Available online at www.sciencedirect.

com

ScienceDirect
A IFAC PapersOnLine 55-32 (2022)
hybrid 188–193
A hybrid dynamical
dynamical systems systems approachapproach to to
A
A
A hybrid
hybrid
hybrid dynamical
smart
dynamical
dynamical
smart systems
irrigation
systems
systems
irrigation approach
approach
approach to
to
to
A hybrid dynamical smart
smart systems
irrigation
irrigation approach to
smart
R. Bertollo ∗∗ G.B. Cáceres
smart irrigation
∗∗
P.
irrigation
∗∗ Millán ∗∗
∗∗ M. Pereira ∗∗
∗∗
R. Bertollo G.B. Cáceres P. ∗,∗∗∗
Millán M. Pereira
L. Zaccarian
∗ ∗∗ ∗,∗∗∗ ∗∗ ∗∗
R.
R. Bertollo
Bertollo ∗ G.B.
∗ Cáceres
L. Zaccarian ∗∗ P.
∗∗ Millán ∗∗ M.
∗∗ Pereira ∗∗
∗ G.B. Cáceres ∗∗ P. Millán ∗∗ M. Pereira
∗ ∗∗ ∗∗ ∗∗
∗∗
R.
R. Bertollo
Bertollo G.B. Cáceres
∗ G.B. Cáceres
L. Zaccarian ∗∗ P.
P. Millán
∗,∗∗∗
Millán
∗,∗∗∗ ∗∗ M.
M. Pereira
Pereira ∗∗
∗∗
∗ R. Bertollo G.B. Cáceres L. Zaccarian P. Millán
∗,∗∗∗
∗,∗∗∗ M. Pereira
∗ Dipartimento di Ingegneria Industriale, University of Trento, Italy
∗,∗∗∗
Dipartimento di(e-mail: Ingegneria L.
L. Zaccarian
Zaccarian
Industriale, ∗,∗∗∗
University of Trento, Italy
∗
∗ Dipartimento di Ingegneria riccardo.bertollo@unitn.it)
Industriale, University of Trento, Italy
∗
∗ Dipartimento
∗∗ di(e-mail:
Ingegneria riccardo.bertollo@unitn.it)
Industriale, University of Trento, Italy
∗∗ Universidad
∗ Dipartimento
∗ Dipartimento di
di Loyola
Ingegneria
Ingegneria
(e-mail: Andalucı́a,
Industriale,
Industriale, Dos Hermanas,
riccardo.bertollo@unitn.it) University
University Spain
of
of Trento,
Trento,(email: Italy
Italy
Universidad
Dipartimento
gbcaceres@uloyola.es, di Loyola
Ingegneria
(e-mail: Andalucı́a,
Industriale, Dos
riccardo.bertollo@unitn.it)
pmillan@uloyola.es, Hermanas,
University Spain
of Trento,
mpereira@uloyola.es) (email: Italy
∗∗
∗∗ Universidad (e-mail: riccardo.bertollo@unitn.it)
gbcaceres@uloyola.es,
∗∗
∗∗ Universidad
∗∗∗ Loyola
(e-mail:
Loyola Andalucı́a,
pmillan@uloyola.es, Dos
riccardo.bertollo@unitn.it)
Andalucı́a, Dos Hermanas,
Hermanas, Spain
mpereira@uloyola.es)
Spain (email:
(email:
∗∗ Universidad
∗∗ Universidad
CNRS, LAAS,
Loyola
Loyola Université
Andalucı́a,
Andalucı́a, Dos
Dosde Toulouse,
Hermanas, France
Spain (email:
∗∗∗
gbcaceres@uloyola.es,
Universidad CNRS,
gbcaceres@uloyola.es, LAAS,
Loyola pmillan@uloyola.es,
Université
Andalucı́a,
pmillan@uloyola.es, DosdeHermanas,
Toulouse,
Hermanas, Spain
Spain (email:
mpereira@uloyola.es)
France
mpereira@uloyola.es) (email:
gbcaceres@uloyola.es,
∗∗∗
gbcaceres@uloyola.es,
∗∗∗ CNRS, LAAS, pmillan@uloyola.es,
pmillan@uloyola.es,
Université de mpereira@uloyola.es)
mpereira@uloyola.es)
Toulouse, France
gbcaceres@uloyola.es,
∗∗∗
∗∗∗ CNRS, LAAS,
∗∗∗ pmillan@uloyola.es,
Université de mpereira@uloyola.es)
Toulouse, France
Abstract: We present∗∗∗a CNRS, novel model
CNRS, LAAS,
LAAS,forUniversité de
de Toulouse,
a drip irrigation
Université Toulouse, system,France
France using a hybrid dynamical
Abstract:
systems We
paradigm present
to a
take novel
into model
account forthea drip irrigation
intrinsic binary system,
nature using
of theaa control hybrid dynamical
input. We
Abstract:
systems
Abstract: We
paradigm
We present
to
present a
take
a novel
into
novel model
account
model for
fortheaa drip irrigation
intrinsic
drip binary
irrigation system,
nature
system, using
of
using the a hybrid
control
hybrid dynamical
input. We
consider
Abstract:
Abstract:
systems an We irrigation
We
paradigm present
present
to pipe
a
a
take withmodel
novel
novel
into a generic
model
account for
forthe aanumber
drip
drip of binary
drippers,
irrigation
irrigation
intrinsic system,
system,analyzing
nature using
using
of the aathehybrid
hybrid
controleffectdynamical
that the
dynamical
dynamical
input. We
consider
Abstract:
systems
transients an irrigation
We
paradigm present
inirrigation
the pipe to pipe
a
take with
novel
into
haveinto a
model generic
account
on account
the forthe
irrigation anumber
drip
intrinsic of
uniformity, drippers,
irrigation
binary system,analyzing
nature
andnature
propose using
of the
an the athehybrid
controleffect
intuitive that
dynamical
input.
irrigation the
We
systems
systems
consider
transients paradigm
paradigm
an in the to take
to
pipe take
pipe
have into
with
on account
a
the generic thenumber
the
irrigation intrinsic
intrinsic of
uniformity, binary
binary
drippers,and nature of the
of
analyzing
propose the
an control
controleffect
intuitive input.
input.
that
irrigation We
We
the
systems
consider based
strategy
consider paradigm
an irrigation
an irrigation
on to
the take
pipe
one
pipe into
with aaccount
currently
with a generic
generic
used the by intrinsic
number
farmers.
number of binary
of drippers,
drippers,
The nature
model is of
analyzing
analyzing the
presented control
the effect
the effect input.
together thatwith
that We
the
the
consider
transients
strategy
consider
transients an
an in
based
in irrigation
the on
irrigation
the pipe
the
pipe pipe
have
one
pipe
have with
on
with
on a
the
currently a
the generic
irrigation
used
generic
irrigation number
by farmers.
number of
uniformity,
of
uniformity, drippers,
The
drippers,and
andmodel analyzing
proposeis
analyzing
propose an
presented
an the
the effect
intuitive together
effect
intuitive that
irrigation
that the
with
irrigation the
simulations,
transients
transients in
in whose
the
the onpipe parameters
pipe have
have on
on are
the
the basedusedon
irrigation
irrigation real farm data.
uniformity,
uniformity, Theand and propose
propose an
an intuitive irrigation
strategy
simulations,
transients
strategy
based
inwhose
based the on
the
pipe
the
one
havecurrently
parameters
one onarethebased
currently onby
irrigation
used by
farmers.
real farm data.
uniformity,
farmers. The andmodel
propose
model is an intuitive
is presented
presented intuitive together
together
irrigation
with
irrigation
with
Copyrightbased
strategy
simulations,
strategy
simulations, © 2022
based whose
whose on
onThe the Authors.
one
parameters
the one
parameters This
currently
are
currently
areis an
based
based open
used
used on
on access
by
by real article
farmers.
farm
farmers.
real farm under
The
data.
The
data. the CC
model
model BY-NC-ND
is
is presented
presented license
together
together with
with
simulations,
simulations,
1. whose
(https://creativecommons.org
whose
INTRODUCTION parameters
parameters are
are based
/licenses/by-nc-nd/4.0/)
based on
on real
real farm
farm
desirable:data.
data.
simulations, whose parameters are based on real desirable:
1. INTRODUCTION farm data. this effect discourages an irrigation strategy
this effect discourages an irrigation
strategy
based on shorter
desirable: this and discourages
effect more frequent an water
irrigation pulses, which
strategy
1. INTRODUCTION
1. INTRODUCTION
INTRODUCTION based
desirable:
on theonon othershorter
thishand and
effectwould more
discourages frequent
significantly water
an irrigation
irrigation
reduce pulses,
the which
strategy
waste
Irrigation is an essential 1.
1. INTRODUCTIONagricultural practice where a based desirable:
desirable: this
this
shorter effect
effect
and discourages
discourages
more frequent an
an irrigation
water pulses, strategy
strategy
which
Irrigation
certain is
amount an ofessential
1. INTRODUCTION
water is agricultural
artificially practice
applied to where
the soil, a on
of the
desirable:
basedwater onother this
shorter
(Lozanohand
effect
and would
et discourages
more
al. significantly
frequent
(2020)). an
For reduce
irrigation
water
this the
pulses,
reason, waste
strategy
which
it is
Irrigation
certain is anofessential
amount water is agricultural
artificially practice
applied to where
the soil, a on based
based
of theon
water on
onothershorter
shorter
(Lozano and
and
handand more
more
etwould
al. frequent
frequent
significantly
(2020)). water
water
reduce
Forwater
this pulses,
pulses, which
which
the which
reason, waste
it anis
Irrigation
depending is
on an
the essential
crop needs, agricultural
see Oborkhale practice
et al.where
(2015). a based
on the
important othershorter
to hand
consider wouldmore
these frequent
significantly
transients reduce
when pulses,
the
modelling waste
Irrigation
certain
depending
Irrigation
certain is
amount
on
is
amount an
the
an of
ofessential
water
crop
essential
water is
needs,
is agricultural
artificially
see
agriculturalOborkhale
artificially practice
applied et
practice
applied to
toal.where
the soil,
(2015).
where
the soil, a
a on
of
on
of the
water
important
the
water other
other(Lozano
to
(Lozanohand
consider
hand would
et
would
et al.
these
al. significantly
(2020)).
transients
significantly
(2020)). For
For reduce
this
when
reduce
this the
reason,
modelling
the
reason, waste
it
waste
it is
an
is
In most amount
certain
certain
depending farms,
amount
on thetraditional
of water
ofcrop
water issurface
is
needs, see water-saving
artificially
artificially appliedetsystems
applied
