INTRODUCTORY LECTURE
UVODNO PREDAVANJE
Proceedings 2010. (2011), Vol 2, ISSN 1986-8154
www.sportkon.com
SPORT I VEŠTAČKA INTELIGENCIJA
SPORTS AND ARTIFICIAL INTELLIGENCE
\orđe Stefanović1
Fakultet sporta i fizičkog vaspitanja, Beograd, Srbija
Faculty of Sport and Physical Education, Belgrade, Serbia
1
INTRODUCTORY LECTURE
doi: 10.5550/SP.2.2010.04
UDK: 796.371.214(437.6)(4)
UVODNO PREDAVANJE
COBISS.BH-ID: 2246936
Summary
Sažetak
The rapid development of computer technologies influenced
the necessity to introduce changes in sport domain. Artificial
intelligence represents a branch within computational science which has the task to create computers that can reason in
a way similar to human behavior.
The aim of this work is to explain how artificial intelligence
can support sport in an increasingly complicated and demanding sports environment. Due to a large number of efficient
and reliable solutions, several areas in the domain of artificial intelligence have become relevant to the process of sport
development.
Some of them will be dealt with in this work: expert systems,
game theory, neurological networks and software agent/based
modeling.
Nagli razvoj kompjuterskih tehnologija uticao je potrebu da
se uvedu promene i u sferi sporta. Ve{tačka inteligencija
predstavlja računarsku disciplinu čiji je zadatak da stvori
računare koji mogu da rezonuju na način sličan ljudskom
rezonovanju.
Ideja ovoga rada je da se objasni kako ve�������������������
{������������������
tačka inteligencija pomaže (mogućnosti i olak������������������������������
{�����������������������������
ice) sportu u sve komplikovanijem i zahtevnijem sportskom okruženju. Usled velikog
broja efikasnih i pouzdanih re{enja pojavile su se vi{e
oblasti ve���������������������������������������������������
{��������������������������������������������������
tačke inteligencije koje su postale bitne u procesu razvoja sporta.
Neke od njih će biti predmet razmatranja u ovom radu:
ekspertni sistemi, igre, neuronske mreže i softverski agenti.
Key Words: sport, artificial intelligence, expert systems,
games theory, neurological networks, software agents.
Ključne riječi: sport, ve��������������������������������������
{�������������������������������������
tačka inteligencija, ekspertni sistemi, igre, neuronske mreže, softverski agenti.
Introduction to the area
of the artificial intelligence
Uvođenje u prostor
ve{tačke inteligencije
Artificial Intelligence (AI) is a branch of computer science
with the objective to create computers with the ability of
reasoning in a similar way to human beings. The AI deals
with the development of programs which show intelligence
similar to the human’s. Basically, AI means processing of
knowledge, not of data.
The objective of AI in sports is to explain how its application
may help in more and more complicated and demanding
sports environment. In other words, the idea of this paper was
to show the possibilities and conveniences which are provided
by the application of the AI in sports. In our days, it is expected to have a quick reaction to changes and problems in sports,
both on the research and practice level. AI-based systems allow,
at least in some segments, a quicker and more accurate obtaining of necessary information, and, which makes them different, to completely substitute routine tasks of a coach in some
cases. Although computers are far from being able to imitate
human thinking, they stand out by logical processes – which
Garry Kasparov (world chess champion) could see for himself
in 2003 during the chess match against the computer Deep
Junior (Figure 1).
AI trend is reflected in several aspects. Some experts find
that the technology will go so far that it will allow a Human
Nervous System to be connected to a computer. It has been
already evidenced that human beings could control a com-
Ve{tačka inteligencija (Vi) predstavlja računarsku disciplinu
čiji je zadatak da stvori računare koji mogu da rezonuju na
način sličan ljudskom rezonovanju. Ona izučava razvoj
programa koji pokazuju inteligenciju nalik na ljudsku. U
osnovi, Vi podrazumeva procesiranje znanja, a ne podataka.
