Professional Documents
Culture Documents
net/publication/290201414
CITATIONS READS
13 4,871
5 authors, including:
Some of the authors of this publication are also working on these related projects:
ROCCO: Robot Assembly System for Computer Integrated Construcción (EU ESPRIT III 6450) View project
All content following this page was uploaded by Juan Jesús Roldán Gómez on 05 November 2017.
Alberto Ruiz-Larrea, Juan Jesús Roldán, Mario Garzón, Jaime del Cerro, and
Antonio Barrientos
1 Introduction
The agriculture in greenhouses is an appropriate field to implement innovative
technologies. In fact, there are many proposals of autonomous systems for the
production monitoring, crop irrigation and nutrition, or ventilation and heating
in greenhouses (e.g. [14]). Nevertheless, the application of these technologies in
the greenhouses is usually restricted to those with experimental purpose or large
production. In the common facilities, the implementation of these technologies
may be difficult due to their cost and their complexity.
The monitoring of the environmental variables is fundamental in greenhouses.
This information is useful to perform the climate control, which can be at global
or local level, according to the resolution of the information and the features of
the greenhouse. In addition, this information is interesting for studying the pro-
duction and traceability of products. A complete log with spatial and temporal
information of the environmental variables allows to know the conditions of each
area of the greenhouse and to determine the optimal ones for plant growth and
maturation.
The aim of this paper is to use an Unmanned Ground Vehicle (UGV) as a
platform for measuring the properties of the soil of a greenhouse. This robot
2 A. Ruiz-Larrea, J.J. Roldán, M. Garzón, J. del Cerro and A. Barrientos
allows moving and placing the sensors in the desired point of the ground and
performing a path in the greenhouse to collect the measures over time. The
challenge of the navigation in a closed and occupied environment is considered
in the proposal of this work.
Section 2 describes the state of art of robotics in greenhouse farming. Section
3 studies the variables that influence the greenhouse climate or the plant growth
and describes the selection of sensors and robot and their integration. Section
4 addresses the navigation strategy in the greenhouse environment, which en-
compasses the path planning and following. Section 5 describes the experiments
performed for validating the platform and their results. Finally, section 6 sum-
marizes the conclusions of the work and the proposals for future ones.
2 State of art
ground platform than an aerial one, particularly if the soil has to be prepared
in order to take the measures.
This work continues this research line and proposes a UGV as a platform for
measuring ground properties, building maps of them and monitoring production.
In the following sections, the sensory and navigation systems are described with
detail, and the experiments and results are presented.
3 Sensory system
The design of sensory system takes into account the environmental variables
that must be monitored. Different climate models for greenhouses and growth
models for plants were studied in order to determine these variables.
The climate model proposed in [7] considers the heating power, ventilation
flow, wind speed, solar radiation, air temperature, air humidity and carbon diox-
ide concentration. On the other hand, the model of [11] takes into account out-
door temperature, solar radiation, heating power and ventilation flow for per-
forming the climate control. Finally, another model [20] includes as variables
the transpiration of plants, the cover condensation and the vapor flow in the
ventilation.
Other publications [10] and [5] show the relation between the maturation of
fruits and the ethylene concentration in the air. Ethylene behaves as a hormone
that controls the processes of growth and maturation in the plants.
The influence of water in the growth of plants is fundamental. It transports
nutrients and other substances required for their development. In addition, it
takes part in the chemical processes that control the metabolism of the plants.
As pointed in [8], a water deficit implies a reduction of the size of roots, tails
and leaves and, therefore, a reduction in the fruit production and quality.
Meanwhile, the ground temperature is a factor that limits the germination
of seeds, the growth of roots and the production of organic material. Therefore,
it is a fundamental variable for the development of plants and their processes of
nutrition and transpiration.
After this study, the variables of ground temperature and moisture were se-
lected for the sensory system. It should be noted that air variables, such as
temperature, humidity or carbon dioxide concentration, were measured in pre-
vious works [17]. In addition, the measurement of the nutrient concentration in
the soil is not considered, due to the sensors have a high response time (around
30 minutes) and, therefore, the monitoring of these values is not possible. Other
variables as ethylene emissions or chlorophyll fluorescence are posed for future
works, due to the complexity of the sensors and their integration.
3.1 Sensors
Fig. 1. Selected sensors: (a) MLX9614 for temperature. (b) SEN92355P for humidity.
– The sensor is able to measure the ground temperature without contact, which
reduces the complexity of system and the risk of damage caused by impacts
or scratches.
– The temperature range and accuracy are appropriate for the application,
due to they allow to measure the potential values of temperature in the
greenhouse with high precision.
– The cost is low in comparison to other non-contact temperature sensors
available in market.
3.2 Robot
The UGV used as sensor platform is a Robotnik Summit XL (figure 2). This
robot has a size of 722x610x392 mm and a weight of 45 kg. It has motors in the
four wheels, so it is able to rotate in place. The robot reaches a speed of 3 meters
per second. It is controlled by an embedded computer with Linux and Robot
Operating System (ROS). The autonomy of its batteries is about 180 minutes.
The UGV is equipped with an Inertial Measurement Unit (IMU) and a Global
Navigation Satellite System (GNSS) receiver for performing navigation. It also
has a linear laser scanner, which has range of 30 meters and angle of 270 degrees,
for recognizing the perimeter and avoiding the obstacles. Finally, it has a pan-
tilt-zoom camera that is useful for manual control or autonomous navigation.
The robot has a load capacity of 20 kg, so it is able to transport the sensors
and the components required to their work. The following section describes the
integration of sensors into the robot at hardware and software levels.
