Recei ed No embe 29, 2021, accep ed Decembe 29, 2021, da e o publica ion Decembe 30, 2021,
da e o cu en e sion Janua y 6, 2022.
Digi al Objec Iden i ie 10.1109/ACCESS.2021.3139764
Gaining S uden Engagemen Th ough
P ojec -Based Lea ning: A Compe i i e
2D Game Cons uc ion Case S udy
INIGO LOPEZ-GAZPIO
Depa men o In o ma ic, Elec onic and Communica ion Technologies, Uni e si y o Deus o, 20012 Donos ia-San Sebas ian, Spain
e-mail: [email p o ec ed]
ABSTRACT In his wo k we conside an open a i icial in elligence game as a ma e o s udy wi hin he
lec u es o a i icial in elligence o comba lack o mo i a ion and inc ease engagemen wi hin he class oom.
Du ing o ma ion, s uden s in compu e science can deal wi h mode a ely complex p ojec s, ne e heless,
dealing wi h such p oblems is elega ed o he Deg ee Final P ojec . In his in es iga ion we show he
p ocedu al s eps o how p ojec -based lea ning combined wi h game cons uc ion can e ec i ely be used o
p omo e engagemen in in o ma ic lec u es a uni e si y. Fo he ask, we build a 2D game engine and p opose
s uden s o en oll in ac i ious esea ch eams wi h he aim o p og amming in elligen agen s ha play he
game employing a i icial in elligence echniques. The in ended p incipal ou come is o show e idence o
he applica ion o p ojec -based lea ning in a i icial in elligence wi hin he lec u es, and how i can be
combined wi h game cons uc ion o inc ease mo i a ion in he class oom. P ojec -based lea ning has he
s uden s lea n, o ganize, and sol e challenges while s uden s hemsel es emain hei own esponsible o
he in es iga ion and p ocess o wo k. We p opose o ollow a se ies o sequen ial phases ha con o m a se
o miles ones ha inco po a e a p ojec -based lea ning app oach o he lec u es. Th ough his wo k we show
ha he use o p ojec -based lea ning combined wi h game cons uc ion p o ides eliable e idence ha a
much deepe unde s anding abou a i icial in elligence is a ained by s uden s pa icipa ing in he challenge.
S uden e alua ion ques ionnai es and inal g ade esul s a ained by s uden s indica e ha s uden s emained
mo e engaged du ing he semes e in compa ison o p e ious semes e s in which lack o mo i a ion was
epo ed.
INDEX TERMS Ac ion esea ch, ac i e lea ning compu ing educa ion, c oss-disciplina y skills, engage-
men , game cons uc ion, p ojec -based lea ning (PBL), ein o cemen lea ning.
I. INTRODUCTION
Resea ch in he ield o educa ional me hodologies and hei
applica ions sugges ha p ede ined labo a o y p ac ices a e
no as much a signi ican con ibu o owa ds s uden pe -
o mance as lec u e s can a i s imagine [1]–[5]. Ac u-
ally, i is an open esea ch ield o unde s and bo h, he
componen s and cons uc s ha enable signi ican s uden
engagemen in o de o achie e high quali y educa ion [6].
Lambo n e al. [7] de ine s uden engagemen as he ‘‘psy-
chological in es men in and e o di ec ed owa ds lea ning,
unde s anding o mas e ing he knowledge.’’ The ci ed s a e-
men has been u he ede ined such ha s uden engagemen
The associa e edi o coo dina ing he e iew o his manusc ip and
app o ing i o publica ion was Aasia Khanum .
is desc ibed as a p ocess in which he s uden ‘‘de elops an
a ec i e and beha io al pa icipa ion in he lea ning expe i-
ence’’ [5], [8]. Thus, hese in es iga ions claim ha ocused
pe sonal e o in lea ning ac i i ies is esponsible o be e
unde s anding and pe o ming in academic wo k.
I we wan s uden s o acqui e new knowledge, skills and
compe encies ha allow hem o ace cu en eali y, lec u -
e s mus assu e hem meaning ul lea ning [9]. Meaning ul
lea ning only occu s when s uden s a e mo i a ed and gi e
pe sonal meaning o wha hey a e ying o lea n. Fo his, i is
essen ial o inco po a e new me hodologies in he class oom
and o e al e na i e scena ios ha enable expe imen a ion
and loca es he s uden s ou o hei com o zone o o e new
engaging expe iences. In 1973, he Ame ican psychologis
B une de eloped he heo y o lea ning by disco e y [10] in
VOLUME 10, 2022 This wo k is licensed unde a C ea i e Commons A ibu ion 4.0 License. Fo mo e in o ma ion, see h ps://c ea i ecommons.o g/licenses/by/4.0/ 1881
I. Lopez-Gazpio: Gaining S uden Engagemen Th ough P ojec -Based Lea ning: Compe i i e 2D Game Cons uc ion Case S udy
which he e e ed o a pedagogical dynamic h ough which
he lea ne disco e s he concep s and hei ela ionships
and ea anges hem o i s own cogni i e schema. Mo e
ecen esea ch sugges s ha inco po a ing eam wo king in o
his me hodology p o ides e en mo e bene i o lea ne s.
Rush e al. no ed ha ‘‘when s uden s a e pa o a lea ning
communi y, hey become posi i e abou hei iden i y as a
membe o he g oup and emain ocused on lea ning’’ [11].
In his sense, we, he au ho s, belie e ha he usual ac i -
i y o enginee s is expe imen a ion and esea ch, he e o e,
expe imen a ion and esea ch mus be he main ool o each-
ing. A he same ime, he p ojec selec ed o expe imen a ion
mus be su icien ly engaging so o mo i a e s uden s and
p omo e commi men . I a su icien ly engaging p ojec is
selec ed, esea ch pe o med by s uden s will e en ually lead
o a meaning ul lea ning. In he pa icula case o enginee -
ing educa o s, hese also ha e he challenge o cul i a ing
a di e se se o cha ac e is ics o s uden s, such as apidly
e ol ing echnical compe encies [12], [13]. This apid de el-
opmen o in o ma ion and communica ion echnologies has
obliged educa ion sys ems o adap o new equi emen s in
li le ime [14]. Teaching is an a ha needs o be engaging so
ha bo h s uden s and lec u e s can collabo a e in he lea ning
p ocess. Fo se e al decades, he p ocess o collabo a i e
lea ning has been s udied as a ma e o g ea in e es [15],
[16]. E en mo e, conside able e o has been pu in o co e-
la ing he le el and quali y o s uden engagemen and pe -
sonal de elopmen wi h he academic p og am and ac i i ies
ca ied ou du ing lec u es h oughou he semes e [1], [5].