Oborkhale to
to the
the
al.the soil,
soil,
(2015). irrigation
are important
of water
of water
water (Lozano system
(Lozano
(Lozano
to consider and
et designing
al.
et these
al. (2020)).
(2020)). a For
transients realistic
For this control
this
when reason, strat-
reason,
modelling it
it anis
is
In most
certain
depending
used, farms,
amount
such on
us traditional
the
dripof water
crop
or surface
is
needs,
sprinkler see water-saving
artificially applied
Oborkhale
irrigation et
systems, systems
toal. are egy.
soil,
(2015).
because irrigation
of
important
Such system
to consider
novel and
et
approach designing
al.
these (2020)).
is a For
transients
carried realistic
this
when
out here.control
reason,
modelling strat-
Moreover,it is
an
depending
In most
used,
depending
In most such on drip
farms,
us
on
farms, the
the crop
traditional
or
crop
traditional needs,
surface
sprinkler
needs,
surface see Oborkhale
water-saving
irrigation
see Oborkhale
water-saving etsystems
systems,
et al.because
al.
systems(2015).
are
(2015).
are important
irrigation
egy. Such
important
irrigation to consider
to consider
system
novel
system and
approach
and these
designing
these
designing is transients
carried
transients when
aa realistic
out
when
realistic here. modelling
control strat-
Moreover,
modelling
control strat-an
an
they
In mostsave more
farms, water or than
traditional alternative
surface solutions,
water-saving seebecause
systems Ohabaare many
irrigation
irrigation of the proposed
system
system and control
designing
and designing
designing strategies
a realistic (Cáceres
control et al.
strat-
In most
In most
used,
they
used,
et al.save farms,
such more
farms,
such
(2012).us
traditional
us dripwater
traditional
drip
In or than surface
sprinkler
surface
sprinkler
sprinkler
water-saving
irrigation
alternative water-saving
irrigation
irrigation,
systems,
solutions,
systems,
water
systems
seebecause
systems
escapes Ohabaare egy.
are
the many
egy.
(2021);
Such
irrigation
SuchofChen novel
the
novelet
approach
proposed
system and
approach
al. control
(2021);
is carried
is carried
Lozoya aa realistic
realistic
strategiesout here.
out
et al. here.
control
(Cáceres
control
(2014);
strat-
Moreover,
et al.
strat-
Moreover,
Giusti
used,
they
et
theyal.
used, such
save
(2012).
such
save us
more
us
more drip
water
In
drip
water or sprinkler
than
sprinkler
or sprinkler
than irrigation
alternative
irrigation,
irrigation
alternative systems,
solutions,
water
systems,
solutions, see
escapes
see because
Ohaba
because
Ohabathe egy.
many
(2021);
egy.
many Such
Suchof
of Chen novel
the
novel
the et approach
proposed
al.
approach
proposed control
(2021);
control is
is carried
Lozoyastrategies
carried
strategiesout
et
outal. here.
(Cáceres
(2014);
here.
(Cáceres Moreover,
et
et al.
Giusti
Moreover, al.
pipes
they
they
et al. through
save
save morerotating
more
(2012). water
water
In nozzles,
than
than
sprinkler with the
alternative
alternative
irrigation, goal ofescapes
solutions,
solutions,
water mimicking
see Ohaba
see Ohaba
Ohabathe and
manyMarsili-Libelli
many
(2021); of
ofChen the et al. (2015)),
proposed
the proposed
proposed control
control
(2021); generate
Lozoya etan
strategies
strategies al.optimal
(2014); control
(Cáceres
(Cáceres et
et al.
Giustial.
al.
pipes
they
et al. through
save more
(2012).
precipitations. rotating
water
In
This nozzles,
than
sprinkler
technique with
alternative
irrigation, the
is useful goal
solutions,
water of mimicking
see
escapes
to irrigate the
larger and
many
(2021);
input Marsili-Libelli
of the
Chen
(namely, et al.
the (2015)),control
(2021);
water flow generate
Lozoyastrategies
out ofetanetanal.
the optimal
(Cáceres
(2014);
emitters) control
et
Giusti
that
et
et al.
al.
pipes (2012).
(2012).
through
precipitations. In
In sprinkler
sprinkler
rotating
This nozzles,
technique irrigation,
irrigation,
with
is the
useful water
water
goal
to ofescapes
escapes
mimicking
irrigate the
the
larger (2021);
and
input Chen
Marsili-Libelli
(namely, et al.
the (2021);
(2015)),
water flow Lozoya
generate
out of al.
the (2014);
optimal
emitters) Giusti
control
that
et
landal.
pipes (2012).
pipes areas
through
through due In sprinkler
rotating
to its
rotating nozzles, irrigation,
with the
wider irrigation
nozzles, with water
the goalgoal escapes
of mimicking
coverage
of mimicking the
(Evans (2021);
and
is
and persistentlyChen
Marsili-Libelli
Marsili-Libelli et al.
varying, (2021);
(2015)),
(2015)),andflow Lozoya
generate
this is not
generate etanal. (2014);
optimal feasible
practically
an optimal Giusti
control
control
pipes
land
pipes through
precipitations.
areas
through
precipitations. due rotating
This
to In
rotating
This its nozzles,
technique
wider
nozzles,
technique with
is the goal
useful
irrigation
with
is the
useful goal
to of mimicking
coverage
to of mimicking
irrigate
irrigate larger
(Evans
larger and
input
is
and
input Marsili-Libelli
(namely,
persistently
Marsili-Libelli
(namely, the
varying, (2015)),
water
the (2015)),
water andflow generate
this out
is not
generate
out of an
the optimal
emitters)
practically
of an
theoptimal
emitters) control
that
feasible
control
that
and King
precipitations.
precipitations.
land areas (2010)).
due This
This
to drip
technique
technique
its wider irrigation,
is
is useful
useful
irrigation on to
tothe other
irrigate
irrigate
coverage hand,
larger
larger
(Evans without
input
input
is significantly
(namely,
(namely,
persistently the
the
varying, increasing
water
water and flow
flow
this the
out
out
is hardware
of
of
not the
the cost.
emitters)
emitters)
practically Indeed,
that
that
feasible
and
land
water King
precipitations.
areas
is (2010)).
due
slowly This
to In
its
supplied drip
technique
wider
to irrigation,
the issoil
useful
irrigation on
throughtothe other
irrigate
coverage
small hand,
larger
(Evans
tubes without
input
is persistently
without significantly
(namely,
the use the
varying,
of increasing
water
expensive and flow
this the
out
is
equipment, hardware
of practically
not thethe cost.
emitters)
only Indeed,
that
feasible
possible
land
and
water
land
and areas
King
is
areas
King due
(2010)).
slowly
due
(2010)). to its
In
supplied
to its
In wider
drip to
wider
drip theirrigation
irrigation,
soil
irrigation
irrigation, on
through
on coverage
the other
small
coverage
the other (Evans
hand,
tubes
(Evans
hand, is
is persistently
without
persistently
without the usevarying,
significantly
of
varying,
significantly expensive and
increasing
and
increasing this
this is
the
equipment,
is
the not
not practically
hardwarethe
practically
hardware cost.
only
cost. feasible
Indeed,
possible
feasible
Indeed,
near
and
and King
water the
King
King crop roots,
(2010)).
(2010)).
is slowly Inthus
In
supplied driptoreducing
drip the soilthe
irrigation,
irrigation, onwater
on
through the loss caused
other
the other
other
small hand,
hand,
tubes without control
withouton thetheusesystem
significantly
significantly is to act
increasing
increasing
of expensive on the
the
the
equipment, valve
hardware
hardwarethe that cost. controls
Indeed,
cost. possible
only Indeed,
near
and
water
by the
is crop
slowly
evaporation, roots,
(2010)). In
supplied
which thus
drip
is to reducing
irrigation,
the
influenced soil the on water
through
by the
wind loss surface
small
and caused
hand,
tubes control
without
without
the water on the the
use
flow system
significantly
toofthe is to whose
increasing
expensive
pipe, act onposition
the
equipment, the valve
hardwarethe is that
cost.
only
by controls
Indeed,
possible
definition
water
near
by
near is
the slowly
crop
evaporation,
water is
the slowly
crop supplied
roots,
which
supplied
roots, thus
is
thus to
to the
reducing
influenced
the
reducing soil
soil through
the
by water
wind
through
the water small
loss
and
small
loss tubes
caused
surface
tubes
caused without
control
the
without
controlwater on
on the
the use
the
flow
use
the toof
system
ofthe
system expensive
is
pipe,
expensive
is to
to equipment,
act
whose on
equipment,
act on the
position
the the
valve
the
valve is only
that
by
only
that possible
controls
definition
possible
controls
runoff,
near theseecrop
nearevaporation,
the Elasbah
crop roots,
roots, etthus
al.is (2019).
thus reducing
reducing Moreover,
the
the water
waterdrip irrigation
loss
loss caused
caused the binary
control
controlwater (either
on
onflow the open
the theor
system
system
to closed).
is
is to
to whoseact
act on on the
onposition valve
the valve
valve that
that controls
controls
by
runoff,
near
by
is theseecrop
evaporation,
typically Elasbah
more
which
roots, etthus
which
efficient,
influenced
al.is (2019).
reducing
influenced
cheaper
by water
Moreover,
the
andby
wind
wind drip
requires
and
loss
andless
surface
irrigation
caused
surface
water binary
control
the water (either
on the
flow open
system
to the orpipe,
closed).
is
pipe, to act
whose the
position
