Abstract—In this paper we use a locally developed adaptive watering system as an example of a remote controlled laboratory (RCL) developed with standard open hardware and using libraries taken from the e-lab. This experiment is a particular case that could benefit from a large number of RCLs proposing different water budget strategies, allowing the studies of the best controller algorithm to save water. The water consumption log can be monitored in real-time and served to any user as a distributed remote laboratory with support of a Raspberry PI and a web connection, using an open source Arduino board and custom made shield. The ultimate goal of RCLs will be achieved when anyone can

1. Turning the internet of (my) things into a remote
controlled laboratory
A. Torres∗, M. Santos†, S. Balula†, J. Fortunato†, H. Fernandes†,
∗Departamento de F´ısica, T´ecnico Lisboa, Portugal
†Instituto de Plasmas e Fus˜ao Nuclear, T´ecnico Lisboa, Portugal
Abstract—In this paper we use a locally developed adaptive
watering system as an example of a remote controlled labora-
tory (RCL) developed with standard open hardware and using
libraries taken from the e-lab. This experiment is a particular
case that could benefit from a large number of RCLs proposing
different water budget strategies, allowing the studies of the
best controller algorithm to save water. The water consumption
log can be monitored in real-time and served to any user as a
distributed remote laboratory with support of a Raspberry PI
and a web connection, using an open source Arduino board and
custom made shield. The ultimate goal of RCLs will be achieved
when anyone can easily publish their own experiment in the
WWW.
Index Terms—e-lab, remote controlled laboratories, internet of
things, arduino, raspberry pi, electrical conductivity
I. INTRODUCTION
Installed in Instituto Superior T´ecnico (IST), Lisbon, the
e-lab laboratory provides remote control of real physics ex-
periments over the internet. In this paper, we describe a plant
watering control system, with which the plant is only watered
once a day, with the adequate amount of water, making the
system as robust as possible, adapting it’s response to the
environmental conditions.
In addition to the implementation at IST, the whole project
(hardware and software) is open source and available online
[1], allowing the reproduction of the experiment by anyone.
Being open source, the user can adapt the code to cope with
the different environmental conditions: different crops, soils
and vases, that will modify the physical response produced by
the watering.
The ultimate goal is to be able to compare different control
strategies, implemented in separate installations, through the
data upload to a common database.
Unlike other e-lab experiments, which aim at demonstrating
physical phenomena, this apparatus is focused in the Internet
of Things (IoT). According to [2], the IoT will grow to 26
billion units installed in 2020 representing an almost 30-fold
increase from 0.9 billion in 2009, making it an important topic
and too large to ignore in the e-lab scope.
By making the experiment available online, we intend to
show to the e-lab target demographics, namely high school
IPFN activities received financial support from “Fundac¸˜ao para a Ciˆencia
e Tecnologia” through project UID/FIS/50010/2013.
Corresponding author: A. Torres (email: andregtorres@tecnico.ulisboa.pt)
Fig. 1. Experimental setup: vases with probes attached and watering system.
teachers and students, that the Internet nowadays is, and will
continue to be integrated in more devices. It is important for
the next generations, of both engineering students and common
tech-aware people, to have the sensibility and understanding
of the implications of this increasing connectivity. With this
mission in mind, the experiment aims to raise the user’s
awareness to the IoT and how the internet could be used not
only in the more conventional and historical uses (scientific
and academic, finance and banking, social media, etc.), but
also in their things, like watering their plants.
II. EXPERIMENTAL SETUP
The main objective of the experiment is to monitor and
control the amount of water supplied to two vases containing
plants (figure 1). To achieve this, the soil’s temperature, the
soil’s conductivity and ambient luminosity are measured and
logged.
As for the hardware used in the experiment, the microcon-
troller chosen was the Arduino Uno board which provides an
easy to build and understand open source approach, which
in the scope of the experiment, is favorable when comparing
to more professional micro controllers, such as Microchip or
even other Atmel boards.
Since there is a need to measure several frequency signals,
and the Arduino Uno only has one Input Capture (IC) module,
a shield containing a multiplexer (MUX) and a driver for the
water valve was devised, taking into account that:
978-1-4673-8246-5/16/$31.00 ©2016 IEEE 24-26 February 2016, UNED, Madrid, Spain
2016 13th International Conference on Remote Engineering and Virtual Instrumentation (REV)
Page 371

2. Fig. 2. Example of the output of a probe, measured on an oscilloscope: In
yellow (CH1) - temperature signal; In blue (CH2)- impedance signal.
TABLE I
CONDUCTIVITY PROBE CALIBRATION.
ρ (g dm−3
) κ (μS cm−1) T (μs)
1 2.2 515
10 21.4 431
100 210 58
1000 1990 6
• As there is only one IC pin, all the probe contacts should
be multiplexed there. Digital pins of the Arduino can be
used to select the MUX entry.
• Since there is the need to control the water admission
valve, measure the amount of water fed to the plants and
read the luminosity sensor, the necessary pins for these
functions must be accessible.