Dis inc al e na i es o inc easing s uden engagemen
ha e been p oposed, including p ojec -based lea ning,
p oblem-based lea ning and lea ning by eaching. As s a ed
by La aza-Mendiluze e al. [17] hese p oposals consis o
h ee cons uc ionis app oaches o human lea ning. I is
well discussed by Ploe zne e al. ha an e ec i e lea ning
p ocess mus in ol e a sel -cons uc ion s ep whe e oppo -
uni ies o lea ning mus be sel -c ea ed [18]. In o de o
lea n, s uden s mus wo k in an engaging en i onmen o
cons uc knowledge a he han ob ain i om he ex e nal
wo ld. I is well documen ed ha he educa i e e ec is
g ea e when knowledge is cons uc ed in his way. Mo e
classical eaching me hodologies aim o deli e concep s o
s uden s while cons uc ionis me hodologies a e ocused on
knowledge building expe iences.
In his con ex , p ojec -based lea ning (PBL) encou ages
cons uc ionis lea ning [19]–[21] in which s uden s build
knowledge ins ead o being simple ecipien s o i , and
encou ages hem o de elop c i ical and e lexi e hink-
ing. Acco ding o he ci ed wo ks PBL is de ined as a
s uden -cen e ed pedagogic me hodology in which dynamic
class oom app oaches a e used. The s uden -cen e ed ped-
agogy assu es ha s uden s acqui e a deepe knowledge o
he ma e s add essed h ough ac i e explo a ion o eal chal-
lenges o p oblems. In he applica ion o he me hodology s u-
den s wo k o ela i ely long pe iods o ime in an ac i e s yle
ying o in es iga e o espond complex challenges [22].
As a esul , PBL is a challenge no only o s uden s bu
also o lec u e s as i mo es away om he mo e adi ional
eaching p ac ices, such as: pape based labo a o y p ac ices
o eache -led ins uc ion p ac ices. PBL implies a change o
oles, wo k me hodology and e alua ion in which s uden s
ac i ely o m eams and ace a complex challenge ha mus
be sol ed wo king al oge he . PBL in i es s uden s o iden i y
wha hey should lea n, o ge in ol ed in he sea ch o
in o ma ion o sol e he p oblem, o de ine he s a egies o
be used and o ag ee on common solu ions. In his con ex ,
he lec u e becomes a guide o u o accompanying s uden s
and helping hem when equi ed, bu ne e as he sole sou ce
o wisdom. F om ou poin o iew, he aim o PBL is o ha e
u u e compe en p o essionals who a e inno a i e and ull o
con idence in hei abili ies and easoning.
The opic o a i icial in elligence (AI) is an impo an ield
wi hin compu e science enginee ing. I is o g ea impo ance
ha compu e enginee s unde s and how an in elligen agen
beha es and how such agen s can be de eloped. S uden s in
compu e science can deal wi h mode a ely complex a i icial
in elligence labo a o y p ojec s; ne e heless, dealing wi h a
eal AI p oblem is elega ed o he inal deg ee p ojec o o a
mas e p og am. In his a icle we conside an open a i icial
in elligence game as a ma e o s udy wi hin he AI lec u es.
The p e ious eaching me hodology goes no u he han he
classical app oach in which s uden s ecei e explana ions and
a e assigned se e al p ede ined labo a o y p ac ices which
leads many s uden s o ela e a i icial in elligence wi h an
un ashionable opic.
In his wo k we add ess he p oblem o inc easing s uden
engagemen h ough he employmen o mode n app oaches
such as p ojec -based lea ning combined wi h game con-
s uc ion. Such an app oach can be success ully used o
p omo e in o ma ics a Uni e si y. Fo he ask, we p opose
o ollow a se ies o sequen ial phases ha con o m a se
o miles ones ha help inco po a e a p ojec -based lea ning
me hodology o he a i icial in elligence lec u es. Th ough
he comple ion o he p oposed phases, eams o med by s u-
den s de elop in elligen agen playe s ha sol e he p oposed
game employing a i icial in elligence echniques. As ega ds
he implemen a ion o he game, we p e iously build a 2D
game engine om sc a ch based on a popula classic game
ha will ul ima ely in e ace wi h he playe s cons uc ed
by s uden s. Th oughou he p oposed phases s uden s a e
asked o en oll in ic i ious esea ch eams o cons uc he
equi ed playe s, based on a i icial in elligence sea ch and
explo a ion echniques discussed in he lec u es. The expe i-
ence also help s uden s acqui e leade ship and knowledge o
c oss-disciplina y compe encies, such as code de elopmen
o so wa e p ocess compe encies.
Based on s uden e alua ion ques ionnai es and inal
g ades, we show ha engagemen wi hin he class oom has
been highe when PBL combined wi h game cons uc ion is
employed. Real engagemen allows s uden s o de elop skills
a a g ea e ex en and when p ope ly s uc u ed, engagemen
can also lead o posi i e in e dependence be ween pee s and
1882 VOLUME 10, 2022
I. Lopez-Gazpio: Gaining S uden Engagemen Th ough P ojec -Based Lea ning: Compe i i e 2D Game Cons uc ion Case S udy
academic s a . This in u n can c ea e a sense o connec ed-
ness and pe sonal in es men in he p ocess o s udy.