is that
is
by definition
by controls
definition
by
by evaporation,
runoff,
is see
typically
evaporation,
runoff, see Elasbah
more
Elasbah which
et
which
et is
al.
efficient,
is
al. influenced
(2019).
cheaper
influenced
(2019). by
Moreover,
andby
Moreover, wind
wind drip
requires
dripand
and surface
irrigation
less water
surface
irrigation the
binary
In
the water
this
binary water (either
work,
(eitherflow
flow to
open
we
to
open the or pipe,
propose
the or closed).
pipe,
closed). awhose
novel
whose position
model
position is
for
is by
by a definition
drip
definitionirri-
flow
runoff,
runoff,
is to see
be operated.
see
typically Elasbah
Elasbah
more et al.
et al. (2019).
al. (2019).
efficient, (2019).
cheaper Moreover,
Moreover,
and drip irrigation
drip less
requires irrigation
water binary
In
binary (either
this line,work,
(either open
we or
openpropose closed).
or aclosed).
closed). a novel model forThe a drip irri-
flow
runoff,
is to see
be operated.
typically Elasbah
more et
efficient, cheaper Moreover,
and drip
requires irrigation
less water gation
binary
In this (either
work, placed
open
we in
or
propose tunnel a greenhouse.
novel model for a model
drip is
irri-
is
is typically
typically
flow to be more
more
operated. efficient,
efficient, cheaper
cheaper and requires
andagricultural
requires less less water
lesssector, gation
In this
water characterized line,
this work,work, byplaced
wethree in
proposea tunnel
main greenhouse.
a features
novel model model
that make The
for aa it model
drip is
irri-
a good
According
is typically
flow to be to
more Water
operated. (2018),
efficient, cheaperin the and requires water In
In this
gation work,
line, we
we
placed propose
propose
in a tunnel a
a novel
novel model
greenhouse. for
forThe a drip
drip
model irri-
irri-
is
flow
flow to
According
to be
be operated.
to Water
operated. (2018), in the agricultural sector, characterized
In this
gation work,
line,
representation by
we
placed three
propose
of three
a ininrealamain
tunnel a features
novel
life features
scenario that
model
greenhouse. make
for
comprising: The a it
dripa
model
(1) good
irri-
theis
irrigation
flow to beis responsible of 70% in of thethetotal water consump- gation line, placed a tunnel greenhouse. The model is
According
irrigation
According
tion. isoperated.
to
to
Additionally,
Water (2018),
responsible
Water in of 70%countries,
(2018),
many of the
in the
agricultural
total wateravailability
agricultural
water consump-
sector,
gation
sector, characterized
gation
characterized
dynamics
line, placed
representation
line, of
placed
by
by
theof three
a in
water
in aamain
real tunnel
life
tunnel
main
in the
greenhouse.
scenario
greenhouse.
features
soil,
that
that
inspired
make
comprising:
makeThe it
The
from
model
aa good
(1)
itmodel the
good
Cáceres
is
is
According
irrigation
tion. is to Water
responsible
Additionally, in (2018),
of
many 70% in
of thetotal
the
countries, agricultural
water
water consump-
availability characterized
characterized
sector, representation
dynamics of by
by
theof three
three
aa real
water main
main
in life the features
features
scenario
soil, that
that make
make
comprising:
inspired from it a
(1) good
itCáceres
a good
good
the
According
irrigation is to Water
responsible (2018),
of 70% in
of the
the agricultural
total water sector,
consump- characterized
representation
et al. (2021), byof
(2)of three real main life features
scenario that make
comprising: it a
(1) the
is limited
irrigation
irrigation
tion. is(De
is Fraiture
responsible
responsible
Additionally, ofand
of
in many 70% Wichelns
of
70%countries,the
of the total
the total
total (2010)),
water therefore dynamics
consump-
wateravailability
water consump- representation
representation
et al.binary
(2021), of the ofthe
(2)of a realdynamics
aaof real
water
the real life
life
inlife
dynamics
of
scenario
scenario
thescenario
soil,
of
theinspired
water infrom
comprising:
comprising:
theinspired
water
the pipe,
(1)
(1) the (3)
the
the
Cáceres
is limited
irrigation
tion.
tion.
is
is(De
Additionally,
policymakers
Additionally,
limited (De
Fraiture
responsible
around in
in
Fraiture
of
many
the
many
and
and
70%
world Wichelns
of
countries,
are creating
countries,
Wichelns
(2010)),
water
water
water
(2010)),
therefore
consump-
availability
strategies
availability
therefore to the representation
dynamics
dynamics
et al. (2021), of
of the
nature
the
(2) water
water
the thein control
in
dynamics the
the
soil,
input.
soil,
soil,
of the The in
comprising:
inspired
inspired
water in
the pipe,
from
from
from
the
(1)
Cáceres
representation
Cáceres
Cáceres
pipe,
(3)
(3)
policymakers
tion.
is Additionally,
limited
achieve (De around
an efficient in
Fraiture
waterthe
manyandworld
usage are
countries,
Wichelns
(EC creatingwater
(2010)),
(2019a); strategies
EC availability
therefore
(2019b)), to the
et
of al.
thebinary
dynamics (2021), nature
of
valvenaturethe
(2) the
position of
water the
dynamicscontrol
in
as control the
a binary of input.
soil,the
state The
inspired
water inrepresentation
from
is possible the Cáceres
pipe,
thanks(3)
is
is limited
limited
policymakers
achieve an (De
(De Fraiture
Fraiture
around
efficient waterthe and
andworld
usage Wichelns
Wichelns
are
(EC (2010)),
(2010)),
creating
(2019a); EC therefore
therefore
strategies
(2019b)), to et
et
the
of al.
al.
the (2021),
(2021),
binaryvalve (2)
(2) the
the
position of dynamics
dynamics
the
as a binary of
of the
the
input.
statewater
water
The
is in
in the
the pipe,
pipe,
representation
possible thanks(3)
(3)
is limited
policymakers
and specialized(De Fraiture
around
agencies the andworld Wichelns
are (2010)),
creating therefore
strategies to et
the
to al.
a (2021),
binary
hybrid (2)
nature the of
dynamical dynamics
the control
systems of the
input. water
The
formulation, in the pipe,
representation
combining (3)
policymakers
achieve
and an
specialized
policymakers
achieve an around
efficient
efficient water
agencies
around waterthesuch
the world
usage
such
world
usage
as(EC
as
FAO
are
FAO
are
(EC
(2022)EC
creating
(2019a);
(2022)
creating
(2019a);
promote
strategies
(2019b)),
promote
strategies
EC (2019b)),
the
the to of
to the
to
the
of the
a
thebinary
binaryvalve
hybrid
valve nature
position
nature
position of the
dynamical
of the
as
as control
aasystems
controlbinary
binary input.
state
input.
state The
is
formulation,
The
is representation
possible
possible thanks
combining
representation
thanks
use of
achieve
achieve
and modern
an
an efficient
specialized technologies
efficient
efficient water
water such
agencies usage
usageto asreduce
(EC
(EC
FAO the
(2019a);
(2019a);
(2022) water
EC
EC consump-
(2019b)),
(2019b)),
promote the continuous-time
of the
of the
to the
a hybridvalve
valve position and discrete-time
position
position
dynamical as
as aaasystemsbinary evolution
state
binary formulation,
state is in the
possible
is possible
possible follow-
thanks
thanks
combining
use
and
tion of
achieve modern
an
specialized
and, at the technologies
water
agencies
same time, usage
such to asreduce
increase(EC
FAO the the
(2019a);
(2022)
crop water
EC consump-
(2019b)),
promote
yield. the
Indeed, continuous-time
of
to
ing aaform valve
hybrid Goebel and
dynamical
et discrete-time
al.as binary
systems
(2012) evolution
state is
formulation, in the follow-
thanks
combining
and
and
use
tion specialized
specialized
of modern
and, at the agencies
agencies
technologies
same time, such
such to as
as FAO
FAO
reduce
increase the (2022)
(2022)
thecrop waterpromote
promote
yield. consump- the
the
Indeed, to
ing hybrid
continuous-time
form Goebel dynamical
and
et al. systems
systems
discrete-time
(2012) formulation,
formulation,
evolution in combining
combining
the follow-
and
use of
the
use specialized
of modernirrigation
classical
modern agencies
technologies
technologies such
strategy as
to reduce
to FAO
reduce
used (2022)
the
bycrop
the promote
water consump-
farmers
water consump- the
consists to a hybrid
continuous-time and
continuous-time dynamical
anddiscrete-time systems
discrete-time evolution formulation,
evolution in combining
in the
the follow-
follow-
use
tion
the
use
tion of modern
and,
classical
of at
modern
and, at the technologies
same
irrigation
the time,
technologies
same strategy
time, to reduce
to reduce
increase
increase usedthe
the the
by
thecrop water
yield.
farmers
water
yield. consump-
Indeed,
consists
consump-
Indeed, continuous-time
ing
ing form
form Goebel
continuous-time Goebel and and
et
et al. discrete-time
discrete-time
al.ẋ(2012)
= f (x),
(2012) evolution
x ∈ C,
evolution in the
in the follow-
follow-
in
tion
tion
the turning
and,
and,
classicalat
at on
the
the the
same
same
irrigationirrigation
time,
time,
strategy systems
increase
increase usedthe
the byonce
crop
crop or
yield.
yield.
farmers twice per
Indeed,
Indeed,
consists ing
ing form
form Goebel
Goebel Het
et : al.

