriculture robots. Their design, as shown in the
figure, is far from small, simple, or cheap. Furthermore, and
as reported by the authors themselves, this work is limited by
the WiFi signal, which is just around 150 meters at best, [14]
This is yet another example of a great concept, but not
a practical implementation. Although this team demonstrated
their ideas and showed a proof of concept, they cannot
implement this in a high volume operations due to the limited
networking capabilities of the WiFi network.
G. Ishibashi et al. (2013)
Recognizing the limitation of WiFi, this team developed a
cost effective, web-based monitoring system for their agricul-
ture robot [15].
The robot’s on-board computer would gather robot’s data,
then this data is sent via bluetooth to a mobile device placed

on the robot. The phone is pre-programmed to send this data
through the telecommunication network to an HTTP server,
where it would be picked up by web-application, running on
a personal computer or another smart phone.
However, this is a one-way communication, as the the
operator can only observe the robot, but cannot control it, or
stop it, diagnose it, or influence it in any other way. Secondly,
the amount and type of data is limited to simple text. higher
level data such as images, or any other type of decoded data
can be transmitted using this approach.
Although the implementation is limited, this is an excellent
example of using existing technologies to develop a cost
effective infrastructure. Apart from the cost of the mobile
device and the Telecom subscription, there was no further costs
needed to achieve this connectivity, [15].
IV. IMPLEMENTING PRACTICAL AGRICULTURE ROBOT S
From the reviews conducted in Parts I and II of this work,
the following understanding can be developed.
For agriculture robots to become practical and feasible,
the following set of unique technological challenges must be
overcome, namely:
1) Hardware Incorporation of Agriculture Machinery: The
first step of automation of agriculture machinery is Hardware
Abstraction; which is developing the software models that
describe the machinery/robots to the automoation system.
Before a computer program is able to tell a robot to perform
an agriculture task, the program needs to know the robot’s
physical attributes and its control system, this includes the
robot’s dimensions, size, sensors, actuators, and controllers.
Whether agriculture robots are built in-house, purchased
from a supplier, or customized from existing agriculture ma-
chinery, its hardware needs to be successfully incorporated
before any automation software is developed.
2) Establishing Wireless Connectivity between robots and
humans: For agriculture robots and humans to interact, there
needs to be a dedicated wireless network to establish commu-
nication. This network must cover the size of the open field,
and it must cater for two modes of interactions needed in
agriculture; close proximity, and long-range connections.
In close proximity connection, the robots and humans are
in direct visual contact and share the work same space,
usually collaborating on an agriculture task, such as picking,
harvesting, and transportation.
In long-range interaction, robots are far out of visual con-
tact, operators monitor the robot from a computer on a remote
location, often via tracking devices and visualization tools.
3) Developing Software for agriculture robots: As dis-
cussed earlier, programming agriculture robots present devel-
opers with a unique set of challenges that are not present for
indoor robots. These agriculture specific challenges must be
tackled and managed when programming for Navigation of
agriculture robots.
This include software for sensor data processing, while tack-
ling illumination issues, obstacle avoidance, both dymanic and
static, localization withing a field of greenery, and handling the
effects of slip and stability on rough terrains.
Furthermore, software that allows the robot perform generic
agriculture task is also needed, such as: harvesting, picking,
watering, transportation, and others.
4) Implementing Human Robot Interaction tools: The role
of human workers in agriculture would not be completely
eliminated by the introduction of robots. Humans are still
needed for supervision and collaborative tasks. In many cases,
humans would still be needed to load/unload robots, help
guide/restablize robots, and generally work with robots.
Also, operators monitoring the robots, be it from near or
far, are surely needed to observe the robot while in action,
perform online diagnostics and analysis, and so insure proper
peerformance.
For all of this to happen, an effective Human Robot Inter-
action (HRI) is needed. This include hardware and software
tools, such as portable command consoles, installed with
proper robot software, to control, monitor, and work with the
robots.
5) Software re-usability and reliability: Generally speak-
ing, agriculture tasks are similar or generic. For example, the
agriculture task of picking in one field, say field A, is more
or less the same picking task in another field, field B. This is
specially true if the crop being picked is the same.
As such, if an organization operating fields A and B would
like to automate their operations, they do not need to create
new software from scratch for field B, but rather re-use
software already utilized in field A.
For this to happen, there needs to be a mechanism for
sharing and re-using software, along with collaboration of
knowledge and expertise. This can also be expanded to all
types of software discussed so far, such as software used in
hardware abstraction, connectivity, and HRI tools.
V. DISCUSSION:
CHALLENGES OF IMPLEMENTING AGRICULTURE ROBOTS
To summarize the points discussed above, implementing
practical agriculture robots requires establishing and perform-
ing the following:
1) Incorporating agriculture machinery
2) Establishing wireless connectivity (short and long range)
3) Developing robot software for agriculture robots
4) Implementing effective HRI tools
5) Enabling software re-usability and reliability.
As it cen be seen, aside from robot design and technology,
practical agriculture robots requires the setup of other factors,
such as an established Connectivity, an effective HRI, and the
re-usability of software. Collectively, let us term these issues
as the Support Infrastructure for agriculture robots.
This explains why agriculture robots today remain exper-
imental and far from being opertational. Researchers, such
as those reviewed in Part I of this work, focused entirely
on robotics design and technology, but not on the Support
Infrastructure related issues, such as connectivity, HRI, and
software sharing and re-usability.

VI. DISCUSSION:
ESTAB LI SH IN G TH E SUP PO RT INFRASTRUCTURE
FO R AGRICULTURE ROBOTS
Establishing the Support Infrastructure for agriculture robots
is not an easy undertaking. Setting up wireless connectivity,
networking, and routing is not simple. HRI tools, software and
hardware, require time and effort to design and build. Finally,
mechanism for software sharing and re-usability needs to be
designed and deployed as well.
