Long-Te m Assessmen o a Se ice Robo in a Ho el
En i onmen
Robe o Pinillosa, Samuel Ma cosa, Raul Feliza, Edua do Zalamab,∗
, Jaime
G´omez-Ga c´ıa-Be mejob
aFundaci´on Ca i . Pa que Tecnol´ogico de Boecillo. Pa cela 205. Boecillo, Valladolid, Spain
bIns i u o de las Tecnolog´ıas A anzadas de la P oducci´on. Uni e si y o Valladolid
Abs ac
The long e m e alua ion o he Saca ino obo is p esen ed in his pape . This
s udy aimed o imp o e he obo ‘s capabili ies as a bellboy in a ho el; walking
alongside he gues s, p o iding in o ma ion abou he ci y and he ho el and
p o iding ho el- ela ed se ices. The pape es ablishes a h ee-s age assessmen
me hodology based on he con inuous measu emen o a se o me ics ega d-
ing na iga ion and in e ac ion wi h gues s. Saca ino has been au oma ically
collec ing in o ma ion in a eal ho el en i onmen o long pe iods o ime. The
acqui ed in o ma ion has been analyzed and used o imp o e he obo ’s ope a-
ion in he ho el h ough successi e e inemen s. Some in e es ing conside a ions
and use ul hin s o he esea che s o se ice obo s ha e been ex ac ed om
he analysis o he esul s.
Keywo ds: Social Robo , Se ice Robo , Robo Assessmen , Me ic
1. In oduc ion and backg ound
Se ice obo ics has had a majo p esence in esea ch cen e s in ecen yea s.
Howe e , he e a e ew applica ions whe e obo s a e pa o ou daily li e ac-
i i ies. A numbe o p oblems a ise in he de elopmen o obo s (localiza ion,
na iga ion, planning, in e ac ion, e c.) which ha e been add essed ex ensi ely5
∗Co esponding au ho
Email add ess: [email p o ec ed] (Edua do Zalama )
P ep in submi ed o Robo ics and Au onomous Sys ems Decembe 22, 2015
This is a p ep in e sion. The inal e sion can be ound he e:
Robo ics and Au onomous Sys ems, Volume 79, Pages 40–57, DOI: 10.1016/j. obo .2016.01.014
in esea ch cen e s and ha e been success ully sol ed o a signi ican ex en .
Howe e , he e has been limi ed success in adap ing he solu ions eached o
he de elopmen o obo s ha can ope a e in eal si ua ions o long pe iods
o ime.
The e a e wo main equi emen s ha a obo mus mee o be b ough o10
he ma ke : i mus o e a good se ice a an a o dable p ice, and i mus
pe o m he asks wi h a minimal, ole able ailu e a e. A ele an example o
a obo ha has success ully ul illed bo h equi emen s is he acuum obo ,
led by iRobo Roomba [11]. In addi ion, social se ice obo s ha e o in e ac
wi h humans and he en i onmen in a use - iendly and socially-compa ible15
way. This equi es obus and e sa ile pe cep ion sys ems and solid in e ac ion
s a egies o be de eloped. To da e, much wo k has been dedica ed o hese
esea ch a eas, looking o easy- o-use in e aces h ough which humans can
communica e wi h obo s in a na u al way [19, 4, 6, 14].
Usually, such c i e ia as he abili y o ge and hold he use ’s a en ion o he20
p oposed se ice, e alua ed h ough di ec obse a ion, a e used o assess he
quali y o he social in e ac ion be ween obo s and people [29]. Se e al s udies
ha e ocused on he unde lying easons o he accep ance o social obo s in
di e en scena ios; hei use ulness, adap abili y, enjoymen , sociabili y, com-
panionship o pe cei ed beha io al con ol ha e been iden i ied as impo an 25
pa ame e s o po en ial use s’ accep ance [22]. In [33] hey sugges ha , in
addi ion o his quali a i e assessmen , me ics assessmen should be used o
p o ide eedback mechanisms aimed a imp o ing he gene al pe o mance o
he obo . Benchma king in obo ics [23] has eme ged as a solu ion o e al-
ua e he pe o mance o obo ic sys ems in a ep oducible way and o allow30
compa ison be ween di e en esea ch app oaches. Howe e , benchma king is
a he di icul in se ice obo applica ions [15], gi en ha humans and eal
en i onmen s mus be explici ly conside ed in he benchma king me hodology.
While many wo ks ha e ocused on analyzing obo s wo king in con olled
en i onmen s [28],whe e hey in e ac wi h ew use s, e.g. in Physically In e -35
ac i e RoboGames (PIRG) [24], only a ew long- e m s udies ha e begun o
2
appea o e he las ew yea s. In his sense, an inc eased e o has been made
o unde s and he pa icula i ies o p olonged in e ac ions wi h obo s.
In [30], one o he i s long- e m s udies in a eal-wo ld se ing in ol ing a
social obo is epo ed. The obo was dedica ed o a a ge impai ed use , and40
was e alua ed o e 3 mon hs. Resul s showed ha i is impo an o obo s
imme sed in public spaces o p o ide clea ins uc ions on how o be ope a ed.
The esul s and also aised some issues such as he pe sonali y o he obo , he
dialogue be ween use s and he obo , and he ele ance o g oup collabo a ion.
Ano he ele an example is he obo- ecep ionis Vale ie de eloped by Gock-45
ley e al. and ins alled in he CMU campus. Resul s om a i s s udy [13]
indica ed ha , a e a ce ain pe iod, only ew use s in e ac ed wi h he obo
o mo e han 30 seconds. To a oid his, he au ho s p oposed some design ec-
ommenda ions such as p ope g ee ing and a ewell beha io s, mo e in e ac i e
dialogue o a obus way o iden i ying epea ed isi o s. A second long- e m50
s udy wi h he same obo [18] was ca ied ou o e nine weeks, in which he
obo was able o display di e en moods. Resul s indica e ha in e ac ions
we e di e en depending on he le el o amilia i y and he obo ’s mood: e-
quen use s in e ac ed mo e o en when he obo was in a posi i e mood, bu
he amoun o ime hey dedica ed o he obo was highe when i was in a55
nega i e mood.
In [17] he obo Robo ie is e alua ed in a shopping mall. The obo had
he abili y o adap i s dialogue o p e ious in e ac ions wi h each use , while
i was also capable o o e ing di ec ions and ad e ising speci ic shops and
se ices in he mall. Thei esul s sugges ha use pe cep ion owa ds he60
obo was posi i e, no only in e ms o pe cei ed amilia i y, bu also ega ding
in en ion o use and guidance. Also, be e esul s we e ob ained om epea ed
isi o s. In addi ion, he s udy also concluded ha people’s shopping beha io
was in luenced by he obo ’s sugges ions.