al.ẋ (2012)
=
(2012) f (x), x ∈ C, (1)
in turning on the irrigation systems once or twice per ing form GoebelHet: al.  ẋ(2012)
tion and, +
the
day,
the
in with at
classical
classical
turning verythe
on
samepulses,
irrigation
long
irrigation
the
time,
strategy
strategy
irrigation
increase
resultingused
used
systems
theinby
by
crop
once
yield.
a farmers
good
farmers
or crop Indeed,
consists
yield
consists
twice per

 x
x
ẋ +==
= =
g(x), x
f (x),
f (x),
g(x),
x∈
x ∈ C,
∈
D,
C,
D,
(1)
day,
the
in
but with
classical
turning
a very
on
significant long
irrigation
the pulses,
strategy
irrigation
waste of resultingused
systems
water inby
(Morillo a farmers
once good
etor crop
al.twice yield
consists
per
(2015)). H
H :: ẋẋ +== ff (x),
(x), x
x ∈
∈ C,
C, (1)
in
day,
but turning
in turning
turning
awith on
on long
very
significant the irrigation
thewaste
irrigation
pulses, systems
systems
resulting
of water once
once
inonce
(Morillo a good or twice
oral.crop
etor twice per
per where f , g are, respectively,
yield
(2015)). H
H ::: x x
ẋ++==
+ = g(x),
f (x), the flow x ∈
∈ D,
C, jump map, and
and
D, (1)
(1)
in
day,
A with
different on
very the
long
irrigation irrigation
pulses,
strategy systems
resulting
involvingin a good
shorter twice
crop
pulses per
yield is where f , g are, H x
respectively, + = g(x), g(x),
the x
x
flow ∈ D,
and jump map, (1)
and
day,
but
A awith
different very
significant long
irrigation pulses,
waste of
strategy resulting
water involvingin
(Morillo a good
et
shorter crop
al. yield
(2015)).
pulses is C, D are the flow and x
x + = g(x),
jump = set,
g(x), x
where
x ∈
∈ D,
solutions
D, may flow
day,
but with
a
difficult
but a tovery
significant
propose
significant long pulses,
waste
without
waste of
of resulting
water
affecting
water in
(Morillo
the
(Morillo a good
et
uniformity
et crop
al.
al. ofyield
(2015)).
the
(2015)). where
C, D
where aref
f ,
, g
the
g are,
flow
are, respectively,
and jump
respectively, the
set,
the flow
where
flow and jump
solutions
and jump map,
may
map, and
flow
and
butdifferent
A aa distribution.
difficult
but significant
irrigation
to propose
significant waste
without
waste of water
strategy
of water
affecting (Morillo
involving et al.
shorter
the uniformity
(Morillo et al. pulses
(2015)).
of the
(2015)). is or jump,
where
where respectively.
ff ,, the
gg are, respectively, the flow and jump map, and
A
A different
water
different
difficult
water to
irrigation
irrigation
propose
distribution.
Thisstrategy
is mostly
strategy
without
This is
involving
because
involving
affecting
mostly the
because
shorter
of
shorterthe pulses
uniformity
of the
effect
pulses
of
effect the is C,
is
of
of
or
C,
D
jump,
where
D
are
are g are,
flow
flow
respectively,
respectively.
f , the are, and jump set,
respectively,
and jump
the
set,
flow
flow and
the where
where and jump
jump map,
solutions
solutions
may flow
map,
may
and
and
flow
A different
difficult
the sloping
difficult to
to irrigation
propose
areas
propose on strategy
without
the
without pipe waterinvolving
affecting
affecting the
distribution
the shorter
uniformity
uniformity pulses
of
uniformity
of the
the is C,
or
The
C, D
jump,
D
or jump, are
rest
jump,are the
of
the flow
respectively.
this
flow and
article
respectively.and jump
is
jump set,
organized
set, where
whereas solutions
follows:
solutions in may
Section
may flow
flow 2
difficult
water
the sloping
difficult
water to
to propose
distribution.
areas
propose
distribution. on without
This
the
without
This is
pipe
is affecting
mostly
water
affecting
mostly the
because uniformity
distribution of
the uniformity
because of the
the of
effect
uniformity the
ofwith
effect the of
of or
The
or rest
jump, respectively.
of this article
respectively. is organized as follows: in Section 2
(Ella
water
water
the et al. (2009)).
distribution.
distribution.
sloping areas on Indeed,
This
This
the is
is
pipe the
mostly
mostly
water transient
because
because
distribution associated
of
of the
the effect
effect
uniformity of
of we
or recall
jump, the dynamics
respectively. of the water in the soil, discussing
(Ella
water et
the sloping
the sloping
pipe al. (2009)).
distribution.
areas on
discharge Indeed,
This
on process
the pipeis the
mostly
pipewhen transient
because
waterirrigation
distribution associated
of the
is turned with
effect
uniformity of
off, The The
we
some rest
recall
rest of
the
of
properties this
this article
dynamics
article
that allow is
isof organized
the
organized water
the as
forwater asinfollows:
the soil,in
follows:implementa-
practical in Section
discussing
Section 2
22
the
(Ellasloping
pipeet al. areas
areas
(2009)).
discharge the
on process
the
Indeed,pipewhen water
water
the distribution
distribution
transient
irrigation uniformity
uniformity
associated
is turned with
off, we The
The
we
some rest
rest
recall of
of
the
properties this
this article
article
dynamics
that is
is
allowoforganized
organized
the forwater
the as
asinfollows:
follows:
the soil,
practical soil,in
in Section
Section
discussing
implementa- 2
the
(Ella
is sloping
known
(Ella et
et al.
to
al. areas
(2009)). on
generate
(2009)). the
Indeed,pipe
uneven
Indeed, water
the
water
the distribution
transient
distributions
transient uniformity
associated
associated with
because
with The
tion
we rest
recall
of
recall our of
the
the this article
dynamics
control
dynamics isof
algorithm;
of organized
the
the in
water asin
Section
infollows:
the
the 3 we
soil,in Section
discussing
present
discussing the 2
(Ella
the
is
(Ellapipe
known et al.
etgravity
al. (2009)).
discharge
to generate
(2009)). Indeed,
process
uneven
Indeed, the
whenwater
the transient
irrigation associated
distributions
transient is turned
associated with
becauseoff,
with we
some
tion
we
some recall
of
recall our the
properties
the
properties dynamics
controlthat
dynamics
that allowof
algorithm;
allowof the
the for
for water
inthe
water
the in
Section
in the
practical
the
practical 3 soil,
we
soil, discussing
implementa-
present
discussing
implementa- the
the
of
the pipe
the discharge
pipegravity
discharge process
action. This
process when
when irrigation
phenomenon
irrigation is turned
becomes
is turned
turned off,
more
off, tion dynamical
some
some properties
ofproperties equations
that
that for
allow
allow the water
for the in the
practical pipe and for
implementa- the
is
of
the
is known
the
pipe
known
noticeable
to
to generate
discharge action.
generate
when
uneven
theprocessThis
uneven
water
water
phenomenon
whenwater
distribution
distributions
irrigation becomes
distributionsis
difference
because
more
because
becomes off, tion
dynamical
some
valve of our equations
properties
our
position,
control
controlthat
then
algorithm;
for
allow
algorithm;
we thefor
complete for inthe
water
inthe practical
Section
in the
practical
Section
the hybrid
weimplementa-
333pipe
we present
and for the
implementa-
present
formulation
the
is
is
of known
known
the
noticeable to
to
gravity generate
generate
when action.
the uneven
uneven
This
water water
water
phenomenon
distribution distributions
distributions
becomes
difference because
because
more
becomes tion
tion of
of
dynamical
valve our
our
position, control
control
equations
then algorithm;
algorithm;
for
we the
complete in
in
water Section
Section
in
the the 3
hybrid we
we
pipe present
present
and for
formulation the
the
the
is
of known
the
comparable to with
gravity generate
action. uneven
thewaterThis
amount water
phenomenon
of waterdistributions
becomes
delivered because
more
during tion of our equations
dynamical
introducing
dynamical control
an intuitive
equations algorithm;
for
for the
control
the instrategy,
water
water Section
in
in the
the 3pipe
which we present
pipe and for
is inspired
and for the
the
of
of the
noticeable
comparable
the
noticeable gravity
when
gravity with
when action.
the
the
action.
the This
amount
This
water phenomenon
distribution
of water
phenomenon
distribution becomes
difference
delivered
becomes
difference more
becomes
during
more
becomes dynamical
valve position,
introducing
dynamical
valve position, equations
an then
intuitive
equations
then for
we
for
we the
complete
control
the
complete water
water in
the
strategy,
in
the the
hybrid
the pipe
which
hybridpipe and
is
and for
formulation
inspired
for
formulation the
the
irrigation.
noticeable
noticeable
comparable Such
when
when
with non-uniformity
the
the
the water
water
amount in
distribution
distribution
of the
water crop
difference
differenceyields
delivered is
becomes
becomes not
during from
valve
valve the
position,
position,“practical”
then
then strategy
we
we complete
complete used the
theby farmers;
hybrid
hybrid lastly,
formulation
formulation in
irrigation.
noticeable
comparable Such
when
with non-uniformity
the
the water
amount in
distribution
of the
water crop
differenceyields
delivered is
becomes not
during introducing
from
valve the
position,
introducing an
an intuitive
“practical”
then
intuitive we control
strategy
complete
control strategy,
used the
strategy,by which
farmers;
hybrid
which is inspired
lastly,
formulation
is inspired in
comparable
irrigation. with
Such the amount
non-uniformity ofin water
the delivered
crop yields during
is not introducing
introducing
from an
an
the “practical” intuitive
intuitive
“practical” control
control
strategy strategy,
strategy,
used which
which
by farmers;
farmers; is
is inspired
inspired
lastly, in
comparable
irrigation.
irrigation. with
Such
Such the amount
non-uniformity ofin water
the delivered
crop yields during
is not introducing
from the an intuitive control
strategy strategy,
used by which is inspired
lastly, in
irrigation. Such
irrigation. Such non-uniformity
non-uniformity in in the
in the
the cropcrop yields
crop yields
yields is is not from
not
is not from
fromthethe
the “practical”
“practical” strategystrategy used
used
used by
by farmers;
by farmers;
farmers; lastly,
lastly, in
lastly, in
2405-8963 Copyrightnon-uniformity
© 2022 The Authors. This is an open access article under theCC “practical”
BY-NC-ND license strategy . in
Peer review under responsibility of International Federation of Automatic Control.
10.1016/j.ifacol.2022.11.137
 R. Bertollo et al. / IFAC PapersOnLine 55-32 (2022) 188–193 189