This a icle is s uc u ed as ollows. Sec ion II p esen s
he s a e-o - he-a analysis comp ising dis inc app oaches
and me hodologies employed in eaching compu e science,
as well as game cons uc ion challenges add essed p e i-
ously in he academia, Sec ion III desc ibes he game engine
cons uc ion and summa izes he cou se in which he chal-
lenge is in oduced and he p oposed me hodology is applied.
Once he challenge and he cou se subjec o modi ica ion
a e desc ibed, Sec ion IV desc ibes he applica ion o he
new me hodology, and Sec ion V p esen s he e olu ion o
he eams du ing he p oposed phases. Finally, Sec ion VI
p esen s he conclusions o he in es iga ion.
II. STATE-OF-THE-ART ON METHODOLOGIES USED TO
TEACH COMPUTER SCIENCE
The s a e-o - he-a (SOA) epo s a a ying se o s yles
o eaching [23], hese me hodologies ange om pu ely
heo e ical app oaches o comple ely p ac ical ones. Fo
he case o compu e science educa ion ou SOA analy-
sis e ealed ha a wide bunch o lea ning me hodologies
is being employed o inc ease engagemen and a ac he
a en ion in a ious science domains. Fo ins ance, an in e -
es ing analysis pe o med by La aza-Mendiluze e al. o
ex ac how compu e a chi ec u e is being augh h oughou
well-known wo ldwide uni e si ies de e mined ha he mos
employed eaching s a egies old in o one o he ollow-
ing se o ca ego ies [24]: i) pu ely desc ip i e app oaches,
ii) pe o mance-based app oaches, iii) p ac ical app oaches
based on p og amming assignmen s and i ) open app oaches.
To ob ain his ca ego iza ion he ci ed wo k analyzed he
educa ional p ocess ca ied ou in 36 Uni e si ies in which
examina ion ques ions, cou se syllabi, ex books and cou se
ma e ials we e exhaus i ely analyzed.
i) Pu ely desc ip i e app oaches only show s uden s
a desc ip ion o he ma e unde analysis. S uden s a e
expec ed o unde s and he lec u e, and he concep s being
desc ibed by hem. Pu ely desc ip i e p ac ices also con-
side ha s uden s can be asked o pe o m some summa i e
assignmen du ing he semes e . This is known o be an easy
way o in oduce opics in he class oom, al hough he le el
o a ac ion o he s uden s is no high. Pu ely desc ip i e
app oaches lack om ob aining high a ac ion since hey
ba ely consis o ex essay ealiza ions.
ii) Pe o mance-based app oaches equi e s uden s o
ealize some kind o compu a ion, so ha he esul s ob ained
a e hen compa ed o some e e ence esul known by he
lec u e . This app oach equi es he applica ion o ce ain
g ade o knowledge o de e mine he app op ia eness o he
echniques used by he s uden [25]. This app oach is also
known in he li e a u e as illus a i e expe imen s [26] which
in ol e expe iences accompanied wi h explana ions gi en by
he eache .
iii) P og amming p ac ical app oaches a e based on
a long- e m p ojec ha mus be elabo a ed by s uden
indi iduals o g oups. This app oach helps o in oduce
abs ac concep s egula ed by he lec u e when de ining he
p ojec scope. I is widely belie ed ha hands-on p og am-
ming helps make concep s clea e . Howe e , i u ns o be a
ough expe ience i he s uden lacks he necessa y abili y o
o e come he p ojec and, consequen ly, engagemen can be
apidly los , especially in small g oups o wo king alone.
i ) Open app oaches a e ela ed o non-guided esea ch
[26] in which he app oach combines he acquisi ion and con-
s uc ion o concep s, he esolu ion o p ac ical p oblems and
p ac ical wo k usually pe o med by s uden s in coope a ing
eams. Open app oaches aim o make s uden s become he
di ec o s o hei own lea ning by p omo ing a new ac i e
lea ning p ocess h ough coope a ion. In his model, lec u e s
ac as acili a o s o he lea ning ha mus come om wi hin
he esea ch pe o med by he eam. F om he didac ic poin
o iew, i is he guided esea ch, and no he s a ic p ac ices
ha allow a comple e and e ec i e lea ning o he p ocedu al,
a i udinal and concep ual concep s. S uden s a e gi en he
oppo uni y o expe ience he usual challenges ha compu e
enginee s encoun e in a daily basis.
Open guided esea ch app oaches ha e shown ha hey
signi ican ly imp o e quali y o s uden s o ace eal p ob-
lems [27]. Acco ding o he ci ed in es iga ion open guided
esea ch app oaches a e capable o s imula e c i ical hink-
ing, o e ing nume ous ad an ages o e adi ional s a ic
app oaches due o hei inhe en open na u e. Wi hou a s ic
ollow-up o a de ailed sc ip s uden s a e gi en he chance o
wo k on small g oups c ea ing he op imal scena ios o g ow
concep s.
On he con a y, as i is also s a ed in [27], e en PBL
in ol es se e al ad an ages i encompasses a ious di icul-
ies along he implemen a ion o he p ocedu e, such as: he
complexi y o handle nume ous p ojec s in he class oom
and i s di icul y o s anda dize he e alua ion. I di icul ies
a e o e come mos o he lec u e s ag ee ha eaching wi h
PBL has nume ous ad an ages, pa icula ly in he ield o
compu e science [27]. Ye , he e is no a one- i s-all-solu ion
on how o apply open guided esea ch in a pa icula scena -
ios [28], and he speci ic p ocedu e on how o apply he open
guided esea ch me hodology a ies among he 500 p ojec s
in es iga ed in [27]. As an example, he au ho s in [29],
[30] e alua ed dis inc s uden g oups augh unde di e en
me hodologies and indica e ha s uden g oups in ol ed in
open guided esea ch ac i i ies ob ained highe g ades and a
g ea e abili y in he ex en o which hey could use co e ideas
o engage science p ac ices.
Open guided esea ch app oaches ha e also al eady been
success ully es ed in dis inc domains o enginee ing sci-
ences, o ins ance, [1], [31] epo ha o enginee ing lec-
u es open simula ion exe cises ha e been widely adop ed
wi h g ea success. In his di ec ion, [32], [33] desc ibe wo ks
ha employ simula ion models o suppo educa ion and o e
mo e in e es ing and ealis ic open en i onmen s o s uden s.