A wide su ey ha add esses he pa icula i ies o p olonged in e ac ions65
wi h obo s can be seen in [20]. This su ey add esses a o al o 24 pape s
o ganized by hei applica ion domain: Heal h Ca e and The apy, Educa ion,
3
Wo k En i onmen s and Public Spaces, and he Home. The expe imen a ion
desc ibed in he said su ey a ied om pape o pape , bu has usually been
done o e se e al sessions and days. Howe e , ew wo ks ha e ex ended o e 70
se e al mon hs. Analysis has been ca ied ou in se e al ways: ideo and di ec
obse a ion, sys em logs and pos - ial in e iews and ques ionnai es. Some
d awbacks ha e been ound ha limi obo pe o mance: i) Robo s lack pe -
cep ual capabili ies o enable ich social in e ac ions and engage spo adic use s,
ii) obo au onomy is o en limi ed, hus p e en ing he obo om ope a ing75
o long pe iods o ime, and iii) pla o ms o en su e om limi ed obus ness
and eliabili y, which esul s in weak suppo ing e idence o he obo ’s e ec-
i eness, while echnological accep abili y is some imes conside ed as one o he
las design s eps.
The p esen pape p o ides he esul s o a long- e m assessmen o a se ice80
obo in a ho el en i onmen . Expe imen s ha e been ca ied ou using Saca ino,
an in e ac i e bellboy obo [37], aimed a p o iding di e en se ices in a ho el:
walking alongside he gues s, p o iding in o ma ion abou he ci y and he ho el
( es au an hou s, menus, e c.), and p o iding ho el- ela ed se ices (calling
axis, guiding gues s o he es au an o o he ooms, e c). Saca ino is designed85
o s ay connec ed o a cha ge in he ho el lobby when i is no doing a speci ic
ask (so i can con inuously p o ide e ec i e se ices) as well as o na iga e
au onomously h ough he ho el acili ies. The app oach p oposed in his pape
in ol es being awa e o cu en limi a ions, a oiding uni e sal solu ions and
es ic ing he applica ion domain o a conc e e use case, ocusing on an i e a i e90
design p ocess. Ou main aim is o ake some o he echnologies in ol ed om
he labo a o y en i onmen o highe echnological eadiness le els.
The es o he pape is o ganized as ollows: A desc ip ion o ou obo
is p esen ed in sec ion 2. The me hodology used o assess he quali y o he
se ices is p esen ed in sec ion 3. The ollowing sec ions 4, 5 and 6, desc ibe he95
di e en s ages o assessmen , including p ocedu e and eedback based on he
analysis o he esul s in each one. The dependabili y o Saca ino is analyzed in
sec ion 7. An enume a ion o lessons lea ned is e iewed in sec ion 8. Finally,
4
sec ion 9 includes he conclusions and u u e wo k.
2. Con ex o Ou Resea ch: Saca ino100
In his pape , he assessmen o Saca ino is p esen ed [37]. This obo has
been designed o ope a e in a ho el. In gene al, h ee le els can be iden i ied
when de ining he s uc u e o a se ice obo (Figu e 1). The i s le el is he
ha dwa e and he mechanical s uc u e, including senso s and ac ua o s. The
second le el comp ises he obo ’s con ol a chi ec u e. A chi ec u e design has105
a ac ed he speci ic a en ion o he scien i ic communi y o e he las ew
yea s, whe e di e en a chi ec u e pa adigms ha e been de eloped (e.g. eac-
i e, delibe a i e and hyb id [12]). A chi ec u e design is a di icul ask as i
comp ises di e en so wa e componen s, p og amming echniques and model-
ing app oaches: na iga ion echniques, simul aneous localiza ion and mapping110
SLAM [8], planning [5], in e aces, human-machine communica ion sys ems in-
cluding dialogue sys ems [21], ecogni ion sys ems [25], cogni i e modeling [35]
and knowledge ep esen a ion [3].
Figu e 1: In as uc u e le els o a se ice obo
5

The hi d le el is he applica ion one. I is a key le el, gi en ha i speci -
ically conce ns he se ices o be p o ided by he obo . Howe e , li le e o 115
has been de o ed o his le el and, ew esul s ha e been ob ained owa ds he
de elopmen o se ice obo ics. The main easons can be ound in:
•A lack o obus ness in p e ious le els and he hie a chy ac oss hem,
which hinde s he de elopmen o se ice applica ions wi h he desi ed
deg ee o au onomy. Fo example, a ailu e in he obo localiza ion may120
cause goods o be d opped du ing loading and unloading, as well as na -
iga ion e o s, collisions and c ashes. While a ce ain ailu e a e may
be accep able in sys ems whe e a human is included in he con ol loop,
he pe missi eness is p ac ically ze o in sys ems ha mus ope a e au-
onomously.125
•A lack o de ini ion a he applica ion le el. Resea che s equen ly ol-
low a bo om-up app oach. They a e usually conce ned abou looking o
uni e sal solu ions wi hou hinking beyond he a chi ec u e le el. Re-
sea ch e o s o en ocus on obo localiza ion, na iga ion, planning o
ace ecogni ion, bu ail in he in eg a ion o di e en echnologies o c e-130
a e use ul applica ions. Thus, we mus combine his bo om-up app oach
wi h a op-down one, whe e he de elopmen is guided by he applica ion
speci ica ions. We mus decide which echnologies a e ma u e enough o
be inco po a ed a he applica ion le el, and pay a en ion o hei lim-
i a ions and he addi ional equi emen s needed o ensu ing ha each135
se ice can be p o ided wi h a high enough quali y le el o be accep ed
by use s. Some examples a e he inclusion o a i icial ma ks in c i ical
a eas whe e localiza ion migh be comp omised o an accu a e posi ion is
equi ed (e.g. loading and unloading a eas), o he inco po a ion o mul i-
modal in e aces (e.g. ouchsc een and oice ecogni ion) in o de o ease140
man-machine communica ion, e en unde high backg ound noise.
•A lack o in ensi e es ing a he applica ion le el. I is necessa y o es ab-
lish app op ia e me ics o quan i ying he obo ’s unc ionali y and he
6
se ice quali y [33]. Fo example, conce ning na iga ion, me ics can p o-
ide a measu e o he na iga ion’s e ec i eness, such as de ia ion om he145
planned ou e, he a ea co e ed by he obo , he obs acles a oided, he
ime aken by he na iga ion asks, o he numbe o imes a human ope a-
o mus in e ene. Conce ning in e ac ion, me ics can p o ide measu es
such as use in e ac ion pe iods, dialogue dep h o use sa is ac ion. Be-
sides, o he ac o s should be e alua ed be o e he commissioning o he150
obo , such as ailu e a e and main enance cos .
