In this paper, we present a mini review to investigate the development progress of agricultural robots from different perspectives including classification, functionality, and characteristics. The current technical situations and features of grafting robot, picking robot, weeding robot, spraying robot, etc. are introduced with their corresponding application contexts. It is analyzed and concluded that high cost and intellectualization are two major factors that may challenge the popularity of agricultural robots. The development trends of agricultural robots are discussed as well.
Robotic technology has been wide used in different areas, such as home service, health care, advanced manufacturing, and agricultural industry [1-4]. Agricultural robot is a kind of automation equipment, which takes agricultural products as the operation objects with environmental perception and automated working function . It is the outcome of the rapid development of information technology. The use of agricultural robots can effectively improve production efficiency, reduce the labor burden of farmers and solve the problem of labor shortage. At present, great progress have been made in the field of agricultural robot. Robots adapting to different needs have emerged one after another, such as pesticide spraying robots, grafting robots, picking robots, harvesting robots [6-9]. The extensive use of agricultural robots is promoting the modern agricultural production to be increasingly intensive and large-scaled.
With the advent of the era of agricultural mechanization, many countries have invested greatly in the research and development of agricultural robots. Various types of robots have appeared frequently [10,11]. According to the different working objects, agricultural robots can be divided into farm robot, fruit and vegetable robot, animal husbandry robot and forest robot. The specific classification is shown in Table 1 .
Compared with industrial robots, agricultural robots have the
A. Complexity and hard predictability of the operating
Crop shape and size always change with time. Crop growth
environment is complex and changeable, which brings a severe
challenge to the operation ability of robots. This requires
agricultural robots to have strong adaptability and achieve high
intelligence in visual reasoning and judgment [13-15].
B. Seasonality of operating
Due to the limitation of technical level, most robots only
have a single function and work in a certain part of agricultural
production, which determines that robotic operation has seasonal
C. Particularity of users and prices
On the one hand, agricultural robots are mostly used by farmers
with low knowledge level, which requires that agricultural robots
must be simple and easy to operate. On the other hand, the overall
profit of agricultural products is not high, which determines that
the price of agricultural robots cannot be too high [16-18].
The research of automatic grafting equipment is mainly
in Japan, South Korea and China. Japan is the first country to
develop grafting robots, and then South Korea began to do this
research. In 1986, Japan Society for research and development
introduced three grafting prototypes, and then Japanese Institute
of Mechanization invented a fully automatic vegetable grafting
machine. Among them, the most representative type is Jing guan
GR800, which was introduced -to the market in October 1993.
GR800 uses artificial rootstocks and joints, and its success rate of
grafting is about 90% [19-21]. TGR Institute of Japan introduced
the grafting machine for melon grafting in 1995.
South Korea began to study the grafting robots in the 1990s.
The needle-type automatic grafting machine introduced by
South Korean Ideal System Company is their typical product. The
machine uses anti-rotation pentagonal ceramic needles to graft.
The whole hole tray can provide seedlings. The grafting speed can
reach 1200 plants/hour, as shown in Figure 1. Chinese research
on grafting robots started relatively late. China Agricultural
University is one of the earliest universities in China to study
agricultural robots [22-25]. The automatic grafting robot invented
by them has been tested in production. The robot solves the
problems of tenderness, vulnerability and inconsistent growth of
seedlings, and can be used in cucumbers, watermelons, etc. The
grafting robots is shown in Figure 2.
For picking robots, Tomato picking robot was the first one
studied. The picking robot developed by Knoto et al. in Japan
was the most influential one at that time . The robot mainly
consists of three parts: manipulator, vision sensor and mobile
mechanism. It separates fruit handle from fruit tree by rotation
of wrist joint. In 2004, California Machinery Company launched
an automatic Tomato Picking robot, which picked fruit and leaves
into the sorting bin, and then selected the fruit through sorting
equipment. The separated leaves were broken and sprinkled into
the field as fertilizer.
The tomato picking robot developed by Japanese Panasonic
Corp is shown in Figure 3. The small lens is used by the robot can
shot more than 70 thousand pixels of color images. Firstly, the
robot detects mature red tomatoes by image sensors, and then
locates the shape and position accurately. When picking, the robot
can only pull the vegetable pedicel without damaging the fruit. It can work continuously at night without human supervision. China
began to contact picking robots in the mid-1990s. For a period
of time, universities and research institutes in China have been
devoting themselves to the research and development of picking
robots and has made some achievements. For example, Shenyang
Institute of Automation has invented a tomato picking robot. As
shown in Figure 4, its manipulator is a four-fingered structure.
The manipulator is equipped with a vacuum suction cup driven by
an air pump, which can be used for fruit picking by the adsorption
of the suction cup.
At present, weed removal in the field is a big problem that
puzzles farmers. The general weeding methods are artificial
weeding and pesticide spraying. These methods not only consume
labor but will also cause soil pollution. Researchers at Wageningen
University in the Netherlands have designed an automatic weeding
robot, which uses a diesel engine as power source and can realize
four-wheel steering. The robot uses D-GPS and machine vision to
navigate . It can collect the boundary information of farmland
in real time and drive along the route of crops independently. The
robot also uses advanced vision technology to identify weeds.