Section 4, we present some numerical simulations based  B+3

θi
on the parameters of a real farm, to show the behaviour Hi (θi ) := Ki (θi ) ψi (θi ) = Ksat ψsat ,
of our model. θsat
Di + Di+1
Di := , (3)
2. WATER DYNAMICS IN THE SOIL 2
with constants B, Ksat , ψsat , θsat being empirical param-
2.1 Physical model and discretization eters depending on the soil composition. Note that, for
compact notation, in the definition of Qi,i+1 we some-
In order to describe the differential equations that charac- times removed the dependence on θ of H and ψ, which is
terize the water dynamics in the soil, according to Cáceres highlighted in their respective definitions. Also note that,
et al. (2021) and references therein, for each crop, the soil as we are considering irrigation in a greenhouse, in the
is divided into N + 1 layers, as shown in the sketch at simulations of Section 4 both the precipitations P and the
the right of Figure 1. The first layer is referred to as the surface evaporation E are set to zero.
surface layer, which is affected directly by precipitations To conclude, the dynamics in (2) can be written in a
and by the irrigation; the layers from 2 to N characterize compact way as follows
the root zone, where the water is actually absorbed by the    
plant roots; the last layer N + 1 is the drainage zone, used θ̇1 f1 (θ1 , θ2 , u, t)
to keep track of the overflowing water, which should be  θ̇2   f2 (θ1 , θ2 , θ3 , t) 
 
minimized to avoid wasting resources. Θ̇ :=  .  = f (Θ, u, t) := 
 ..  , (4)