Fo ins ance, he success o he Roomba obo owes much o a ligh and
modula mechanical design, a small di e en ial d i e which p o ides a g ea
maneu e abili y and he possibili y o na iga e unde ables and beds, and a
simple senso y sys em o ien ed owa ds na iga ion and he de ec ion o a eas155
whe e di is mo e in ense. Mo eo e , i s a chi ec u e is mainly o ien ed o he
de elopmen o eac i e beha io s. The obo only needs a p ecise localiza ion
when i mus e u n o he cha ging s a ion, o which i uses in a ed beacons
in he s a ion c adle. Fu he mo e, he applica ion laye includes speci ica ions
ha ex end o p e ious le els, such as adap a ion o di e en su aces (ca pe ,160
wood, e c.), co ne cleaning b ush, cleaning ime adap ed o he size o he
oom, and au oma ic sea ch o he cha ging s a ion and subsequen na iga ion
o i (when he ba e y le el d ops below a gi en h eshold).
2.1. Ha dwa e Le el
The Saca ino obo comp ises wo main pa s: a mobile base o na iga ing165
h ough he ho el and an an h opomo phic body o in e ac wi h people and
ho el gues s.
The Saca ino base (see Figu e 2) is con olled by ou double wheels a anged
in a sync od i e con igu a ion. The ou wheels pull and o a e a he same ime,
d i en by wo mo o s, one o ac ion and ano he o u ns. Two eme gency170
s op bu ons on bo h sides o he pla o m allow any use o p e en po en ially
dange ous si ua ions.
7
Figu e 2: Saca ino
Saca ino’s body, moun ed on he mo ing base, suppo s he “social” com-
ponen s o he obo , as well as some na iga ion senso s (a SICK LMS-100
ho izon al lase scanne and a se o ul asonic senso s). The body can o a e175
360 deg ees synch onously wi h he wheels so ha he social pa o he obo
always aces he di ec ion o mo ion.
2.2. A chi ec u e Le el
The de elopmen o Saca ino’s a chi ec u e has been conduc ed unde he
p inciple o componen -based in eg a ion (componen -based obo ics amewo k).180
Figu e 3 shows a unc ional block diag am o Saca ino’s a chi ec u e, including
he linking wi h some applica ion le el unc ionali ies. The e is a se o unc-
ional modules, each one in cha ge o a speci ic ask. Module in eg a ion and
communica ion has been ca ied ou using he ROS amewo k [26]. Modules a e
g ouped in wo majo unc ional subsys ems. The social subsys em includes in-185
e ac ion modules o ges u e con ol (body con ol), isual pe cep ion, cha bo
[1] o gene a e dialogue and an Au oma ic Speech Gene a ion and Recogni ion
8
Figu e 3: Saca ino’s a chi ec u e
sys em [32]. Fu he mo e, his subsys em is esponsible o he beha io o
Saca ino. The obo ’s beha io a ies acco ding o a p ede ined s a e machine.
S a e ansi ion depends on he s imulus ecei ed (p esence o use , in e ace in-190
pu s, execu ion o scheduled ask). An o e all s a e machine diag am is shown
in Figu e 4. The na iga ion subsys em includes such na iga ion modules as
a planning, localiza ion and eac i e na iga ion, and he con ol modules ha
communica e wi h he con olle boa d (p o iding p ope abs ac ion o he
ha dwa e le el comp ising senso s and ac ua o s). When a na iga ion ask is195
equi ed by he beha io module, ei he scheduled o eques ed by he use ,
i sends a na iga ion eques o he na iga ion subsys em, making he obo
change o na iga ion s a e.
2.3. Applica ion Le el
Cu en ly, Saca ino can p o ide he ollowing se ices:200
•Gi ing in o ma ion abou he ho el acili ies. This includes audio- isual
in o ma ion abou he ho el, meal imes and es au an se ices. The
9
•Ques ion: Reques made by he use and sen o he cha bo o subse-
quen p ocessing. The ques ion may be selec ed on he ouchsc een o 325
spoken and ecognized by he ASR (Au oma ic Speech Recogni ion).
•Answe : Answe e u ned by he cha bo , which is displayed on he sc een
and spoken simul aneously (by means o he ex o speech module).
•Topic / A ibu e / Emo ion / Ac ion: Sequence o pa ame e s e u ned
by he cha bo . The opic co esponds o he con ex o he cha bo 330
esponse (e.g., es au an , ime, e c.) and he a ibu es o he speci ic
de ails. Emo ion e e s o he emo ional exp ession gene a ed by Saca ino
du ing in e ac ion, such as joy, ange o neu al. Ac ion epo s whe he
he cu en cha bo esponse has in ol ed a speci ic ac ion o no (e.g.
elling a joke, gi ing headline news, na iga ing o a loca ion, e c).335
Ac ions a e g ouped in o use in e ac ions. An in e ac ion has been consid-
e ed o begin when a use /gues , alone o accompanied by a g oup o people,
makes he i s social ac ion wi h he obo ; he in e ac ion has been conside ed
o inish when he use lea es he in e ac ion a ea o exp essly says goodbye. I
does no ma e i he in e ac ion is s a ed by he obo o no . A ime h esh-340
old be ween in e ac ions has been used o de ec when he use has changed. In
case o doub , changes can be se manually wi h he aid o he images eco ded
by he obo .
The ollowing a iables a e compu ed:
•A e age in e ac ion ime.345
•Ques ions asked mos o en du ing in e ac ions. This allows us o disce n
wha in o ma ion p o ided by he obo is mo e demanded and whe he
use s eally eques in o ma ion o simply play wi h he obo .
•In e ac ion channel: which channel is he mos used in in e ac ions ( ouch-
sc een, oice ecogni ion o bo h).350
16

•Idle ime: ime du ing which he obo is idle wi hou in e ac ing wi h
anyone.
•In e ac ions agains ime o he day: numbe o in e ac ions a e e y hou
o he day, which allows a weekly/mon hly a e age o he numbe o in-
e ac ions h oughou he day o be compu ed. This is o know he imes355
a which Saca ino is busie .
•Numbe o in e ac ions pe day/week: The cus ome p o ile is qui e di -
e en on wo king days (business a ele s) and weekends ( amilies and
ou ism). The e o e, i is in e es ing o analyze sepa a ely bo h kinds o
days.360
3.2. Mo ion Me ics
This sec ion desc ibes he me ics de ined o e alua ing Saca ino’s mobili y
on si e. The obo mus a el o ce ain loca ions a speci ic imes o day
and/o on speci ic da es ( es au an a b eak as ime, mee ing a ea, e c). The
displacemen s a e scheduled by he ho el s a h ough he websi e hos ed by365
he obo . In addi ion o his, any use a any ime can eques he obo o
a el o a gi en loca ion (e.g. o he es au an , he ele a o e c., o guiding
pu poses). Finally, he ho el s a can also ask he obo o a el o a gi en
loca ion. The ollowing da a a e eco ded o each na iga ion ask:
•Na iga ion o igin and des ina ion.370
•O igin and a i al da e and ime.
•Da e and ime he obo s ops.
•Whe he he o de is synch onous (ope a ional) o asynch onous ( eques ed
by he use ).
•Da e and ime o e en s ha esul in he cancella ion o ou e (eme gency375
s op bu on, ouch senso ac i a ion, joys ick con ol.)