• The peripherals and chips used have different operating
voltages, as the probes, luminosity sensor and the the
valve control driver need a 12V DC tension but the MUX
and pull-up resistors associated with the probes must be
supplied 5V DC.
• To prevent power surges, there is the need for decoupling
capacitors in the power lines.
III. PROBE SIGNAL AND CALIBRATION
The used probe provides two outputs: the temperature and
the impedance between it’s two contacts. These signals are
square waves of varying frequency as pictured in figure 2. As
both the temperature and conductivity increase, the frequency
of the respective signals will also increase. These variations are
nonlinear, thus each probe must be calibrated and the signal
frequency compared in a look-up table in the memory of the
Arduino.
As an example of a calibration, for the conductivity, the
probes were submerged in four solutions of NaCl with known
conductivities [3]. As for the temperature, the probe was
immersed in running water under agitation to ensure homo-
geneity at different temperatures, measured by a thermocouple.
TABLE II
TEMPERATURE PROBE CALIBRATION.
θ (oC) T (μs) θ (oC) T (μs)
13,7 501 28,1 298
16,3 472 29,9 287
18,4 416 31,9 266
20,2 392 34,2 239
22,0 372 35,8 223
24,1 347 37,3 219
26,0 326 39,7 158
Hardware server
Initialization
Data
aquisition
Data output
Time to
initiate
control?
Control code
Compute
responce
Actuation
Sleep
until next
aquisition
Y
N
Fig. 3. Simplified block diagram of the Arduino program.
IV. ARDUINO PROGRAMMING
The microcontroller used is an Arduino Uno, which is
fitted with an ATmega328 microprocessor. Figure 3 shows a
simplified block diagram of the Arduino’s programming. In
the data acquisition phase, the proceeding is the following:
(i) Enables the multiplexer inputs and selects one of the
probes. (ii) Averages 30 consecutive measurements for the two
signals, thus preventing some unwanted signal fluctuations.
This process is (iii) repeated for all probes before the code
compares the measured values in the lookup tables and inter-
polate/extrapolates the values in scaled SI units. Along with
a time-stamp and the luminosity measured through an array
of photo-sensitive diodes, these values are sent through the
serial UART (Universal Asynchronous Receiver/Transmitter)
port to a Raspberry Pi (RPi) connected to the internet, which
saves them in a database, allowing the remote querying of
the said data. Since the total experiment time has a span of
days, and since there is no need for a high sampling frequency,
the Arduino enters sleep mode, which keeps only one timer
powered, which is set to wake the board up at a pre-set time
interval.
Another component of the implemented code is the control
cycle. Unlike the data acquisition process, which runs at
a user-set frequency, 30 minutes, for instance, the control
978-1-4673-8246-5/16/$31.00 ©2016 IEEE 24-26 February 2016, UNED, Madrid, Spain
2016 13th International Conference on Remote Engineering and Virtual Instrumentation (REV)
Page 372

3. Fig. 4. Serial port result for a sampling time of 4s in an one probe system.
cycle runs daily. This function(s) are defined by the user
connecting to the experiment. There is however a sample
code with a Proportional-Integral-Derivative (PID) controller
implemented, that may serve as a model for the user to
base its controller. In this sample code, the microprocessor
computes the mean conductivity, which, in the assumption
that the conductivity is proportional to the soil hydration,
is in turn representative of the water consumed. This value
is then compared with a setpoint, obtaining the error (e(t))
conductivity and the adequate response u(t) is calculated
through the PID controller equation (1).
u(t) = kP · e(t) + kI
t
0
e(t)dt + kD
de(t)
dt
(1)
All the proportional, integral and differential dependencies
are absorbed by the kP , kI and kD, respectively. These factors
can be determined by the methods of Control Theory, but
since the actuation time for this particular system is so long, a
theoretical model was devised to be able to computer simulate
the control system and thus obtaining the said coefficients.
V. RESULTS AND THEORETICAL MODEL FOR THE WATER
CONSUMPTION
Figure 4 illustrates an example of the sensor output through
the serial port with only one probe and no luminosity sensor.
In order to output integer values, the units used were sub-
multiples of the conventional units in pedology: do
C and
μSm−1
for temperature and conductivity, respectively.
Figure 5 shows the output of this sensor over one acquisition
to serve as reference, using a small vase with a plant. The
plant was watered in the beginning of the time scale (t = 0
corresponding to 12 June 2015, 00:15), with enough water to
ensure the soil is wet and uniformly irrigated.
The second day data 5b was used to generate a mathematical
model of the conductivity response over time (σ(t)) to com-
(a)
(b)
Fig. 5. Reference acquisition 5a: Three days scope; 5b: Second day scope.
pute the optimal PID coefficients for the sample code. This
model has the following expression, on equation (2):
σ(t) = σ0 − α (1 + βl(t) + γT(t)) dt (2)
where l(t) is the simulation of the light intensity as in equation
(3) and T(t) the temperature sensor data:
l(t) =
sin 2π
24 t − π
2 , x ∈ [6, 18]
0, , otherwise
(3)
By allowing every user to implement a different control cycle,
we will be able to engineer an efficient watering system by
comparing the plant’s response to different control algorithms.