 . . .
θ̇N +1 fN +1 (θN , θN +1 )
where the dependence on t is through the external inputs
P , E and Eroot

2.2 Choice of the parameter N (monotonicity)

The model parameters to be selected are the layer thick-

nesses Di , which are directly related to the number of
layers N , since the soil depth is a given, constant value. A
large number of layers N (or, which is the same, a small
value of the layer thicknesses Di ) increases the accuracy
of the model in approximating the original continuous
version in Sellers et al. (1996). On the other hand, practical
feasibility suffers from the choice of a high number of
Fig. 1. Graphical representation of the irrigation line: view layers, which would require a higher number of sensors,
from above (left diagram), view from the side (middle to be placed at small distances; thus, a trade-off is needed.
diagram) and sketch of the soil layers (right diagram).
An acceptable value of Di in this sense can be selected as
According to Cáceres et al. (2021), the dynamics of the the maximum value such that the following assumption is
soil moisture θi , i ∈ {1, . . . , N + 1}, for each layer, can be satisfied.
expressed though the following differential equations for Assumption 1. For all i ∈ {1, . . . , N }, Di and Di+1 are
each m2 of field chosen so that ∂Qi,i+1 /∂θi > 0 and ∂Qi,i+1 /∂θi+1 < 0
1 for all the values in the considered soil moisture interval
θ̇1 = f1 (θ1 , θ2 , u) := (P (t) − Q1,2 (θ1 , θ2 ) − E(t) + u)
D1 [θmin , θmax ].
θ̇i = fi (θi−1 , θi , θi+1 ) i ∈ {2, . . . , N }
Remark 1. As we are dealing with an approximation of
1 the theoretical model, it makes sense to select a portion of
:= (Qi−1,i (θi−1 , θi ) − Qi,i+1 (θi , θi+1 ) − Eroot (t))
Di the state space where such an approximation needs to be
θ̇N +1 = fN +1 (θN , θN +1 ) accurate, as we did in Assumption 1. The lower limit θmin
of the interval corresponds to the soil moisture at which
1 the plant will die, while the upper limit θmax corresponds
:= (QN,N +1 (θN , θN +1 ) − KN +1 (θN +1 )) ,
DN +1 to the soil saturation limit for the specific case study, i.e.
(2) the maximum value of soil moisture before water starts
where constant Di is the thickness of the i-th layer, the overflowing. ◦
input P represents the precipitations, the input E repre-
sents the water evaporation at the surface, Eroot represents Assumption 1 holds in the continuous version of the model
the water absorbed by the roots, u is the irrigation input, Sellers et al. (1996), and such a requirement makes sense
and the water flow Qi,i+1 between consecutive soil layers intuitively; indeed, we expect the water flow to increase
is computed as when the soil moisture in the upper layer increases, and
  the converse to happen with the soil moisture in the lower
B 1 1 layer.
Qi,i+1 (θi , θi+1 ) := + (Hi − Hi+1 ) ,
B + 3 Di ψi+1 − ψi To this end, we split the soil into equal layers (except
 2B+3
θi for the first one, i.e. the surface layer) and we represent
Ki (θi ) := Ksat , the surface Qi,i+1 (θi , θi+1 ), for increasing values of D =
θsat
 −B Di = Di+1 . From the graphical analysis of the surfaces (see
θi Figure 2 for some examples), we select the value D = 3cm,
ψi (θi ) := ψsat ,
θsat which satisfies Assumption 1.
190 R. Bertollo et al. / IFAC PapersOnLine 55-32 (2022) 188–193

3. HYBRID DYNAMICAL MODEL

3.1 Water dynamics in the pipe

When we consider a set of M crops sequentially placed

along an irrigation line, we must take into account the
dynamics of the water in the pipe. In particular, we can
assume a very fast filling, due to the high water pressure,
Fig. 2. Graphical representation of the water flow surface, while the emptying process is slower and non uniform, due
for D = 9cm (left) and D = 3cm (right). to the difference in height among the crops along the same
irrigation line (see the diagram in the middle of Figure 1).
According to these observations, we choose to model the
pipe filling with an instantaneous change of the flow
jumping to its steady-state value. Conversely, the pipe
emptying can be modeled introducing a state representing
the residual volume of water still contained in the pipe (or,
which is the same up to rescaling, the length of the water
column measured along the pipe axis ). This value can
be retrieved considering the whole pipe-crops system as a
compartmental model, which gives
t
j Qj (τ )dτ
(t) = L − 0 , (5)
πr2
where L is the length of the pipe, r is its radius, and Qj is
the volumetric flow out of the j-th dripper. The value of
Fig. 3. Detail of the simulation in Section 4: the order of the flow Qi out of a dripper as a function of the (relative)
the soil moistures is not the same during the whole pressure pi in the pipe can be modelled using the equation
evolution of the system. Qj = kpzj ,
where k is a constant given by the dripper type, and z is
an empirical exponent whose value goes from about 0 (in
In view of Assumption 1, we can derive the following state- the case of auto-compensated drippers) to 0.5 (in the case
ments, whose proof is omitted due to space constraints. of non auto-compensated drippers).
Proposition 1. Under Assumption 1, each constant input Given the very slow discharge velocities in this application,
u(t, j) = u corresponds to a unique equilibrium configu- we can consider the emptying process as a quasi-static
ration Θeq (u) = (θ1,eq (u), θ2,eq (u), . . . , θN +1,eq (u)). More- phenomenon, therefore Bernoulli’s principle gives
over, if u2 ≥ u1 then θi,eq (u2 ) ≥ θi,eq (u1 ) for all i ∈ H
{1, . . . , N + 1}. pj = ρw g max{0, ( − j )} ,
L
Remark 2. Proposition 1 ensures that, if the crops along where ρw is the water density, g is the gravity acceleration,
an irrigation line receive decreasing amounts of water, then j is the position of the dripper, measured along the pipe
the average soil moisture decreases too. This fact can be axis, and H is the height difference between the starting
intuitively confirmed, observing that the crops at the top and the ending point of the pipe. Combining the equations
of an irrigation line are drier than the ones at the bottom. above, and differentiating (5), we obtain the dynamics of
◦ the state : 
Q
Remark 3. Selecting N sufficiently large, as suggested in ˙ = f () := − j j
this section, raises issues concerning the practical feasi- 2
πr z 
bility of the control scheme. Indeed, measuring all the k ρw gH z
states θi would require a significant increase in expenses. =− 2 (max {0, ( − j )}) .
πr L
However, given the compartmental nature of this model, j
the soil moisture of the first and last layers of the root (6)
zone are reasonable upper and lower bounds for the soil Summing up, the dynamics of the state  depends on
moisture level of all the layers between them. This is the valve position, and is represented by the following
confirmed by the simulations, see Section 4, and together equations
with the monotonicity property shown in Proposition 1 it
may allow us to use only two sensors for each irrigation ˙ = f (), with (OFF),
line, which is reasonable. ˙ = 0, with (ON), (7)
+ = L, with (OFF → ON),
Moreover, the monotonicity property along the irrigation +
 = , with (ON → OFF).
line, shown in Proposition 1, suggests the possibility to
exploit some similar monotonicity property across the The (hybrid) dynamics in (7) can be written in a compact
layers in the soil. However, this property is disproven by form by introducing a logic state q, representing the
simulations, as shown in Figure 3, depicting a zoomed position of the valve. More precisely, we assign q = 0 when
portion of the simulation in Section 4, where we can see the valve is closed and q = 1 when it is open. Using this
that the order of the soil moistures is reversed along the notation, we can rewrite (7) as
evolution of the system. This result highlights the need of ˙ = (1 − q)f (), with (valve still)
different mathematical tools, and it will be investigated in
future works. ◦ + = L, with (valve switch).
 R. Bertollo et al. / IFAC PapersOnLine 55-32 (2022) 188–193 191