17
This in o ma ion is p ocessed o ob ain mobili y me ics, among which we
can men ion:
•Des ina ions eques ed mos equen ly by use s.
•Time pe ou e (which allows si ua ions o be analyzed in which he obo 380
has aken mo e ime han expec ed o each a gi en des ina ion).
•To al na iga ion ime pe day.
•Numbe o imes he na iga ion ask could no be comple ed (due o he
ac i a ion o eme gency s ops, manual con ol, joys ick, e c.).
Conce ning he las i em in he lis abo e, we di e en ia e be ween na iga-385
ion incidences and na iga ion disca ds. Na iga ion incidences e e o he case
in which he obo cancels na iga ion on- he- ly because an unexpec ed e en
has been ound (a s a iona y obs acle has been de ec ed on he obo pa h o
an eme gency s op has been p essed by a use ). In his case, he ho el s a
has o guide he obo manually o a known posi ion ( he cha ging s a ion) and390
emo e he obs acle o elease he s op. Na iga ion disca ds co espond o he
case in which he obo does no ini ia e mo emen because, ei he he ba e y
le el is unde a gi en h eshold o an eme gency s op has been p essed be o e
s a ing o mo e. In gene al, hese disca ds do no co espond o any obo
mal unc ion.395
4. S age I: Se up and S udy on P oxemics
This s age ex ended om Janua y o Oc obe 2012. The implemen a ion o
he obo in he ho el and he de ini ion o some asks and se ices ook place
du ing he i s mon hs o his pe iod. The e alua ion was pe o med by di ec
obse a ion o gues s in e ac ing wi h Saca ino. I is wo h no ing ha when400
in oducing an au onomous obo in a new en i onmen , he usage and wo king
p ac ice o his new o m o echnology by non-expe use s is commonly missing.
De elope s mus make a signi ican e o o e alua e his ci cums ance.
18
This s age ended wi h a pe iod ( om Augus o Oc obe ) du ing which a
numbe o expe imen s on p oxemics and he ex en o in e ac ions we e pe -405
o med [27]. This esea ch analyzed he aspec s o he obo ’s design and be-
ha io ha we e ele an o use engagemen and com o . The expe imen s
ocused on he in luence on p oxemics, du a ion and in e ac ion e ec i eness
o a numbe o dicho omous ac o s ela ed o he obo design and beha -
io : obo embodimen ( ouchsc een wi h/wi hou a obo ic body), s a us o 410
he obo (awake/asleep) and who s a ed communica ion ( obo /use ). The
expe imen s we e done by di ec obse a ion o obo use in e ac ion. The ob-
se a ions we e made wi hou he use s being awa e hey we e being wa ched,
in an a emp o ge he mos na u al condi ions achie able. The da a collec ed
o each in e ac ion was: sample numbe , da e, ime, gende , age es ima ion,415
in e ac ion dis ance, who s a s he in e ac ion, in e ac ion ype (1. The use
in ol ed in he in e ac ion uses speech and ouchsc een o 2. The uses akes a
passi e ole) and o he addi ional commen s. In he expe imen wi h he obo
in he awake s a us and aking he ini ia i e o in e ac ion, we obse ed he
in e ac ion o 95 people wi h Saca ino. F om hose 95 in e ac ions, 53 we e420
held by a single use whe eas 42 we e held by a single use bu accompanied
by mo e indi iduals. 74 we e male and 21 emale. The a e age du a ion o he
in e ac ion o each use was 59.51 seconds.
Some in e es ing conclusions: young and old people seem o eel mo e com-
o able close o Saca ino han middle-aged people; use s end o main ain a425
highe in e ac ion dis ance owa ds an embodied agen e sus only ouchsc een;
embodimen engages use s in main aining longe in e ac ions; in e ac ion ime
inc eases when he obo s a s he in e ac ion. O e all, he ob ained esul s
sugges ed in e es ing guidelines abou how Saca ino should be p esen ed o he
use s in a ho el en i onmen , along wi h o he ine uning ad ice. A de ailed430
desc ip ion o expe imen s and conclusions o his s age can be ead in [27].
4.1. Feedback and imp o emen s a ha dwa e le el
Se omo o s: The expe imen s ha e shown ha child en end o s and close
19
o he obo han adul s. Usually, child en play wi h he obo , g ab i by he
a ms and head and y o o ce hese elemen s, in mos cases wi hou pay-435
ing a en ion o he sc een. These ac ions ha e esul ed in damage o se e al
se omo o s, whose eplacemen has been equi ed. Consequen ly, sp ings and
complian mechanisms ha e been inco po a ed a a m and head join s o p e en
damage when ex e nal o ces a e applied.
Sc een: Ano he conclusion de i ed om he expe imen s has been he need440
o conside mul imodal in e ac ion ( oice and ouchsc een) o mos se ices.
The p oxemic analysis has also shown ha he close in e ac ion dis ance equi ed
by he ouchsc een mus be compa ible wi h he p e e ed pe sonal in e ac ion
dis ance du ing ace o ace oice communica ion. As a esul , la ge on s and
images we e inco po a ed o he sc een, hus equi ing he ini ial 10-inch sc een445
o be eplaced by a 17-inch one.
Au oma ic ba e y cha ging sys em: Some manual in e en ions we e ini ially
equi ed o cha ge he obo . The ho el s a was esponsible o doing his ask
a he obo ’s demand ( he obo moni o s i s ba e y le el). Howe e , he s a
occasionally pos poned o igno ed his demand. So he de elopmen o a ully450
au oma ed cha ging sys em was en isaged (Figu e 8). This cha ging sys em has
been designed acco ding o no m IEC 60204-1 o elec ical appliances and low
ol age (24 ) unde IEC 61140. A se o de ec o s in bo h he obo and he
cha ge , along wi h a p essu e senso , gua an ees ha he connec ions a e ac i e
only when he obo is coupled o he dock, as ecommended by he ISO 13482.455
O cou se, he au oma ic connec ion o he cha ge equi es qui e p ecise
maneu e ing. The e o e, a new guidance sys em was de eloped using an in a ed
ma k a he cha ge s a ion, along wi h wo cha ging con ac s and an in a ed
wiimo e con olle came a on Saca ino’s back. The in eg a ion o a da a ma ix
a he cha ge and a second came a has also been conside ed [7].460
Gy oscope: Du ing he es s, we ha e ound ha he obo loca ion de-
g aded unexpec edly h ough ime. This was a consequence o a small wheel
misalignmen in he synch od i e pla o m and esul ed in odome y d i and a
subsequen de e io a ion o he localiza ion and na iga ion sys ems. The SLAM
20
Figu e 8: Au oma ic ba e y cha ging sys em
sys em was able o co ec his e o in sho displacemen s in he labo a o y,465
bu his was no he case in he long e m expe imen a ion in he ho el. The e-
o e, a gy oscope was added o he obo . This de ice p o ides in o ma ion
abou he Roll, Pi ch and Yaw angles o he obo base. The Yaw angle is
in e p e ed as a ela i e angle and is in eg a ed wi h odome y da a o co ec
he said odome y d i s.470
4.2. Feedback and Imp o emen s a A chi ec u e Le el
When he obo was wo king o many hou s, some lack o obus ness was
de ec ed a he au oma ic eco e y om ce ain unexpec ed obo /en i onmen
s a es. This p oblem was sol ed by e ining he ROS a chi ec u e o he obo .