In he ecen pas PBL p ac ices ha e been success-
ully employed in dis inc a eas o knowledge o sol e
VOLUME 10, 2022 1883
I. Lopez-Gazpio: Gaining S uden Engagemen Th ough P ojec -Based Lea ning: Compe i i e 2D Game Cons uc ion Case S udy
open challenges. Some in es iga ions on PBL combined
wi h open p oblems a e hose o [34], [35] in which he
au ho s ied o p o ide a solu ion o he poo mo i a ion
o s uden s in labo a o ies wi h ew esou ces by employ-
ing s uden -cen e ed me hodologies and ques ionnai es. The
s uden -cen e ed me hodology is p o en o be o use allow-
ing o mo e enjoyable, in e es ing, c ea i e, and mo i a ing
p ac ices. The e a e many e e ences in he li e a u e ela ed
o PBL ha s ess i s impo ance and e ec i eness, as an
example he in es iga ions o Puche e al. in [27] enume a es
mo e han 500 p ojec s in ol ing PBL wi h a ious ou comes.
Ne e heless, he wo k ema ks ha p ojec ideas sugges ed
by s uden s u n ou o yield excep ional good esul s in mos
o he cases. This ema k sugges s p e ious in es iga ions
ha sugges ed an e ec i e lea ning p ocess mus in ol e a
sel -cons uc ion s ep whe e oppo uni ies o lea ning mus
be sel -c ea ed. The wo k also s a es he bene i s and di i-
cul ies o he me hodology and poses he conce n ha many
imes he bene icial e ec o he me hodology is based on
he con ex o he p oblem and he subjec i e ision o he
eache . Mo e speci ic in es iga ions on he applica ions o
PBL include hose o La aza-Mendiluze e al. in which he
au ho s p o ide an ex ensi e enume a e o plausible p ojec s
ha could be ca ied ou in he a ea o compu e a chi-
ec u e [36]. In [37], [38] he au ho s elabo a e on dis inc
p ojec s om he compu e a chi ec u e domain in which
PBL me hodology has been inco po a ed o lec u es wi h
success con on ing s uden s wi h eal wo ld open challenges
ha mus be esol ed.
The p oposal o open challenges based in games also i s
pa icula ly well wi h PBL in he sense hey de ine and
con ex ualize a eal o simula ed challenge ha mus be
add essed. In his ega d, he wo k desc ibed in [39] collec s
expe iences om he pas wo decades in which compu e
games ha e been used o esea ch in he pa icula case
o compu e science and compu a ional in elligence, among
o he s. Acco ding o he au ho s games ha e always been
a popula es bed no only o scien i ic esea ch, bu also
educa ional esea ch due o he ac games o e a big ange
o lexibili y and a iable p oblem complexi y. The wo k
s a es ha games se led in an open wo ld en i onmen and
consis ing o a ela i ely small se o ac ions can pose an
in e es ing challenge o con olled esea ch.
III. COURSE OVERVIEW AND CONSTRUCTION OF THE
GAME ENGINE
A i icial in elligence is a i h-yea second semes e cou se
in he Depa men o In o ma ic, Elec onic and Communica-
ion Technologies a he Uni e si y o Deus o (San Sebas ian
acul y, Spain). The cou se in oduces classic elemen s o
a i icial in elligence o s uden s, such as: ee and g aph
sea ch heu is ics, local sea ch op imiza ion, popula ion-based
op imiza ion and in oduc o y machine lea ning concep s.
The cou se p o ides an ini ial unde s anding o wha ac ually
is inside an in elligen agen implemen a ion, and i is based
on a classical app oach based on he well-known e e ence
FIGURE 1. Ini ial disposi ion o he i een s icks laye ed in ows. The
game always s a s wi h se en, i e and h ee s icks o ganized in h ee
ows. Playe s, in u ns, keep on emo ing s icks un il he las s ick is
aken om he boa d, which dic a es he game lose .
by Russel and No ig [40]. The cou se comp ises ou con-
ac hou s pe week, and an addi ional u o ing hou . I has
adi ionally been deli e ed in ace- o- ace desc ip i e o ma
using a chalk-and- alk s yle, aided wi h some labo a o y p o-
g amming assignmen s.
The lea ning ou comes o he cou se a e ha upon suc-
cess ul comple ion o he cou se s uden s should be able
o: (i) iden i y, analyze and de ine he signi ican elemen s
making up a p oblem in o de o sol e i wi h c i e ia and
e ec i ely; (ii) use one’s own expe ience and c i e ia in he
analysis o he causes o a p oblem and build up a mo e
e ec i e and e icien solu ion; (iii) unde s and he unda-
men al p inciples and basic echniques o in elligen sys-
ems and hei p ac ical applica ions; (i ) o mula e sea ch
p oblems and iden i y and apply an app op ia e sol ing
echnique; ( ) de ine and apply good heu is ics o sol e
di e en p oblems conside ed di icul ; ( i) apply machine
lea ning echniques as a way o an in elligen sys em o gain
a ce ain deg ee o au onomy; and ( ii) analyze p oblems
whose esolu ion equi es empi ical knowledge and o design
knowledge-based sys ems.
Thus, indi iduals who g adua e in compu e enginee ing
a e able o apply hei knowledge o sol e AI p oblems in
di e en a eas ela ed o in o ma ion echnologies, p o id-
ing he mos app op ia e solu ions in each case. The main
con ibu ion o he subjec In elligen Sys ems o he deg ee
in compu e enginee ing is he esolu ion o p oblems and
design o applica ions, based on exis ing equi emen s and
applying he p ecise c i e ia o e ec i eness, e iciency, cos s
and bene i s in ol ed. In pa icula , he subjec o In elli-
gen Sys ems emphasizes he esolu ion o di icul p oblems,
many o hem NP-comple e, h ough he design and use o
heu is ics in a i icial in elligence algo i hms; he design and
use o machine lea ning echniques, and knowledge-based
sys ems.