Fig. 2. (a) Robot coming through the plants, (b) Robot with sensory system, (c)
Integration of sensors.
3.3 Integration
The integration of the sensors in the robot has required the following actions:
– The sensors are connected to a controller that reads the signals, computes
the values and stores the information. In this work an Arduino UNO has
been chosen for its compatibility with the sensors (the temperature sensor
is connected by I2C and the moisture sensor is connected directly) and its
ease of use.
– The controller is connected to the embedded PC of the robot, in order to
synchronize the sensor and path following algorithms. Thus, when the path
following algorithm arrives to a measure point, it passes the control to the
6 A. Ruiz-Larrea, J.J. Roldán, M. Garzón, J. del Cerro and A. Barrientos
sensor algorithm. On the other hand, when the sensor algorithm obtains the
measures, it returns the control to the path following algorithm. In this work,
the sensor controller is connected to the robot computer via USB and it acts
as a ROS node in the robot architecture.
– An autonomous tool is developed in order to drill the ground and place the
moisture sensor for obtaining measures. This tool is shown in figure 2 and
is based on a threaded bar, which converts the motor rotation to a vertical
movement that drills the soil and place the sensor. The bar is moved by a
stepper motor, which is connected through a driver to the Arduino UNO.
An ultrasonic sensor controls the depth of the hole, while a contact sensor
detects the collecting of the tool. A structure of bars and plates ensures the
stability of the mechanism.
– The sensor algorithm is executed in the Arduino UNO. It takes the control
when the robot arrives to a measure point. At first, it controls the deployment
of the measurement tool and the placement of moisture sensor. Later, it
takes ten measures of temperature and humidity and computes the average
values. Finally, it sends the values of temperature and humidity and returns
the control to the path following algorithm.
4 Navigation system
A greenhouse is a closed place that commonly has high occupancy. It has regular
properties such as the crop layout, which is usually composed of crop lines (as
seen in figure 3), but it also has irregular elements such as the plants, which are
planted in regular places but grow irregularly.
Figure 3 shows the common exploitation in Almerı́a (Andalucia, Spain),
where there is the largest agglomeration of greenhouses in the world. The green-
houses of Almerı́a have an average surface of 6,200 m2 [9], which is clearly less
than the fields of outdoor agriculture. These greenhouses often present a front
side with one or more doors that can be used for input and output of machinery,
a set of main corridors with a width of around 2 meters, and a series of side
corridors with a width of around 1 meter.
A back and forth strategy [3] is selected for path planning. This strategy allows
the robot to cover all the surface of greenhouse and pass next to all the plants for
obtaining the measures (as shown in figure 4). Back and forth motion is described
in the algorithm 1. The algorithm has the following inputs: l, the length of the
crop lanes, w, the width of the sets of two crop lanes, L, the length of greenhouse,
and W, the width of greenhouse.
The path planning algorithm has been implemented in a ROS node, which
receives the parameters of the greenhouse and their corridors and returns the
trajectory as a list of points. The path must pass through all the points, with
the shortest possible longitude and without changing over time. In fact, the
A UGV approach to measure the ground properties of greenhouses 7
Fig. 3. Greenhouse views: (a) Outdoor, (b) Main corridor, (c) Side corridor
monitoring of the ground properties requires measuring at the same points over
time.
There are some possible differences between theoretical and real paths. On the
one hand, the robot can find obstacles in its path that it should avoid. On the
other hand, it should stop at the measure nodes, in order to obtain the ground
temperature and humidity. Figure 5 shows the ROS architecture that performs
the perception, guidance, navigation and control functions.
The navigation is performed by using the navigation stack of ROS [13]. It
receives the position and orientation through the integration of IMU, GPS and
odometry data in a Extended Kalman Filter (EKF). This method enhances
the accuracy to the order of a few decimeters. It also receives the location of
the obstacles around the robot detected by the laser scanner. This information
allows compensating the position errors. Finally, it receives the coordinates of
the next goal from the path planning and following module. The navigation
stack computes the trajectory that reaches the goal and converts it to speed
commands for the controller of the robot.
The path following node sends the goals to the navigation stack one by one.
This node controls the stop and start in the measure points, which allows the
deployment and work of sensors. In addition, it is able to cancel the goals when
A UGV approach to measure the ground properties of greenhouses 9
the robot cannot reach them: e.g. when the goal is unaccessible or when the time
to reach it is high.
6 Conclusions
Following the line of the previous works, this paper addresses the application of
robotics in greenhouse farming. Specifically, it proposes the use of a UGV as a
platform for measuring the ground properties of greenhouses.
A study of the literature about both greenhouse climate models and crop
growth models is performed in order to select the ground variables that should
10 A. Ruiz-Larrea, J.J. Roldán, M. Garzón, J. del Cerro and A. Barrientos
be measured: the ground temperature and the soil humidity. The MLX90614
temperature sensor and SEN92355P moisture sensor are selected and integrated
through an Arduino UNO controller and a mechanism to pick and place them.
A back and forth strategy is proposed for the movement in the greenhouse.
The path planning and following algorithms are implemented in Robot Op-
erating System (ROS) and connected to the sensor control algorithms. These
algorithms are adapted to the navigation in a closed and occupied place such as
a greenhouse.
Finally, the systems are validated in a test field, where maps of ground tem-
perature and humidity are performed. The application of this system in a pro-
ductive greenhouse and the integration with other systems (e.g. climate control
or irrigation) are proposed for future works.
References
1. I.F. Akyildiz, W. Su, Y. Sankarasubramaniam, and E. Cayirci. A survey on sensor
networks. Communications magazine, IEEE, 40(8):102–114, 2002.
A UGV approach to measure the ground properties of greenhouses 11