The inal unc ional diag am o he a chi ec u e has al eady been shown in475
Figu e 3.
Localiza ion and Na iga ion: Na iga ion is a undamen al ask o Saca ino,
bo h o accompanying gues s h ough he ho el dependencies and eaching he
loca ions chosen by he ho el s a . In gene al, na iga ing om A o B equi es
he obo o de e mine whe e i is (A), whe e i is equi ed o be (B), how480
i should ge he e (pa h planning), and how o deal wi h s a ic and dynamic
en i onmen al ac o s such as obs acles encoun e ed on he way.
21

The localiza ion sys em ini ially ope a ed on a ange-lase onboa d he obo
(see sec ion 2.1 and he odome y da a, h ough a lase mapping sys em. How-
e e , we e i ied ha loca ion deg aded h ough ime, especially unde dynamic485
condi ions (people mo ing abou he obo ). The addi ion o a gy oscope, al-
eady desc ibed in he p e ious sec ion, led o a la ge imp o emen . Fu he -
mo e, a opological map wi h 7 des ina ions and waypoin s was selec ed o
planning pa hs, ins ead o he gene al ee space map p e iously employed.
The use o his opological map has simpli ied pa h planning and has p o ided490
p ecise con ol o e he pa hs and he a eas h ough which he obo mo es.
Figu e 9 shows he lase map o he ho el g ound loo used o localiza ion and
na iga ion and he opological map wi h he di e en goals.
Reac i e na iga ion is also used, so ha Saca ino s ops a obs acles and
wai s un il hey a e emo ed. O cou se, ci cum en ing he obs acle and e-495
calcula ing he ajec o y would also be possible, bu his ac ion is o en less
e ec i e because i can lead he obo o lea e he ope a ing a ea, he e o e
mee ing unexpec ed si ua ions. In addi ion, na iga ion is cu en ly es ic ed
o he g ound loo , whe e he e a e no s ai s o gaps ha migh comp omise
Saca ino’s s abili y. The in eg a ion o senso s o de ec s ai s and elec onics500
o emo ely manage ele a o s will be add essed in he nea u u e in o de o
ex end na iga ion o o he loo s.
4.3. Feedback and Imp o emen s a Applica ion Le el
Beha io : The esul s o ou concu en esea ch on p oxemics and in e -
ac ion [27] show ha a obo in he awake s a us and aking he ini ia i e o 505
in e ac ion clea ly a o s use engagemen . I was he e o e p oposed, o he
second and hi d s ages, ha Saca ino should be ’awake’, wi h i s a ms sligh ly
ben a he elbow, he head held high and he eyes open and li up. The obo
should andomly make gen le mo emen s wi h i s a ms and i s head, and he
sc een should be on and displaying he main menu sc een.510
Howe e , while Saca ino is a he cha ging s a ion, i should be ’asleep’,
wi h i s a ms in an ex ended posi ion, he head held high and he eyes open,
22
Figu e 9: Topological map o g ound loo wi h des ina ions
bu wi hou mo ing.
In bo h cases, when he lase senso de ec s a pe son wi hin a 3 me e
semici cle a ound Saca ino, i should look in he di ec ion o he app oach and515
make a g ee ing o inci e in e ac ion. The g ee ing should include sen ences
such as “Hello” and “Come close and alk o me”.
In e ace: Acco ding o he use e iew in S age I, he in o ma ion displayed
on he ouch sc een was imp o ed. A mo e a ac i e in e ace was de eloped,
simpli ying he in o ma ion displayed on i and designing mo e in ui i e menus520
and bu ons. Figu e 10 shows an example o he esul ing in e ace.
Mo eo e , in o de o inc ease he obo ’s (ba e y) au onomy and ack he
esul s o he p oxemics s udies, a sc eensa e showing andom pic u es o he
ho el should be launched when he e is no pe son nea he obo . Saca ino
would immedia ely espond when a pe son comes close o i , pu ing he main525
menu on he sc een and making a g ee ing o inci e in e ac ion. In addi ion, i
was no iced ha mos ho el gues s did no ac ually know wha se ices we e
o e ed by he obo . The e o e, a Help menu was added o he in e ace, o
show speci ic ins uc ions.
23
Figu e 10: Saca ino’s in e ace
Dependabili y: The p esence o he obo in he ho el mean ex a wo k o 530
main enance pe sonnel. These pe sonnel we e in cha ge o u ning he obo on
and o and connec ing i o he ba e y cha ge (when equi ed). The la e
can be a he demanding because he ba e y au onomy may be less han 4
hou s, depending on he asks ca ied ou by he obo . (In pa icula , asks
in ol ing na iga ion esul in impo an ba e y consump ion). The au oma ic535
cha ging sys em has allowed his o be deal wi h, he ho el main enance s a
only becoming esponsible o u ning he obo on/o . The obo gene a es a
na iga ion ask o he cha ging s a ion a any low ba e y le el de ec ion and
plugs i sel in au oma ically.
5. S age II: Me ics Assessmen 540
This s age ex ended om Sep embe 2013 o Janua y 2014. Du ing he i s
wo mon hs, es s we e ca ied ou unde ou di ec obse a ion. The o e all
ope a ion o he obo and he imp o emen s in oduced in eedback I we e
checked (in pa icula , hose conce ning he na iga ing asks and he maneu e s
24
o each he cha ging s a ion). Fo he las wo mon hs o he cu en s age,545
Saca ino was ope a ed in no mal condi ions in he ho el wi hou he p esence
o ou esea ch eam. The ho el main enance s a was esponsible o u ning
he obo on and o , while he obo e u ned and au oma ically plugged i sel
in o he cha ge s a ion a any ba e y le el d op. The me ics de ined in 3 we e
eco ded o assess he abili ies o he obo in ela ion o he se ices p o ided.550
As p e iously men ioned, Saca ino has been designed o s ay connec ed o a
cha ge in he ho el lobby when no doing a speci ic ask, so i can con inuously
p o ide e ec i e se ices, and o na iga e h ough he ho el’s acili ies. A web-
si e hos ed by he obo (Figu e 11) allows he s a o p o ide he obo wi h
e en in o ma ion and o schedule na iga ion asks. The e a e 2 p ep og ammed555
ba e y cha ging asks a day, du ing which na iga ion is disabled.
Figu e 11: Saca ino’s schedule
All he da a we e aken du ing a 60 day pe iod, including wo king days
and weekends. Saca ino was u ned on o 23 o hese days (acco ding o he
wo kload o he ho el main enance se ice manage ). Table 1 epo s he o e all
ope a ion da a. The obo ope a ed o 220 hou s, 9.56 hou s on a e age pe 560
25
in he p esence o o he people, he speech ecogni ion a io is a ec ed by he680
ollowing e idence:
•Ho el en i onmen is e y noisy, in con as wi h he lab en i onmen .