Fo many yea s, a i icial in elligence has been augh
using a adi ional app oach based on labo a o y p ac ices,
in which s uden s comple e well-known a i icial in elligence
1884 VOLUME 10, 2022
I. Lopez-Gazpio: Gaining S uden Engagemen Th ough P ojec -Based Lea ning: Compe i i e 2D Game Cons uc ion Case S udy
p oblems om he s a e o he a , such as he N-Queens p ob-
lem o he T a elling Salesman P oblem (TSP). Despi e he
ac he classical algo i hms o sol ing hese p oblems need
o be discussed in class, hey o e an old- ashioned app oach
ha ails in engaging s uden s. Rega ding he subjec ecen
eedback epo ed ha a signi ican numbe o s uden s do
no engage wi h he con en o he lec u es. Repo s mainly
ocus on he una ac i eness o classical AI me hods seen
du ing he i s weeks o he semes e . S uden s jus do no
eel he classical app oaches a e o any mo e use in he ecen
o nea by u u e. This esul s in a d op o mo i a ion o
s udy AI algo i hms which epe cusses in poo class pe o -
mance. Some s uden s e en mani es he non-use ulness o
he classical AI app oaches when machine lea ning o deep
lea ning is cu en ly a aining such a g ea impac on socie al
applica ions.
Based on bo h, ying o make a mo e iendly and a ac-
i e lea ning s yle, and a mo e engaging lea ning p ocess,
he au ho s conside ed he oppo uni y o inco po a e a
p ojec -based lea ning app oach based on game cons uc ion
o engage s uden s. In his se ing lec u e s build a 2D game
engine ha ep esen s a classic well-known challenge and
p opose s uden s o en oll in ac i ious esea ch eams wi h he
aim o p og amming in elligen agen playe s ha play and
sol e he game employing a i icial in elligence echniques.
Following [39] in o de he game o expose an in e es ing
esea ch en i onmen , he p oposed game should be open,
wi h a ela i ely small amoun o ac ions and pose a eal
wo ld challenge o s uden s. Ye , in o de o main ain s uden
engagemen high he p oposal o how o code he in elligen
AI playe mus come wi hin he s uden g oups [27]. To his
aim, he game engine is designed so ha i can in e ace wi h
any kind o playe ha ollows a de ined in e ace o pe o m
game ac ions (e.g. emo e a ce ain amoun o s icks om he
boa d).
We selec ed he opic o he game so as o amilia ize
and engage he age o he s uden s (23-24 yea s). Fo he
ask, we selec ed an a i icial in elligence game ha is being
popula ized by he well-known ‘‘El Ho migue o’’ maximum
audience ele ision b oadcas , which among o he s in ends
o deli e science o young adul s: he i een s icks game.
The i een s icks game consis s o an open AI p oblem ha
can only be sol ed by complex ma h-based algo i hms as
i is a a he complex game wi h a high numbe o s a e-
space ou comes, bu a ela i ely low ac ion space se ; in
con as o games like Chess o Go!, he numbe o ac ions
and s a e-space ou comes is ye ac able compu a ionally
and makes he esolu ion easible o s uden s. The i een
s icks game1se es a pe ec scena io o challenge s uden s
as i consis s o a classical 2-playe compe i i e game ha
consis s o emo ing s icks om a boa d in u ns, as many as
eques ed by playe s, bu only om one o he h ee a ailable
ows a a ime. The playe who ca ches he las s ick is
1The i een s icks game is a edesign o he popula Nim game
h ps://en.wikipedia.o g/wiki/Nim
FIGURE 2. Full sequence o a game played by wo human agen s ha
communica e h ough he keyboa d wi h he game engine. Top le
sub igu e shows he ini ial game s a e and each o he ollowing
sub igu es ep esen an ac ion aken by each o he human playe s
(playe 1 selec s s icks o be emo ed in shaded g ay and playe 2 selec s
s icks o be emo ed in ed). A each ime s ep any numbe o s icks can
be emo ed om a single ow, un il a single s ick is le . The playe ha
emo es ha inal s ick is he game lose .
he one who loses he ma ch. The game always s a s wi h
i een s icks, dis ibu ed in h ee ows as shown in Figu e 1.
We implemen ed he game engine using py hon p og am-
ming language and he pyGame2game de elopmen lib a y.
The game engine is designed in such a way ha di e en
ac o playe s can be employed o launch he game, a i s
2h ps://www.pygame.o g/news
VOLUME 10, 2022 1885
I. Lopez-Gazpio: Gaining S uden Engagemen Th ough P ojec -Based Lea ning: Compe i i e 2D Game Cons uc ion Case S udy
FIGURE 3. Block diag am desc ibing he de ined miles ones o he
applica ion o he PBL me hodology.
we implemen ed a andom playe and a human playe ha
communica es wi h he game engine h ough he keyboa d.
A ull sequence o a game played using wo human playe s is
shown in Figu e 2.
The challenge p oposed o s uden s is hen o en oll in
ic i ious esea ch eams o analyze he p oblem and de elop
an in elligen agen ha is able o play he game wi h a
high win a io, p e e ably wi h supe human pe o mance. Fo
he ask, hey mus employ AI sea ch echniques discussed
wi hin he lec u es (e.g. ee-sea ch based heu is ics, g aph-
sea ch based heu is ics, local sea ches, machine lea ning, . . .)
F om he me hodological poin o iew, his p ac ice aims o
s uden s o make a i s con ac wi h he wo ld o esea ch,
as well as os e ing a c i ical and inno a i e spi i . I in ends
ha g oups o mo i a ed s uden s apply he concep s o he
In elligen Sys ems subjec in he speci ic game con ex ha
has been desc ibed, and de elop an in elligen agen able o
e icien ly play he game.