Note that we simplified the jump map, since the pipe is τ̇ = ρ1 τ̇ = ρ1

always full while the valve is open, so in the ON-OFF 1 − γ2
transition we get + =  = L.
1 − γ1
Using the value of , we can also (algebraically) compute
the value of the input uj () to the j-th crop, using the t t
value of the volumetric flow Qj out of the corresponding τ̇ = ρ2 τ̇ = ρ2
sprinkler. Since u is a linear flow in the vertical direction, 1 − γ2
averaged over a certain area, we must divide Qj by the 1 − γ1
area W ∆L that each sprinkler is irrigating, where W is
the tunnel width and ∆L is the distance between two t t
consecutive sprinklers, measured along the pipe axis. This
gives Fig. 4. Evolution of the timer τc (black, thin line), and the
 z corresponding value of q (blue/red, heavy line) given
Qj k ρw gH
uj () = = max {0,  − j } . (8) by (9): γ = γ1 > γ2 (left plots) generates a signal
W ∆L W ∆L L with a larger mean value; ρ = ρ1 > ρ2 (upper plots)
3.2 Proposed control strategy generates faster switching.

Irrigation Θ1 , . . . , ΘM
Since the control input is binary (valve open or closed),
system
the control strategy corresponds to choosing the switching
logic. From Proposition 1, it is clear that increasing the q
total amount of opening time of the valve results in a
switching ρ, γ
higher soil moisture. On the other hand, we know that g ρ , gγ
frequent switching generates a non-uniform distribution logic
of water among the crops due to the pipe depletion.
Fig. 5. Block diagram representation of the closed-loop
According to these facts, we propose here an intuitive
system.
control algorithm, consisting in a periodic switching (to be
activated only at daytime, when the water absorption of automated context, i.e. providing pulses of water once or
the roots Eroot is non-zero) regulated by two parameters, twice per day. While the final goal of a smart irrigation
γ and ρ. In particular, the switching logic can be described system is to optimize the pulse distribution in order to
using the following hybrid formulation reduce the waste of resources, the feedback law proposed

τ̇c = ρ, in this work only aims at showing the behaviour of the
with τc ∈ [0, 1], improved model of the irrigation line, without any claim
q̇ = 0,
 + of (9) being the optimal control strategy. The study of an
τc = 0, improved control law will be the focus of future work. ◦
with τc = 1, (9)
q + = q,
 + 3.3 Complete hybrid model
τc = τc , with (τc ∈ [0, γ] and q = 0)
q + = 1 − q, or (τc ∈ [γ, 1] and q = 1) .
The complete hybrid dynamics of the feedback system of
In this hybrid dynamics, τc is a timer state that continu- Figure 5 is as follows. The flow dynamics stems directly
ously evolves with derivative ρ and is reset to zero when from (4), (7), (8) and (9), and is given by
its value reaches 1. The valve is opened when the value of    
the timer becomes larger than the threshold 1 − γ. Thus, Θ̇1 f (Θ1 , u1 (), t)
practically speaking, ρ > 0 regulates the switching rate  ..   .. 
 .   . 
and consequently the irrigation uniformity, while γ ∈ [0, 1]    
regulates the average opening time, in a PWM fashion.  Θ̇M   f (Θ , u (), t) 
  
M M
Some examples of this hybrid dynamics are depicted in  ˙   (1 − q) f ()  
   0 
Figure 4, which clearly shows that a larger value of γ re-  q̇   ,
ẋ :=   = f (x, t) :=  0  x ∈ C,
sults in a larger percentage of time when the valve is open,  ċ   
whereas a larger value of ρ generates faster switching.  ρ̇   0 
   
 γ̇   0 
In view of the considerations above, the inputs γ and    cρ 
 τ̇c   
ρ can be selected based on some feedback quantity, in τ̇  1/60
day
order to reach the desired distribution of soil moisture. τ̇a 1/60
Given the nature of this process, the (mean) effect of
changes in γ and ρ on the soil moisture will take some (10)
time to be observed, thus the solution that we propose where the flow set C will be precisely characterized in
is to to keep γ and ρ constant during flows, and then the following, and we introduced three additional states
periodically update them (in our simulations we perform (c, τday , τa ) to the description.
updates every two days) according to some functions
gγ (Θ1 , . . . , ΘM ), gρ (Θ1 , . . . , ΘM ), where Θj denotes the The timer states τday and τa in (10) keep track, respec-
vector of the soil moistures at the j-th crop, as defined tively, of the hour of the day (in order to activate the
in (4). control scheme) and of the elapsed time since the last
adaptation of γ and ρ. The third state c is another logic
A block diagram of the complete closed-loop system is state, which is set to 1 when the control algorithm is
depicted in Figure 5. activated (i.e., at certain daytime hours), and to 0 when
Remark 4. The proposed control algorithm is inspired by the controller is deactivated. Note that in (10) the flow
the irrigation strategy used by the farmers in a non map of τc contains this new state c, so that when c = 0
192 R. Bertollo et al. / IFAC PapersOnLine 55-32 (2022) 188–193

the control timer is frozen and the switching is deactivated,

as desired.
Concerning the jump dynamics, we need to define different
maps and sets to describe
(1) the valve switch
(2) the reset of τc when it reaches 1
(3) the reset of τday at the end of the day
(4) the adaptation of γ and ρ
(5) the control activation/deactivation
We now explicitly characterize each one of these maps
g1 (x), . . . , g5 (x), with the corresponding jump sets. Since
they only affect some of the states, for compact notation
we only report the jump dynamics of the affected states,
while assuming that the other ones remain unchanged
across the corresponding jump
 +   Fig. 6. Soil moisture evolution in the layers of three crops:
l L bottom crop (top-left plot), middle crop (top-right
= g1 (x) := 1 − q , x ∈ D1 , (11a)
q+ plot) and upper crop (bottom-left plot).
[τc+ ] = g2 (x) := [0] , x ∈ D2 , (11b) 4. SIMULATION RESULTS
 + 
τday = g3 (x) := [0] , x ∈ D3 , (11c)
 +   In this section we present the results of simulations 2
ρ gρ (Θ1 , . . . , ΘM ) performed on the proposed model. For our simulations,
γ + = g4 (x) := gγ (Θ1 , . . . , ΘM ) , x ∈ D4 , (11d) we selected the adaptation laws for γ and ρ to be
τa+ 0  t
 +   
c 1−c gγ (Θ1 , . . . , ΘM ) := −kγ (ΘM,N (τ ) − µ)dτ,
= g5 (x) := , x ∈ D5 . (11e)
τc+ 0 t−τa
gρ (Θ1 , . . . , ΘM ) := −kρ max{0, Γ(t)}
The corresponding jump sets are given by  t 
D1 := {x ∈ X : (τc ≤ (1 − γ) and q = 1) Θ1,2 (τ ) + Θ1,N (τ )
Γ(t) :=
or (τc ≥ (1 − γ) and q = 0)}, (12a) t−τa 2