•Use s end o main ain a pe sonal dis ance om he obo .
•Saca ino was con inuously analyzing he ecei ed sound, looking o any
oice eques .685
In o de o deal wi h hese ac s, we ha e added a ouch- o-lis ening mech-
anism so he use mus s a ecogni ion by ouching a bu on on he in e ace.
This me hod, al hough less na u al han di ec speech, is e ec i e and well
accep ed by use s who o en use i o in e ac wi h hei mobile de ices.
Na iga ion: Conce ning incidences du ing he na iga ion asks, he as 690
majo i y o cases we e no caused by localiza ion inaccu acy, poo planning o
cha ging maneu e ailu e, bu by he said misuse o he eme gency bu ons o
he p esence o s a ic obs acles in he obo ’s way. The wai ing imeou s we e
uned highe in o de o sol e some o hese ci cums ances. Clea sc een and
oice messages ha e been added oo, in o de o in o m use s abou an imminen 695
na iga ion cancella ion i hey s ay in he obo ’s pa h.
5.5. Feedback and imp o emen a applica ion le el
Beha io (Eme gency bu ons): I is ema kable ha he eme gency s ops
we e ac i a ed du ing 12% o he o al unning ime o Saca ino (mos imes
wi hou any ac ual eason). The e o e, a ques ion ha a ises is how o ac when700
an eme gency bu on is p essed:
•Saca ino can con inue wi h i s social skills ac i a ed, in e ac ing wi h peo-
ple and disca ding only na iga ion asks when hey a e eques ed. Wi h
his beha io , Saca ino is a ailable o in e ac and p o ide se ices o
longe , bu main enance s a ha dly no ices ha he obo needs assis-705
ance.
32

•On he con a y, i any ope a ional e o occu s, all Saca ino’s capabili-
ies can become blocked, hus no allowing any in e ac ion. Saca ino can
p o ide only sc een and oice messages epo ing he need o assis ance
so ha he ho el s a can quickly no ice he need (di ec ly o h ough a710
gues ’s wa ning). Howe e , he obo may con inue o p o ide no se ice
o an ex ended ime unnecessa ily.
In he p esen wo k, we ha e aimed a maximizing he obo ’s unc ionali y, so
we ha e chosen he i s op ion.
6. S age III: Re inemen 715
This s age ex ended om May o Augus 2014. Du ing his pe iod, Saca ino
ope a ed in he ho el in he same condi ions as in S age II. The imp o emen
o he obo ’s pe o mance was de i ed om he modi ica ions in oduced in
eedback II, measu ed using he same quan i a i e me ics, h oughou he same
pe iod o ime (60 calenda days). The ho el main enance s a was esponsible720
o u ning he obo on/o and, when necessa y, eco e ing i om ope a ional
e o s (by b inging i back o he cha ge s a ion).
Saca ino was u ned on 51 days (acco ding o he wo kload o he ho el
main enance-se ice manage ), while in S age II i was u ned on only 23 days.
This inc ease comes om he imp o emen s in oduced in he p e ious s ages,725
which ha e esul ed in enhanced se ice and deg ee o dependabili y (dec eas-
ing main enance pe sonnel equi emen s). The e was also a sligh dec ease in
unning ime (abou 1 hou a day o e 8 hou s), bu he o al unning ime was
abou wice ha o he p e ious s age (435 hou s e sus 220).
The o e all ope a ion da a in his S age III is summa ized in Table 2. Mo e-730
o e , we ha e disc imina ed be ween wo king days and weekends ollowing he
ho el manage ’s sugges ion. I is wo h no ing ha he gues p o ile is qui e
di e en on wo king days (gene ally business a ele s) and weekends ( amilies
and ou ism). The e o e, a sepa a e analysis is use ul o adap ing he obo
se ices and schedule o he day ype.735
33
Table 2: O e all ope a ion ime da a summa y in S age III
Days To al To al social To al ime To al numbe To al numbe A e age
obo wo king in e ac ion in mo ion social use s/ in e ac ion
wo king ime ime asks ac ions in e ac ions du a ion
Wo king days 43 350.22 h 35.7 h 6.35 h 9232 677 2.44 min
(mon- i) (8.14 h/day) (10.2%) (1.8%) (214.69/day) (15.74/day)
Weekends 8 84.76 h 12.39 h 0.31 h 3616 209 2.61 min
(sa -sun) (10.5 h/day) (14.6%) (0.3%) (452/day) (26.1/day)
To al 51 434.98 h 48.09 h 6.66 h 12848 866 2.58 min
S age III (8.52 h/day) (11%) (1.5%) (251.9/day) (17.3/day)
6.1. Social in e ac ion
The e was a signi ican inc ease in he obo use in S age III, as can be seen
in able 2. A o al o 12848 social ac ions om 886 di e en use s ook place,
which ep esen s mo e han wice hose co esponding o he p e ious s age
( able 1). Saca ino was in e ac ing wi h people o 11% o he o al ime (4% in740
S age II) and was unning na iga ion asks o e 1.5% o he o al ime (1.11%
in S age II). Thus, he ime du ing which Saca ino was busy inc eased om 5%
o 12%. The o he 88% o he ime, Saca ino was awai ing use demands.
The obo ope a ed on 43 wo king days (Monday o F iday) and 8 weekend
days (Sa u day and Sunday). On a e age, Saca ino’s unning ime (pe day)745
was highe on weekends han on wo king days. Mo eo e , mo e use s in e ac ed
wi h he obo on weekends and o mo e ime, as was expec ed. Conce ning
na iga ion, he obo spen mo e ime a he cha ge on weekends because, in
gene al, he e we e no na iga ion asks scheduled on hese days. The imp o e-
men s in oduced in eedback II also esul ed in a dec ease in he numbe o 750
obo eboo s o less han once a day.
The a e age numbe o social ac ions h ough he day ime is epo ed in
Figu e 17. Saca ino was busies a abou 10am and 10pm. The a e noon max-
imum happened wo hou s la e han in he p e ious s age due o he di e en
ime o he yea (sp ing and summe in he cu en s age). The a e age use in-755
e ac ion du a ion inc eased om 1.54 o 2.58 minu es, hanks o he inc eased
34
Figu e 17: S age III: Numbe o social ac ions e sus ime o he day.
pe o mance and smoo hness o he in e ac ion.
The numbe o social ac ions, conside ing he day ype, is shown in Figu e 18.
The dis ibu ion o hese ac ions h oughou he day is also di e en depending
on he day ype. Noon, a e lunch and mid-a e noon a e p e e ed on weekends.760
Figu e 18: S age III: Numbe o social ac ions e sus ime o he day (wo king days e sus
weekends)
The dis ibu ion o he se ices eques ed by he use s (o e 51 days) is
epo ed in Figu e 19.a. The cha is simila o ha o he p e ious s age (Figu e
13). In o ma ion abou Saca ino i sel is s ill he mos eques ed opic (43%).