IV. APPLICATION OF THE PBL METHODOLOGY
Inco po a ing game in e ac ion is a e y lexible and e icien
solu ion o cons uc a su icien ly complex p oblem ha
poses some deg ee o con olled challenge o he s uden s
FIGURE 4. Kick-o mee ing o eam pa icipa ing in he challenge.
while i allows o elabo a e pa icula sel de ined solu ions
o he AI playe s. In he ecen pas , games and gaming pla -
o ms a e becoming mo e and mo e ele an in educa ional
en i onmen s. Fo ins ance, in he compu e science and com-
pu e enginee ing domain he majo i y o he li e a u e ci es
games designed o each p og amming languages [41]–[44].
Addi ionally, simila a emp s ha e also been pe o med in
o he compu e science domains in o de o each compu e
a chi ec u e [1], [36] o o each compu e memo y [45].
As epo ed by [27], mos o he s uden g oup ailu es
come om he nonexis ence o o ganiza ion wi hin s uden
g oups o he lack o eam wo king expe ience. To o e -
come hese di icul ies, guided by he lec u e he s uden s
employ he phases p oposed in he me hodology p esen ed in
Figu e 3 o guide hei p ojec li e-cycle. The PBL me hod-
ology o he cons uc ion o he AI playe is composed o
se en phases, as deno ed in he men ioned block diag am.
These se en phases in ol ed in he me hodology co e he
mos ele an aspec s o agile so wa e de elopmen and a e
o ganized in sequen ial miles ones ha s uden eams mus
adop in hei imeline in o de o de elop he p ojec . The
phases a e enume a ed as a se ies o sequenced miles ones:
1) P oblem explo a ion, 2) S a e-o - he-a and ele an ech-
nology analysis, 3) Design o he solu ion, 4) Implemen a ion
o he solu ion, 5) Tes ing and e alua ion o he solu ion,
6) Documen a ion o he solu ion, and 7) Dissemina ion o
documen a ion.
Du ing he li e-cycle o he p ojec lec u e no es on AI
o build he so wa e agen along o he compu e science
compe encies and knowledge a e o be used, as he phases
de ine asks ela ed o con o ming g oups, de e mining oles
and objec i es, coding, es ing, deploying and documen ing
he solu ion. Among o he s, i is expec ed o b ing a leas
so wa e p ocess and quali y, so wa e design and p og am-
ming compe encies o e on . The p ojec also bene i s om
collabo a i e lea ning as s uden s ac as eache s o hei
eamma es. Taking he dual oles o lea ne s and eache s help
s uden s u he hei knowledge o in o ma ics [46].
A. DESCRIPTION OF MILESTONES AND EVOLUTION OF
THE STUDENT TEAMS
1) PROBLEM EXPLORATION
In his phase he eam conce ns abou he p oblem o be
add essed. As he p oblem is a challenge ha mo i a es he
1886 VOLUME 10, 2022
I. Lopez-Gazpio: Gaining S uden Engagemen Th ough P ojec -Based Lea ning: Compe i i e 2D Game Cons uc ion Case S udy
FIGURE 5. The igu e shows h ee game s a es om he i een s icks challenge ha will e en ually
lead he agen o lose.
eam chances o lose in e es du ing he li e cycle o he
challenge a e low, and he eam soon s a s delimi ing he
scope o he wo k o be done. A his s age i is impo an
o s a unde s anding and de ining clea objec i es. Figu e 4
shows one p ac ical session in which a eam is b ains o ming
on he p oblem explo a ion o he game.
2) STATE-OF-THE-ART AND RELEVANT TECHNOLOGY
ANALYSIS
In his phase, he eam wo ks on explo ing a possible solu ion
and ga he ing use ul echniques and knowledge, mee ings
in ol ing b ains o ming a e especially ele an a his poin .
Explo ing AI echniques o collec ing u he equi emen s
ega ding he unc ionali y o he so wa e o be buil a e
example ou comes o his phase. In he p oposed challenge,
he ou come consis ed o a b oad explo a ion o sea ch ech-
niques used in he s a e o he a in he domain o AI.
3) DESIGN OF THE SOLUTION
In his phase eams ocus on designing a so wa e a chi-
ec u e ha esponds o he equi emen s iden i ied in he
second phase. Uni ied Modeling Language (UML) diag ams
and o he ypes o so wa e maps is he key o a success ul
solu ion design phase. The in elligen agen o be de eloped
in he eam mus be able o model he sea ch s a e-space ee
o he possible occu ences o he game, and i mus be able o
classi y game-s a e ep esen a ions acco ding o some u ili y
unc ion (used o de e mine goodness o badness o s a es and
ac ions). In ha sense he agen mus be able o de e mine
use ul game mo es ha e ol e owa ds he agen winning
he play e en ually. Iden i ying he game s a es ha lead he
agen o lose is also an in e es ing aspec discussed wi hin
eams. A his phase, eams o med by s uden s we e able
o model game s a es and iden i y he losing s a es as shown
in Figu e 5. Du ing his phase alida ion wi h he lec u e is
necessa y in o de o main ain he scope o he wo k and no
o e load he objec i es o he eam and keep hem ealis ic
and doable.
4) IMPLEMENTATION OF THE SOLUTION
Du ing his phase, he eam ocuses on implemen ing he
design made using a ious p og amming languages lea n
h oughou he in o ma ics deg ee. A his phase, he eam
ocuses on scheduling and bu ning miles ones ega ding
iden i ied unc ionali ies. Fo he speci ic case o he
desc ibed p ac ice, many eams decided o implemen he
code in Py hon p og amming language and employ some kind
o heu is ic sea ch o ein o cemen lea ning app oach so ha
he agen was able o ind he bes ac ion se in o de o win
he game. Equa ion 1shows an example o a eam making
use o ein o cemen lea ning (Tempo al di e ence online
lea ning) app oach, which employs he VS unc ion as he
alue unc ion o ob ain he u ili y (goodness o badness) o
s a es om he game. The VS unc ion is shown below, and
i can be sho ly summa ized as a elaxa ion o he Bellman
equa ion [47]. The VS unc ion g ades goodness o badness
o s a es be ween 0 and 1 acco ding o how nea he e is a
winning o losing s a e om he ac ual game s a e. Winning
and losing s a es as de ined by a s uden eam a e shown
in Figu e 5, and Figu e 6shows a play sequence o a ull
episode o he game.