D2 := {x ∈ X : τc = 1}, (12b) ΘM,2 (τ ) + ΘM,N (τ )
− − σ dτ,
D3 := {x ∈ X : τday = 24}, (12c) 2
D4 := {x ∈ X : τa = Tadapt }, (12d) where Θj,i represents the soil moisture θi of the j-th crop,
µ is the desired average value for the soil moisture of the
D5 := {x ∈ X : (τday ∈ [hON , hOFF ] and c = 0) N -th layer of the highest crop, and σ is the tolerated
or (τday ∈ [0, hON ] ∪ [hOFF , 24] and c = 1)}. (12e) mismatch between the average soil moistures of the first
In (12), hON , hOFF and Tadapt are tuning parameters, and last crop.
corresponding to the hour of the day when the irrigation All the numerical values with the corresponding measure-
begins and ends, respectively, together with the amount ment units are reported in Table 1, with the exception
of hours between two consecutive adaptations of γ and ρ. of Eroot , which is a time varying signal; the plot of its
Lastly, the complete state space X is defined as evolution during the day is not reported due to space con-
X := R(N +1)M × [0, L] × {0, 1}2 × R>0 straints, but it can be retrieved from the typical values of
(13) the evapo-transpiration in this application, see for example
×[0, 1]2 × [0, 24] × [0, Tadapt ]. (Cáceres et al., 2021, Figure 6), rescaling them according
Using all the definitions in this section, we can characterize to the surface W ∆L considered here.
the hybrid system in the form (1), where the flow map f The simulation spans a period of 20 days, and the re-
is defined in (10), the jump map is defined in terms of the sults are depicted in Figure 6, for some of the crops. In
graphs of the individual jump maps 1 in (11), namely particular, we choose to report the evolution of the soil

gph g(x) := gph gi (x), (14) moisture for 3 crops, located respectively at the bottom,
in the middle and at the top of the irrigation line. Even
i∈{1,...,5}
if the average value of the soil moisture of the last crop
the flow set C is given by the closure of the complement of is not yet settled, we choose to show this portion of the
D simulation, to appreciate the spikes due to the irrigation
C := X \ D, (15) in the single days.
, and the jump set D is defined as the union of the jump From Figure 6, we can clearly see that the water dis-
sets in (12), namely tribution behaves as expected, and the lower crops in
 the irrigation line are characterized by larger average soil
D := Di . (16)
moisture. We can also observe that the initial values for
i∈{1,...,5} γ, ρ cause a transient increase of the water content of the
1 As the intersection among some of the jump sets in (12) is non- last crop above the desired value, but then the feedback
empty, strictly speaking the result of this operation would be a set- (9) effectively reduces γ and consequently the average soil
valued map, generating a hybrid inclusion. However, in view of the
practical focus of this paper, we used this slight abuse of notation to 2 The simulations have been performed using the Matlab HyEQ

avoid complicating the exposition. toolbox Sanfelice et al. (2013)

 R. Bertollo et al. / IFAC PapersOnLine 55-32 (2022) 188–193 193

Parameter Value Parameter Value REFERENCES

D1 1 cm Di,i=1 3 cm
P 0 E 0 Cáceres, G., Millán, P., Pereira, M., and Lozano, D. (2021).
B 4.05 Ksat 1.056 cm/min Smart farm irrigation: Model predictive control for
ψsat 12 cm θmin 0.09 cm3 /cm3 economic optimal irrigation in agricolture. Agronomy,
θmax = θsat 0.395 cm3 /cm3 N 13 11(9).
L 5 · 103 cm r 0.8 cm Chen, W.H., Shang, C., Zhu, S., Haldeman, K., Santiago,
k 2.8 · 10−4 z 0.5 M., Stroock, A.D., and You, F. (2021). Data-driven
ρw 0.997 g/cm3 g 3.53 · 106 cm/min2 robust model predictive control framework for stem wa-
H 100 cm ∆L 20 cm ter potential regulation and irrigation in water manage-
W 60 cm Tadapt 48
hON 7 hOFF 20
ment. Control Engineering Practice, 113, 104841.
kρ 1 · 10−5 µ 0.15 cm3 /cm3 De Fraiture, C. and Wichelns, D. (2010). Satisfying
kγ 1 · 10−4 σ 0.05 cm3 /cm3 future water demands for agriculture. Agricultural water
Variable Initial condition Value Initial condition
management, 97(4), 502–511.
Θj,i,init 0.15 cm3 /cm3 γinit 0.13
EC (2019a). European Commission: Farm to fork strategy.
ρinit 3 · 10−3 URL https://ec.europa.eu/food/farm2fork_en.
Table 1. Numerical values of the parameters EC (2019b). European Commission: The European Green
and initial condition used in the simulation. Deal. COM/2019/640 final.
Elasbah, R., Selim, T., Mirdan, A., and Berndtsson, R.
(2019). Modeling of fertilizer transport for various
0.14
fertigation scenarios under drip irrigation. Water, 11(5),
;
.
893.
0.12
Ella, V.B., Reyes, M.R., and Yoder, R. (2009). Effect of hy-
0.1 draulic head and slope on water distribution uniformity
0.08 of a low-cost drip irrigation system. Applied Engineering
0.06
in Agriculture, 25(3), 349–356.
Evans, R.G. and King, B.A. (2010). Site-specific sprinkler
0.04
irrigation in a water limited future. In Proceedings of
0.02 the 5th National Decennial Irrigation Conference, 1.
0
0 50 100 150 200 250 300 350 400 450
FAO (2022). URL https://www.fao.org/home/es.
time [h] Giusti, E. and Marsili-Libelli, S. (2015). A fuzzy decision
support system for irrigation and water conservation in
Fig. 7. Evolution of γ (red) and ρ (blue) during the agriculture. Env. Modelling & Software, 63, 73–86.
simulation. Goebel, R., Sanfelice, R.G., and Teel, A.R. (2012). Hybrid
Dynamical Systems: Modeling, Stability, and Robust-
moisture, as we can see in Figure 7, where we depicted the ness. Princeton University Press.
evolution of ρ and γ. Lozano, D., Ruiz, N., Baeza, R., Contreras, J.I., and
Gavilán, P. (2020). Effect of pulse drip irrigation
duration on water distribution uniformity. Water, 12(8),
5. CONCLUSIONS 2276.
Lozoya, C., Mendoza, C., Mejı́a, L., Quintana, J., Men-
In this work we proposed a novel model for irrigation doza, G., Bustillos, M., Arras, O., and Solı́s, L. (2014).
lines with multiple crops and drippers. We used hybrid Model predictive control for closed-loop irrigation. IFAC
dynamical systems tools to model both the water dynamics Proceedings Volumes, 47(3), 4429–4434.
in the pipe, when the latter is characterized by some slope, Morillo, J.G., Dı́az, J.A.R., Camacho, E., and Montesinos,
and the binary nature of the control input, corresponding P. (2015). Linking water footprint accounting with
to the valve position. irrigation management in high value crops. Journal of
cleaner production, 87, 594–602.
Taking inspiration from the non-automated irrigation Oborkhale, L., Abioye, A., Egonwa, B., and Olalekan,
strategy applied by farmers, we also proposed a prelim- T. (2015). Design and implementation of automatic
inary hybrid feedback scheme, to stabilize the soil mois- irrigation control system. IOSR Journal of Computer
ture close to a desired value, reducing the non-uniformity Engineering (IOSR-JCE), 17(4), 99–11.
induced by the transients in the pipe. The behaviour of Ohaba, M., Abidin, M., Li, Q., Shibusawa, S., Kodaira,
the model and the control scheme was illustrated through M., and Osato, K. (2012). Adaptive control of capillary
simulations, based on real farm data. water flow under modified subsurface irrigation based
Future work will include a deeper analysis of the properties on a SPAC model. In Proc. of 11th International
of this compartmental-like hybrid model, the design of Conference on Precision Agriculture.
an optimal control law and the analysis of the stability Sanfelice, R., Copp, D., and Nanez, P. (2013). A toolbox
properties of the closed loop. for simulation of hybrid systems in matlab/simulink:
Hybrid equations (hyeq) toolbox. In Proceedings of
the 16th International Conference on Hybrid Systems:
ACKNOWLEDGEMENTS Computation and Control, 101–106.
Sellers, P., Randall, D., Collatz, G., Berry, J., Field, C.,
Dazlich, D., Zhang, C., Collelo, G., and Bounoua, L.
The authors thank the Spanish Agency for International (1996). A revised land surface parameterization (SiB2)
Cooperation and Development (AECID) and the Junta de for atmospheric GCMs. Journal of Climate, 9, 676–705.
Andalucı́a, for partially funding this work, under projects Water, U. (2018). Nature-based solutions for water. 2018
Agricultura 4.0. (Grant 2020/ACDE/000192) and IRRI- UN World Water Development Report.
GATE (Grant PY20-RE-017-LOYOLA), respectively.