Howe e , “Dialog” has inc eased signi ican ly. This opic conce ns in o ma ion
ha is only accessible h ough oice (no h ough he ouchsc een). Ce ainly,765
his inc ease owes much o he imp o emen o he speech ecogni ion sys em
add essed in S age II. In ac , oice in e ac ion inc eased om 0.05% in he
p e ious s age o 30% in he cu en one, as can be seen in Figu e 19.b. This
inc ease ep esen s a signi ican ad ance, and u u e e o s should ocus on
35
making u he imp o emen s, because oice is conside ed he mos na u al way770
o in e ac ion o humans.
Figu e 19: S age III: a) Topics dis ibu ion. b) In e ac ion channel
6.2. Na iga ion
The na iga ion asks ep esen abou 1% o he o al unning ime o he
obo (as in he p e ious s age). Howe e , a o al o 240 na iga ion asks
we e launched, which ep esen s wice he numbe o na iga ion asks launched775
in S age II. Thei dis ibu ion is shown in Figu e 20. 5.65% (13 asks) we e
disca ded be o e s a ing, which compa es a o ably o he 16% ound in S age
II. Mo eo e , he asks we e no comple ed in 15 cases (6.52%) due o incidences
p oduced du ing execu ion. In o al, 212 na iga ion asks we e success ully
accomplished, i.e. 88%, hus esul ing in an inc ease o 16% wi h espec o780
S age II.
I is wo h emembe ing ha disca ds do no ac ually co espond o a na i-
ga ion sys em mal unc ion, bu o a good managemen o he obo ’s p io i ies
and i s s a e machine. In ac , jus 227 o he na iga ion asks equi ed mo e-
men , 94% o which we e comple ed success ully (while only 6 % inished due785
o an incidence).
Disca ded asks d opped om 16% in S age II o 6% in he cu en s age
(see Figu e 20), which sugges s ha he eme gency s ops emained p essed o
36
Figu e 20: S age III: Na iga ion asks
less ime ( hanks o he added LED indica o s).
The e ec o he ba e y cha ging sys em modi ica ions (a ha dwa e and790
a chi ec u e le els) is also no iceable. The maneu e ailu es dec eased om 57%
in he p e ious s age o 28% in he cu en one. The connec ion/disconnec ion
maneu e s we e no accomplished only 4 imes.
Conce ning he pa h acking, on- he-way ask cancella ions a e simila (in
%) o hose o S age II. The mos equen eason o cancel na iga ion (43%, 3795
imes) was an obs acle s aying in he obo ’s way o an ex ended ime.
Figu e 21: a) Des ina ions. b) E ec i eness na iga ion
The 212 na iga ion asks comple ed in his s age co e ed a o al dis ance o
2,490 me e s (mo e han wice he dis ance co e ed in he p e ious s age). The
dis ibu ion o he na iga ion asks, acco ding o hei des ina ions, emains
37

simila o ha o he p e ious s age, as can be seen in Figu e 21.a. The cha ge 800
con inues o be he mos common des ina ion.
Finally, he na iga ion e ec i eness inc eased om 73% o 88%, as can be
seen in Figu es 15.b. and 21.b. This means ha he na iga ion sys em imp o ed
signi ican ly hanks o he modi ica ions made in eedback II: Saca ino cancels
i s na iga ion due o he p esence o obs acles less han be o e.805
7. Main enance
In his sec ion, he dependabili y o Saca ino is analyzed, hus p o iding
in e es ing guidelines aimed a he ma ke ing o his echnology. The obo mus
pe o m a eal se ice wi h accep able obus ness and dependabili y wi hou
equi ing cos ly main enance o de elope s’ assis ance.810
The assessmen p esen ed in his sec ion co e s s ages II and III, he obo
ope a ing in no mal condi ions in he ho el wi hou he a endance o he de el-
ope s. The se up pe iod in ol ing he de ini ion o se ices and he cons uc ion
o he localiza ion and na iga ion map has no been included.
The in oduc ion o he au oma ic ba e y cha ging sys em g ea ly inc eases815
Saca ino’s au onomy. The ho el main enance s a is only esponsible o u ning
he obo on and o . Mo eo e , when a mal unc ion happens, gene ally du ing
na iga ion, a ho el employee has o check whe he he eme gency s ops a e
p essed o Saca ino is blocked a a s a iona y obs acle. Then, he/she mus
unlock he eme gency s op (in he o me case) o mo e he obo o an open820
a ea (in he la e ) and, e en ually, o he cha ging s a ion i Saca ino is unable
o elocalize ( he na iga ion sys em is es a ed a he cha ging s a ion).
Some ha dwa e p oblems could no be sol ed by he ho el s a and equi ed
ou assis ance. Du ing S age II, he s a epo ed ha Saca ino’s body had been
damaged and los some o i s mo emen s. The neck il ing join (up/down mo e-825
men s), he igh a m se omo o and an eyelid se omo o had been damaged
due o ex e nal o ces (applied gene ally by child en). Sp ings and complian
join s had been included du ing eedback II so he p oblems no longe happened
38
in S age III.
Fu he mo e, he ho el s a wice epo ed ha he obo could no be830
u ned on. The eason was ha he ba e y had go ully discha ged due o a
mal unc ion o he p essu e senso onboa d he obo . The p oblem was sol ed
by imp o ing he docking sys em ha dwa e and ela ed so wa e du ing eedback
II. The o al cos o man-hou s employed on main enance is a ound 22 hou s.
A he end o s age III, he ho el s a has no epo ed any o he need o 835
assis ance, hanks o he desc ibed imp o emen s.
8. Lessons Lea ned
As a esul o he ield expe imen a ion app oached in his wo k, some im-
po an lessons ha e been lea ned ha may guide esea che s and de elope s o
se ice obo s. These lessons a e summa ized below.840
•The obo mus be ully au onomous. In a i s expe imen a ion s age,
Saca ino was placed a he cha ge s a ion when i s ba e ies we e low,
bu manual connec ion o he cha ge s a ion was equi ed. This ask,
al hough simple and quick, was some imes o e looked by he ho el main-
enance s a (due o hei o he du ies), hus limi ing he unning ime845
o he obo . The de elopmen o a ully au oma ic cha ging s a ion has
allowed his p oblem o be o e come. Mo eo e , Saca ino’s au onomy has
inc eased signi ican ly, gi en ha he obo can cha ge du ing idle pe i-
ods, e en while in e ac ing wi h use s. A manual on/o ou ine is s ill
used because i allows he obo o be disconnec ed du ing nigh s ( hus850
educing powe consump ion) and can be assumed by non- echnical pe -
sons (e.g. he ecep ionis s). O cou se, i desi ed, his ope a ion could
be au oma ed by using a simple elay-based ci cui pe manen ly supplied
wi h a low powe .