Cilj Vi u sportu je da se objasni kako njena primena pomaže u sve komplikovanijem i zahtevnijem sportskom okruženju. Drugim rečima, ideja pisanja bila je da se prikažu
mogućnosti i olak{ice koje daje primena Vi u sportu. Danas
se očekuje brza reakcija na nastale promene i probleme
na polju sporta, kako na istraživačkom planu tako i u praksi. Sistemi zasnovani na Vi omogućavaju, bar u nekim
delovima, tačnije i brže dobijanje potrebnih informacija,
a ono {to ih čini posebnim, je to {to u nekim slučajevima
i potpuno zamenjuju rutinske zadatke trenera. Iako su
kompjuteri daleko od opona{anja ljudskog razmi{ljanja,
oni se ističu logičkim postupcima – u {ta se Gari Kasparov
(svetski prvak u {ahu) uverio 2003. tokom {ahovskog meča
s kompjuterom „Dip Džunior“ (Deep Junior) (Slika 1).
Trend Vi se ogleda u nekoliko aspekata. Neki stručnjaci
smatraju da će tehnologija otići toliko daleko da će omogućiti da ljudski nervni sistem bude povezan sa kompjuterom.
Već je pokazano da ljudi mogu upravljati kompjuterima
22
Stefanović, \.: SPORT I VE[TAČKA INTELIGENCIJA
Zbornik radova 2010, 22-33
puter with nothing but their thoughts, using electrodes attached to the scalp, or wires connected to their nerves. Such
technology could help in designing perfect artificial limbs,
or hands, and enable making more complex movements.
Improvements of our bodies are the area of interest for the
AI. As a result, prosthetic body parts nowadays serve solely
for reconstructive purposes; they restore body function of
their users, which was lost due to illness or injury.
Some believe that the Evolution of Cyborgs (half machineshalf humans, as in the movie “Terminator”) is not only
inevitable, but also necessary. They perceive a future in
which people will have to either integrate with machines or
to accept the domination of robots. Others imagine future
humans who would improve their abilities using external
devices instead of implants. Maybe we will become surrounded by more and more gadgets, such as palm computers
or smart phones, in enormous wireless network which would
enable us to be constantly connected to the Internet and to
each other.
In 2020, a home will have its own machine for virtual reality – virtual Robo-Pal (Figure 2). This machine will allow
family members to play dangerous sports such as mountain
climbing, or bungee-jumping, but also to visit exotic places
during their virtual vacation, all that with one click. A virtual station joints together the mind-control technology and
artificial sensor feedback.
Rising trend of miniaturization will eventually result in miniature computers that can be sewn into clothes. Scientists are
even now developing computers which can be washed, made
of fabrics which conduct current. These fabrics contain a
microchip, and they are supplied by solar energy, or by devices producing electricity from body movements. Figure 3 shows
computerized clothes – the “smart” jacket, which contains a
microchip, keyboard made of fabrics, on one of the sleeves.
It is supposed that sportsmen will be able to use it too.
Which subfields are within the AI? A large number of efficient and reliable solutions were invented and improved. This
explains why AI is divided into a number of subfields, from
shape recognition to artificial life, including evaluative calculation and planning. The syncretism of AI and sports can
INTRODUCTORY LECTURE
UVODNO PREDAVANJE
Slika 1: Gari Kasparov u {ahovskom meču s kompjuterom
Figure 1: Garry Kasparov in the chess match against the computer
samo pomoću misli, koristeći elektrode pričvr{ćene za kožu
lobanje, ili žice koje su povezane s njihovim nervima. Takva
tehnologija mogla bi pomoći u pravljenju savr{enih ve{tačkoh
udova, ili {aka, omogućavajući sve složenije pokrete.
Usavr{avanje na{eg tela je prostor u kome je Vi na{la interesovanje. Tako protektički delovi tela danas služe isključivo
u rekonstruktivne svrhe; svojim korisnicima vraćaju telesnu
funkciju koja je izgubljena usled bolesti ili povrede.