VS(s)=VS(s)+alpha ∗(VS(snex )−VS(s)) (1)
whe e s deno es he cu en s a e, snex deno es he nex s a e
and alpha is he eloci y o lea n, a model hype pa ame e .
The ma hema ical model le s he agen lea n backp opaga ing
he alue unc ion o s a es once he episode is inished
un olling he pe o med s eps.
5) TESTING AND EVALUATION OF THE SOLUTION
A his phase eams expe imen ed wi h se e al ac o s, such
as he i e a ion numbe o ain he agen and i s success a e o
win, he lea ning a e hype pa ame e and o he explo a ion
hype pa ame e s. Fo solu ion e alua ion, eams epo ed
di e en kinds o expe imen s and esul s such as he e ec
he agen has been aining wi h ega d o success a e. The
eam discussions o igina ed a his s ep a e e y en iching o
he whole class oom as hey analyze pe o mance and gain
p ac ice o ganizing and showing expe imen esul s.
6) DOCUMENTATION OF THE SOLUTION
A his phase s uden s a e asked o o e come he unnel
ision acqui ed du ing p e ious phases and documen he
solu ion explaining he AI echniques employed. One o he
bes pe o ming eams managed o p oduce a scien i ic pape
in a jou nal om he Uni e si y o he Basque Coun y. Fo
his pa icula eam he ex en o he e o was g ea e han
o o he eams, bu he ewa d o wo king h ough he pee
VOLUME 10, 2022 1887
I. Lopez-Gazpio: Gaining S uden Engagemen Th ough P ojec -Based Lea ning: Compe i i e 2D Game Cons uc ion Case S udy
FIGURE 6. VS example alues o a sequence play pe o med by he in elligen agen
de eloped by a eam. No e ha a e each ac ion he opponen mus also pe o m
some ac ion, ha ac ion is unknown and no e lec ed in he agen ’s his o y.
e iew p ocess o he jou nal and ha e i published [48]
consis ed o a g ea ewa d o hei e o .
7) DISSEMINATION OF DOCUMENTATION
A his phase eams communica e hei wo k o eamma es
and a gene al discussion was ca ied ou in class. Dis inc
solu ions a e explained, analyzed, and discussed by eam
membe s and s uden s lea n how a scien i ic discussion is
ca ied ou . Some eams also pa icipa ed in a ideo eco ding
o show o he s uden s and u u e incoming s uden s wha
kind o ac i i ies hey pe o med in class. Fo his, s uden s
es ablished con ac wi h he ma ke ing and communica ion
depa men om he Facul y o Deus o in San Sebas ian
and Bilbao. The ideo eco ded is expec ed o be use ul o
opening days o dissemina ion ac i i ies ela ed o he deg ee.
V. STUDENT ENGAGEMENT EVALUATION
This sec ion desc ibes he ou e alua ion scena ios employed
o measu e he s uden engagemen du ing he PBL p ac ice.
Fo he e alua ion we ha e aken in o accoun : A) he in e -
nal conclusions o he s uden eams, which co esponds o
he in a-e alua ion epo s o he s uden s ex ac ed om
assignmen s, B) imp essions o he lec u e s h oughou he
semes e , C) inal lec u e g ades, and D) s uden engagemen
e alua ion epo s collec ed using ques ionnai es by he Uni-
e si y o Deus o.
A. INTERNAL CONCLUSIONS OF STUDENT TEAMS
The aim o he challenge p oposed has been non- i ial as i
in ol ed analyzing, s udying and applying he use o a i icial
in elligence in he con ex o a game. To do his, we i s
b ains o m and a gue abou di e en ways a i icial in elli-
gence could be applied in his con ex , such as wha a ional
beha io is and by wha means we could possibly e lec his
beha io in an in elligen agen . A e explaining classical
me hods om he s a e o he a , we we e gi en addi ional
ins uc ions on echniques, which could be applied in he
i een s icks game. The cu en a icle desc ibes all he wo k
pe o med du ing he semes e , including bo h: he desc ip-
ion o he game and he a i icial in elligence echniques used
o sol e i . We also sha ed he code in Gi hub o bo h: he
game engine p og ammed in py hon and he in elligen agen
class ile, which we belie e o be easily adap able o sol e
any o he simila game o use in o he con ex s o eaching
and lea ning. We also de eloped an e alua ion me ic o
measu e he pe o mance o ou in elligen agen , based on
he p obabili y o na iga e o good and bad s a es. The esul s
ob ained showed ha ou agen pe o med a supe human
le el, wi h e y low p obabili ies o lose.
B. LECTURERS’ IMPRESSIONS
The e is no doub ha me hodologies aimed o engage s u-
den s need u he e alua ion, and ha o his, educa ional
esea ch is necessa y. Many ools ha e been de eloped, bu
mos o hem ha e no been e alua ed no ha e clea ou -
comes. In his expe imen , we ha e challenged s uden s o
o m small uni s o ic i ious esea ch g oups and sol e a chal-
lenge ha was ne e compu a ionally ackled be o e employ-
ing p ojec -based lea ning and game cons uc ion. O e all,
he de elopmen o he p ojec has been posi i ely e alua ed
by he au ho s because o he ollowing ou comes:
1) I gene a es a high in e es as s uden s ace wi h a
eal challenge hey ha e been wa ching ecen ly on he
media.
2) Se e al di e en speci ic, gene ic and ans e sal com-
pe ences om he compu e science deg ee a e applied
o sol e he challenge.
3) S uden s p ac ice coding heo y lec u es seen in class,
and ac as acili a o s o he knowledge as needed
by eamma es. They cons uc knowledge as he eam
e ol es. In his ega d, communica ion and con lic
1888 VOLUME 10, 2022
I. Lopez-Gazpio: Gaining S uden Engagemen Th ough P ojec -Based Lea ning: Compe i i e 2D Game Cons uc ion Case S udy
sol ing skills a e de eloped wi hin all eam discus-
sions.