VI. E-LAB COMMUNICATION PROTOCOLS
A generic communication protocol, common to all other
experiments at e-lab, connecting the computer and the micro-
controller board via a serial connection is used. This standard
hastens the deployment of new experiments and facilitates
debugging. It comprises a state machine with four different
states: stopped, configured, started and reseted, as well as
the corresponding transitions, ordered by the generic driver,
a program running in the host computer, which could then be
connected to the e-lab server using Remote experiment Control
978-1-4673-8246-5/16/$31.00 ©2016 IEEE 24-26 February 2016, UNED, Madrid, Spain
2016 13th International Conference on Remote Engineering and Virtual Instrumentation (REV)
Page 373

4. Fig. 6. Experimental data and theoretical model
Fig. 7. e-lab infrastructure overview
(ReC) [4]. The open-source library for arduino is available at
[5].
The hardware server used is a Raspberry Pi (RPi), which
has been successfully implemented in previous experiments in
the e-lab, bringing improvements in power consumption and
overall cost, not compromising the performance [6].
The general structure [7]of e-lab is shown in figure 7.
VII. REMOTE ACCESS AND LIVE VIDEO STREAM
As the apparatus will be installed in a remote location,
without a static external ip address, open-vpn [8], a reliable
open source VPN implementation, was used. It provides plug
and play functionality (i.e. the remote apparatus can be moved
without any further configuration), can overcome restrictive
networks and allows remote programming, control and data
transmission.
As in [6], video streaming is performed using RPi’s pro-
prietary camera, at a maximum 1080p30 (H.264 encoding),
although resolution is reduced given the limited resources of
the clients’ networks.
VIII. ONLINE INTERFACE
The user will be able to interact with the ongoing experi-
ment in two different ways:
• First, by inputing the control algorithm, either setting
it up from scratch or by modifying the suggested PID
controller, tuning the parameters and the setpoint or the
observables used definition of the error (1) from the
measured data.
• In order for data to be accessible to both the user who set
up the experiment and other users to compare the results
of the different configurations, it is periodically uploaded
to a SQL database in the e-lab servers. In the case of this
experiment, the measured quantities are plotted using the
free JpGraph [9] library for PHP.
This implementation ensures that the experiment is reusable,
as this system and code for the control and data acquisition
can be used and modified by anyone, whilst merging the data
in a common database.
IX. CONCLUSIONS
The experiment was successfully implemented at IST, prov-
ing the concept. The data was continuously logged to the
database, allowing the monitoring of the watering effect on
conductivity, and indirectly the soil moisture.
The use of this experiment shows great potential, if used in
large scale, in achieving the main goal of comparing different
watering user solutions.
REFERENCES
[1] A. Torres, “Collaborative-smart-watering-repository,” 2015, [Online;
accessed-27-December-2015]. [Online]. Available: https://github.com/
e-lab-tecnico-ulisboa-pt/Collaborative-smart-watering
[2] Gartner, “Gartner says the internet of things installed base will grow to
26 billion units by 2020,” 2013, [Online; accessed 04-November-2015].
[Online]. Available: http://www.gartner.com/newsroom/id/2636073
[3] Foxboro, “Conductivity ordering guide,” 1999, [Online; accessed
June-2015]. [Online]. Available: https://www.grc.com/dev/ces/tns/
Conductivity v Concentration.pdf
[4] R. B. Henriques, A. S. Duarte, H. Fernandes, T. Pereira, J. Fortunato, and
J. Pereira, “Generic protocol for hardware control,” in 1st Experiment@
International Conference, 2011.
[5] S. Balula, “e-lab’s rec generic driver for arduino,” 2015, [Online;
accessed 11-November-2015]. [Online]. Available: https://github.com/
sbalula/rec-arduino
[6] S. Balula, R. Henriques, J. Fortunato, T. Pereira, H. Borges, G. S.
Amarante-Segundo, and H. Fernandes, “Distributed e-lab setup based
on the raspberry pi: the hydrostatic experiment case study,” in 3rd
Experiment@International Conference - exp.at’15, 2015.
[7] R. Neto, H. Fernandes, J. Pereira, and A. Duarte, “E-lab remote laboratory
integrated overview,” in Remote Engineering and Virtual Instrumentation
(REV), 2012 9th International Conference on. IEEE, 2012, pp. 1–7.
[8] D. Wiki, “Openvpn,” 2014, [Online; accessed 15-December-2014].
[Online]. Available: https://wiki.debian.org/OpenVPN
[9] JpGraph, “Graph creating library for php,” 2016, [Online; accessed
4-January-2016]. [Online]. Available: http://http://jpgraph.net/
978-1-4673-8246-5/16/$31.00 ©2016 IEEE 24-26 February 2016, UNED, Madrid, Spain
2016 13th International Conference on Remote Engineering and Virtual Instrumentation (REV)
Page 374