Researchers of the Helmstad University in Sweden have
developed a weeding robot, which is controlled by a DC servo motor
and uses the Ackerman steering on a car as a control mechanism
. The robot has two cameras, one of which is a white camera
for identifying crop routes and guiding them. The robot travels
along the crop and another color camera is used to identify the
weeds. It is proved by experiments that the navigation accuracy of
the robot can reach ±2cm. as shown in Figure 5. German scientists
have developed a weeding robot named BoniRob . As shown
in Figure 6, the GPS positioning system of the robot can accurately
record the position of weeds and control the accuracy within 2cm.
BoniRob can remove 120 weeds in a minute, much faster than
manual and medicines.
Holland researchers have developed a tunnel type circular
sprayer, as shown in Figure 7. The robot has a tunnel shaped
cover. When spraying, the cover covers the crops to be sprayed,
so that the whole spraying process is carried out in a relatively
confined space. This reduces the influence of air flow on droplets.
The cover also has a liquid recovery system avoiding the waste of
liquid medicine. The robot is mainly used for pest control of crops
with lower crowns.
In 1990, Yamamoto launched the first unmanned helicopter
for pesticide spraying in the world. As shown in Figure 8, the
agricultural unmanned aerial vehicle is efficient, safe and free
from terrain restrictions. However, at present, spraying equipment
in China are mostly manually-driven, and only a few can self-walk.
A tractor sprayer has been developed by Chinese researchers,
which combines tractor with sprayer, and realizes indiscriminate
spraying, as shown in Figure 9.
Since the beginning of the 21st century, the research and
development of agricultural robots in various countries have
developed vigorously. Various agricultural robots adapting to
different needs have emerged, such as the seedling-raising robot
developed in Boston, USA. It consists of rolling tires, grippers and
trays. Workers only need to set location parameters on the touch
screen, and the robot can automatically move to the designated
working place and effectively shorten the seedling period.
A fruit sorting robot has been developed in the U.K., as shown
in Figure 10. The robot is equipped with an electronic image
sensor. It can recognize not only the size and shape of the fruit, but
also the color of the fruit, and judge the maturity of the fruit. The
robot picks the fruit by the giant pincers connected to the vacuum
tube to ensure that the fruit is intact. A grazing robot invented
by an Australian inventor uses two-dimensional and threedimensional
sensors, and the built-in global positioning system
can automatically detect the movement speed of cattle and drive
them away, as shown in Figure 11.
Although great achievements have been made in the
development of agricultural robots so far, there is still a long
way to go before the popularization of such robots. By taking the
references of robot applications in other fields, some development
trends of agricultural robots can be foreseen as follows:
A. Developing more advanced action planning approaches
for agricultural robots to improve the agricultural task
quality. In some cases, when the robot conducts an assigned
agricultural task, it may be inevitable have some negative
influences on other targets. For example, when a picking robot
picks some apples from the three, it may produce some harmful
actions on thin branches or leaves by its links or end-effector.
Therefore, some advanced robot action planning algorithms
based on cost functions can be developed to address such
issues to improve the agricultural task quality .
B. With the increasing complexity of agricultural tasks,
traditional agricultural robots cannot finish the tasks
effectively and accurately. Therefore, humans need to be
involved in such situations, which have similarly happened
in manufacturing contexts . Human farmer-agricultural
robot collaboration will be a research topic on the way of
agricultural robot development.
C. Making the agricultural robot more considerate will
be a trend in its development. With the development of
agricultural requirements, the agricultural robot will transfer
from a machine to be a “human partner” in most of agricultural
contexts . Therefore, the intelligence of the agricultural
robot such as understanding human intentions will be a large
gap to be filled .
D. Developing easy-to-use programming approaches
for agricultural robots. As we know, it is difficult for the
farmer users to program robots with professional expertise
when they want to update the robot to new agricultural
tasks. Therefore, developing new approaches by integrating
multi-modal human information such as natural language
or wearable sensing to have farmers program agricultural
robots intuitively and to make robot understand or predict
human intentions effectively will also be a distinct trend for
agricultural robots .
The cost is too high: In order to adapt to the complex working
environment, agricultural robots generally have complex structures
and control systems, which directly leads to the increase of
the manufacturing costs. In addition, due seasonal characteristics
of these robots, the low working time indirectly leads to the increase
of the cost, which makes the current cost-value ratio of agricultural
robots far from the market demands.
Lack of intelligence: In order to adapt to the complex
agricultural production environment, agricultural robots must
have a high level of intelligence. For example, in order to accurately
locate crops, agricultural robots must have the functions of
automatic path planning, navigation and obstacle avoidance.
Although the research in recent years has made breakthroughs
in machine vision, artificial intelligence and other aspects, the
degree of intelligence of agricultural robots cannot meet the needs
of the market.
Great achievements have been made in the development
of agricultural robots. Some agricultural production activities
can be completed independently by these robots. However, the
overall popularization of agricultural robots still faces some
challenges. With the development of automation technology,
more new techniques will be applied in to this field. In the future,
agricultural robots will be developed rapidly in the direction to
highly intelligent, comprehensive functions and environmental
protection. Agricultural robots will make great contributions to
the development of agriculture.