4) S uden s simula e a esea ch g oup and can y o pub-
lish hei esul s, which b ings hem an ex a mo i a-
ion o inish he wo k and p esen esul s. Also, hey
ge o know how he esea ch communi y wo k, he
pee e iewing p ocess and hey ob ain eedback om
expe s dis inc o hei habi ual lec u e s. In-class dis-
cussions abou all hese aspec s also a o communica-
ion be ween s uden s. To be mo e speci ic, du ing he
challenge we no iced ha ha ing an ex e nal e alua ion
boa d om he jou nal esul ed in an addi ional engage-
men componen in eams ha decided o p esen hei
wo k as i ex ends he ba ie s o he lec u e.
I is also o in e es o men ion ha due o he na u e and
cha ac e is ics o he challenge, an agile so wa e me hod-
ology can be u he de ined o ca y ou he p ojec . Agile
me hodologies ocus on os e ing communica ion and he
i e a i e and inc emen al de elopmen o p o o ype-based
so wa e. This is achie ed by assigning oles and esponsibil-
i ies among eam membe s wo k, which a o s he consolida-
ion and in eg a ion o he eam. Agile so wa e de elopmen
is cu en ly a ma e o g ea in e es in compu e science
deg ees. O e all, we hink i has been a e y ewa ding expe-
ience o see s uden s manage o build an in elligen agen
ha sol es an open game challenge ha is cu en ly being
b oadcas ed on TV.
C. STUDENT’S FINAL GRADE PERFORMANCE ANALYSIS
In o de o asse ha he ins uc ional s a egy in ol ing
p ojec -based lea ning wi h game cons uc ion is an e ec-
i e app oach o help s uden s lea n subjec s, we compa e
among dis inc s uden pe o mances ac oss wo consecu i e
academic cou ses in which he a i icial in elligence lec u e
has been augh ollowing he same schedule and e alua ion
c i e ia bu ha e been augh unde di e en lea ning p o-
cesses. One o he cou ses has been augh in ol ing adi-
ional p ac ices (2018/2019) while he second cou se has been
augh wi h he PBL me hodology (2019/2020).
We hink ha quali y and quan i y o e idence used o
suppo he s a ed asse ion o ins uc ional e ec i eness is
highly ela ed o he inal pe o mance o s uden s mea-
su ed as lec u e g ades once he lea ning p ocess is in-
ished. Final e alua ion g ades ha e been ex ensi ely used o
de e mine cou se o e all pe o mances and measu ing s u-
den engagemen h ough he analysis o inal g ades is a
common p ac ice in he s a e-o - he a wi hin educa ional
esea ch [1]. The dis ibu ion o s uden s alongside ob ained
s a is ics o g ades is shown in Table 1 o he academic
cou ses 2018/2019 and 2019/2020.
Assuming an unde lying no mal dis ibu ion on he g ades
ob ained by s uden s as indica ed by a Shapi o-Wilk es
o no mali y (p- alue >0.05), we conduc ed a one-sample
Wilcoxon signed- ank non-pa ame ic es o de e mine
whe he he pe o mance mean alues we e signi ican ly
di e en among dis inc academic yea s. The esul s o he
Wilcoxon es (V =17, p- alue <0.001) indica e ha s uden
a e age pe o mance has been highe when p ojec -based
lea ning ac i i ies ha e been ca ied ou wi h ega d o he
classical app oach. Signi ican ly be e s uden pe o mance
sugges s ha a highe le el o engagemen has been a ained
in academic cou se 2019/2020, in which p ojec -based lea n-
ing me hodology was employed.
D. STUDENT ENGAGEMENT QUESTIONNAIRES
In addi ion o he inal g ades ob ained by s uden s we also
include he S uden Engagemen E alua ion Repo s (SEER)
ha he Uni e si y o Deus o collec s e e y academic yea
wi h ega d o all subjec s augh a he Uni e si y collec ed
by means o ques ionnai es designed and o e ed o s uden s
by hi d-pa y p o essionals. SEER ques ionnai es include
a o al o 30 i ems o e alua ion ega ding he imp essions
o s uden s wi h ega d o engagemen in e en ion. SEER
ques ionnai es a e comple ely anonymous o lec u e s o
which hey can only access he inal mean esul o each i em
unde analysis. The undamen al pu pose o his e alua ion
is o asses lec u e s in hei con inuous imp o emen . In his
ega d, he assessmen made by s uden s on di e en aspec s
o dimensions o he eaching me hodology can be a aluable
aid as a basis o e lec ion and iden i ica ion o s eng hs and
weaknesses, as well as o quan i y s uden engagemen along
he semes e . The i ems unde e alua ion old in o one o he
ollowing ca ego ies:
1) Design and Planning: de ini ion and o ganiza ion
skills, con en , esou ces, ma e ials, me hodologies,
u o ing sys ems and e alua ion ha , acco ding o
he Uni e si y o Deus o T aining Model (MFUD),
ensu es an au onomous and signi ican lea ning
p ocess.
2) Lea ning Managemen : de elop he se o ac i i ies
ha acco ding o he MFUD ac i ely in ol e s uden s
in he lea ning p ocess.
3) Tu o ing and E alua ion: o ca y ou e ec i e mon-
i o ing and suppo du ing he s uden ’s lea ning p o-
cess ha allows o e i y, communica e and acili a e
he imp o emen o he deg ee o de elopmen and
achie emen o compe ences h oughou he p ocess
and as a inal esul .
4) Re iew and Imp o emen : e i ica ion o he deg ee
o compliance wi h he planning and he adequacy o
he esou ces used, making he consequen decisions
o hei imp o emen .
5) Collegiali y: ac i i ies ha lec u e s ca y ou oge he ,
which a e common and ha e a collec i e and com-
muni y o ien a ion (all ha e e s o wo k be ween
eache s as membe s o he depa men , deg ee, acul y
and uni e si y). Ac i e pa icipa ion wi h o he eache s
who in luence he uni o de elop he di e en pa s
o he p ocess, ensu ing au onomous and meaning ul
lea ning on he pa o he s uden s.
VOLUME 10, 2022 1889