As such, the technological and monitory costs of imple-
menting the Support Infrastructure is just too high and might
be prohibitive for many. This may offset the very benefit of
implementing agriculture robots in the first place.
For many organizations and conglomerates, this renders
agriculture robots impractical and infeasible. A visual repre-
sentation of this situation is shown in figure 4.
Figure 4. The Challenge of Implementing Agriculture Robots
As such, the key to implementing practical agriculture
robots, is solving the technological challenges associated with
establishing the support infrastructure needed by agriculture
robots. Specifically, challenges to establishing connectivity,
HRI tools, and software sharing and re-usability.
To achieve this, a novel approach (or a collection of ap-
proaches) is needed to solve these technological challenges,
help setup the required support infrastructure, and so facilitate
the implementation of practical agriculture robots.
VII. CONCLUSIONS
Interest in agriculture robotics have soared in recent years.
Global spending and research activities on the subject is
experiencing a near exponential growth.
Part I of this work reviewed over 50 of these activities, high-
lighting their achievements. These include agriculture-specific
navigation, image processing, and other robotics challenges
agriculture robots face. However, despite these successes,
agriculture robots remain far from being operational. This
indicates that something else is needed.
This paper, Part II of this work, showed that agriculture
robots need a support infrastructure. This would provide
a reliable wireless connectivity for the robots in the field,
an effective HRI tools between robots and humans, and a
framework for robot software sharing and re-usability.
At the moment, implementing such infrastructure is very
challenging, the technological and monitory costs of imple-
menting such infrastructure could be prohibitive.
This in turn, renders agriculture robots impractical and
infeasible. Therefore, The key to implementing agriculture
robots is to to find a novel, practical, and a reliable approach
to implementing the support infrastructure it needs.
VIII. ACKHNOWLEDGEMENTS
The authors would like to thank the College of Engineering
and the Innovative & Research Manangement Centre (iRMC),
UNITEN, for their continued support of this work.
REFERENCES
[1] The United States Department of Agriculture, “USDA agriculture pro-
jections to 2024,” Office of Cheif Economist, Tech. Rep., 2015.
[2] The Food and Agriculture Organization of the UN, (FAO), “The state of
food insecurity in the world 2015. meeting the 2015 international hunger
targets: taking stock of uneven progress,” World Food Organization,
Tech. Rep., 2015.
[3] M. Sahi. (2015, February) The rise of agrcultural robots. https://www.
tractica.com/automation-robotics/the- rise-of- agricultural-robots/. Trac-
tica, a market research andintelligence firm that focuses on human
interaction with technology. 1111 Pearl Street, Suite 201, Boulder, CO
80302 USA. Accessed=2016-06-21.
[4] The Radiant Insights Institute, “Global agricultural robots market -
global industry analysis, size, share, growth, trends and forecast, 2014
to 2020,” Radiant Insights, Tech. Rep., 2104.
[5] F. Tobe. (2016) Are agriculture robots ready? https://www.
therobotreport.com/news/ag-in-transition-from-precision-ag-to-full.
The Robot Report. Accessed=2016-06-21.
[6] D. Lane and R. Buckingham, “Ras 2020 national strategy for robot and
automation,” Robot and Automatino Systems, Special Interest Group,
Tech. Rep., 2014.
[7] S. S. H. Hajjaj and K. S. M. Sahari, Review of Research in the
Area of Agriculture Mobile Robots. Singapore: Springer Singapore,
2014, pp. 107–117. [Online]. Available: http://dx.doi.org/10.1007/
978-981- 4585-42- 2 13
[8] H. Durmus, E. O. Gunes, M. Krc, and B. B. Ustundag, “The design
of general purpose autonomous agricultural mobile-robot: Agrobot,” in
Agro-Geoinformatics (Agro-geoinformatics), 2015 Fourth International
Conference on, July 2015, pp. 49–53.
[9] K. Kashiwazaki, Y. Sugahara, J. Iwasaki, K. Kosuge, S. Kumazawa,
and T. Yamashita, “Greenhouse partner robot system,” in Robotics (ISR),
2010 41st International Symposium on and 2010 6th German Conference
on Robotics (ROBOTIK), June 2010, pp. 1–8.
[10] G. Amer, S. M. M. Mudassir, and M. A. Malik, “Design and operation
of Wi-Fi agribot integrated system,” in Industrial Instrumentation and
Control (ICIC), 2015 International Conference on, May 2015, pp. 207–
212.
[11] J. Tardaguila, “The vinerobot project,” http://www.vinerobot.eu/, 2014,
accessed=2016-06-16.
[12] Sciencedaily, “Agriculture robot,” https://images.sciencedaily.com/2015/
01/150128113713 1 900x600.jpg, 2015, accessed=2016-06-14.
[13] R. Eaton, J. Katupitiya, K. W. Siew, and K. S. Dang, “Precision guidance
of agricultural tractors for autonomous farming,” in Systems Conference,
2008 2nd Annual IEEE, April 2008, pp. 1–8.
[14] L. Emmi, M. G. de Soto, G. Pajares, and P. G. de Santos, “New trends
in robotics for agriculture: Integration and assessment of a real fleet of
robots,” The Scientific World Journal, vol. 2014, p. 21, March 2014,
article ID 404059.
[15] M. Ishibashi, M. Iida, M. Suguri, and R. Masuda, “4th ifac
conference on modelling and control in agriculture, horticulture and
post harvest industry remote monitoring of agricultural robot using
web application,” IFAC Proceedings Volumes, vol. 46, no. 18, pp. 138
– 142, 2013. [Online]. Available: http://www.sciencedirect.com/science/
article/pii/S1474667015349739