•Na iga ion mus be obus , hus a oiding localiza ion losses ha may p e-855
en he obo om eaching i s des ina ion. A de ailed in- ield analysis is
39
ecommended o add ess his issue. Many cases can be sol ed by changing
he obo ’s ou es o a oid undesi able si ua ions ( ou es passing h ough
la ge ea u eless a eas o nea windows, mi o s, one-legged ables and
o he u ni u e, c owded a eas, e c.). Some landma ks can also be added860
o he en i onmen o educe localiza ion e o s. In ou case, we added a
da a-ma ix and some in a- ed LEDs o he cha ging s a ion o guide he
cha ging maneu e . Mo eo e , he obo localiza ion is es a ed a he
cha ging s a ion (uniquely iden i ied wi h he said da a-ma ix) o a oid
cumula i e e o s.865
•Voice in e ac ion is especially challenging because o he cu en limi a-
ions o speech ecogni ion and he noisy ho el en i onmen . Mul i-modal
in e ac ion sys ems allowing bo h speech and ouch in e ac ion is ecom-
mended. In his case, i is also necessa y o econcile he close in e ac ion
dis ance equi ed by he ouchsc een wi h he pe sonal dis ance used in870
ace- o- ace oice in e ac ion. The equi ed comp omise can be alle ia ed
by including la ge ouchsc eens and on s. In addi ion, he in e ac ion
should include help and eedback mechanisms so ha he use can know
abou he obo ’s capabili ies and i s scope o unde s anding.
•Ano he ele an aspec is how o ini ia e he in e ac ion. I is ecom-875
mended ha he obo g ee s and inci es in e ac ion when a pe son is
de ec ed nea by. Howe e , ou obse a ions ha e shown ha speaking oo
loud may cause eluc ance in some gues s, gi en ha hey become he
cen e o a en ion o he su ounding people, which may no be pleas-
an . S a ing in e ac ion wi h a g ee ing olume adap ed o he gues ’s880
dis ance (which can be measu ed by he lase o o he dis ance senso )
a o s in e ac ion.
•The mos eques ed in o ma ion om Saca ino has been abou he obo
i sel (i s capabili ies and unc ionali y), ollowed by eques s abou he
ho el acili ies and se ices. The echnological no el y o he obo a ouses885
gues s’ cu iosi y, which eases engagemen and can be exploi ed. Howe e ,
40
added alue se ices mus be inco po a ed o he obo o p e en use s
om losing in e es o e ime. Rega ding he eques ed se ices, he mos
popula one is accompanying gues s h ough he ho el acili ies. This owes
much o he expec a ion gene a ed by obo mo emen . Howe e , i is890
assumed ha o he se ices, such as axi calls, check-ou o i em deli e y,
will gain ele ance in he u u e, p o ided ha dependable beha io is
a ained.
•Child en ge close o he obo han adul s, as expec ed, and end o
play se e ely wi h i . They g ab he obo by i s a ms ying o mo e895
hem and manipula e he head and ela ed elemen s (eyelids, mou h), in
mos cases, wi hou paying any a en ion o he sc een. The e o e, special
ca e should be aken o build a obus obo ic pla o m o p e en damage.
Sp ings and complian join s a e ecommended o p o ec he obo agains
ex e nal o ces and, he e o e, minimize main enance equi emen s. The900
mos c i ical poin s a e he a ms and he neck.
•The obo mus be accep ed no only by use s bu also by he ho el s a ,
looking o hei complici y and in ol emen in he in eg a ion o he obo
in he ho el’s e e yday ac i i y. O cou se, he s a should no pe cei e
he obo as an incon enience o a h ea o hei jobs. Mo eo e , an easy-905
o-use in e ace should be p o ided so ha he s a can emo ely plan he
obo ’s ac i i y and modi y in o ma ion con en s (menu o he day, e en s,
mee ings and con e ences, e c).
•The obo mus p o ide added- alue se ices. I is able o a ac g ea
a en ion om use s, bu hey lose in e es o e ime i he obo can-910
no o e any hing else. Quan i iable, added- alue se ices ha can be
amo ized in a sho pe iod o ime mus be p o ided o in oduce his
echnology in he ma ke .
•A inal issue o ca dinal ele ance conce ning he ma ke ing o his kind o
obo s is ha hey mus comply wi h all sa e y s anda ds o human obo 915
41
[36] Wisspein ne , T., an de Zan , T., Iocchi, L., and Schi e , S. (2009).1080
RoboCup@Home: Scien i ic compe i ion and benchma king o domes ic se -
ice obo s. In e ac ion S udies, 10(3):392–426.
[37] Zalama, E., Ga c´ıa-Be mejo, J. G., Ma cos, S., Dominguez, S., Alonso,
R. F., Pinillos, R., and L´opez, J. (2013). Saca ino, a se ice obo in a ho el
en i onmen . In ROBOT 2013: Fi s Ibe ian Robo ics Con e ence - Ad ances1085
in Robo ics, Vol. 2, Mad id, Spain, 28-29 No embe 2013, pages 3–14.
48

Table 3: O e all assessmen summa y
S age I Se up and S udy on P oxemics
Resul s
Use s: A e age in e ac ion du a ion:
95 0,59 minu es
Feedback
Le el Weaknesses Imp o emen s
Ha dwa e Se omo o s damage s due o misuse. Add complian mechanism.
Use s in e ac ion dis ance is no
close .
Inc ease ouchsc een size (17-inch).
Lack o obo au onomy . De elop au oma ic ba e y cha ging
sys em.
Odome y d i s. Add Gy oscope.
A chi ec u e Na iga ion and localiza ion is no
good enough .
Use opological map o planning a-
jec o ies .
Saca ino s ops a obs acles and wai s
un il hey a e emo ed.
Applica ion Beha io . The obo akes he ini ia i e o in-
e ac ion.
Lack o simplici y and cla i y o he
in e ace.
Design menus and bu ons mo e in u-
i i e.
S age II Me ics Assessmen s
Resul s
Use s: A e age in e ac ion du a ion:
349 (15,17/day) 1,54 minu es
Feedback
Le el Weaknesses Imp o emen s
Ha dwa e Mic ophone: inpu noise om he
head mo o s.
Mic ophone is eloca ed.
The ho el s a can eadily no ice
ha he Eme gency bu ons ha e been
p essed.
Add LEDs s a us .
A chi ec u e The speech ecogni ion a io is low. Include ouch- o-lis ening bu on on
he in e ace o s a ecogni ion.
The majo i y o incidences du ing he
na iga ion asks due o p esence o ob-
s acles.
Inc ease imeou s be o e cancella ion.
Add clea sc een and oice messages
o p e en use be o e cancella ion.
Applica ion Beha io . Saca ino can con inue wi h i s social
skills e en hough he eme gency bu -
ons a e p essed
S age III Re inemen
Resul s
Use s: A e age in e ac ion du a ion:
349 (15,17/day) 1,54 minu es
49