Neki smatraju da evolucija kiborga (bića koja su pola ma{ine,
a pola ljudi, npr. u filmu “Terminator“) nije samo neizbežna,
već i neophodna. Oni vide budućnost u kojoj će ljudi morati, ili da se spoje s ma{inama, ili da prihvate dominaciju
robota. Drugi zami{ljaju buduće ljude koji će unaprediti
svoje sposobnosti putem spoljnih naprava umesto implantata. Možda ćemo postati okruženi sve većim brojem
spravica, kao {to su kompjuteri veličine {ake i inteligentni
telefoni, ogromne bežične mreže putem kojih ćemo neprestano biti u univerzalnom kontaktu sa Internetom i jedni s
drugima.
Dom iz 2020. imaće vlastitu ma{inu za virtuelnu stvarnost
– virtuelni robopal (Slika 2). Pomoću nje, članovi porodice
moći će da upražnjavaju opasne sportove kao �������������
{������������
to su planinarenje, ili skakanje na bandžiju, kao i da posećuju egzotična mesta tokom virtuelnog godi{njeg odmora, a sve to na
pritisak dugmeta. Virtuelna stanica spaja tehnologiju upravljanja mislima i ve{tačku senzornu povratnu spregu.
Trend minijaturizacije na kraju će proizvesti siću{ne računare
koji će moći da se u{ivaju u odeću. Naučnici čak razvijaju
računare koji se mogu prati, a koji su napravljeni od tkanina
koje provode struju. Te tkanine sadrže mikročip, a napajaće
se solarnom energijom, ili uređajima koji stvaraju elektricitet
od pokreta tela. Na Slici 3 je prikazana kompjuterizovana
odeća – „pametna“ jakna, koji sadrži mikročip, tastaturu od
materijala na jednom rukavu. Pretpostavka je da će moći da
je koriste i sportisti.
Koje oblasti čine prostor Vi? Veliki broj efikasnih i pouzdanih
re{enja se pojavio i unapredio. Ovo daje obja{njenje za{to
je Vi podeljena na vi{e grana, od prepoznavanja oblika, do
ve{tačkog života, uključujući evolutivno izračunavanje i
planiranje. Sinkretizam Vi i sporta moguće je da se objasni
23
INTRODUCTORY LECTURE
UVODNO PREDAVANJE
Stefanović, \.: SPORTS AND ARTIFICIAL INTELLIGENCE Proceedings 2010, 22-33
be explained through a short introduction to fields such as:
expert systems in sports, electronic games, neural networks
and software agents.
kroz kraće upoznavanje sa oblastima kao {to su: ekspertni
sistemi u sportu, elektronske igre, neuronske mrže i softverski
agenti.
Slika 2: Virtuelni robopal
Figure 2: Virtual Robo-Pal
Slika 3: „Pametna jakna“
Figure 3: „Smart jacket“
Expert systems and sports
Ekspertni sistemi i sport
The demand for Expert Systems (ES) has been growing as the
rhythm of people’s life became faster; on the other hand,
the “loss” of time in waiting for experts in specific fields to
come and try to solve or help in solving certain problems.
ES is a computer program which emulates problem solving
in the way an expert does it. In order to be named an ES, a
program has to contain expert knowledge in the specific field and to provide automated reasoning. ES is developed
using relevant software tools (shell). The development of ES
involves an expert, knowledge engineer and a user (Figure
4).
An Expert (lat. expertus) is a professional, adept, master in a
given subject. Who is an expert in sports? Usually a distinguished professor – scientist/researcher among internationally
recognized national and foreign university lecturers. His/her
role is to: “lend” (give, transmit) knowledge and assist in the
verification (testing) of ES’s knowledge. Problems may occur
if the expert is: inaccessible, uncommunicative; if he tends
to emphasize the obvious and cannot remember everything.
Knowledge engineer is a person who interviews the expert
and collects (“retrieves) knowledge from him/her. Then the
person is supposed to select appropriate techniques for
presenting the knowledge and appropriate techniques for
deduction; to select the development tools, formalize,
formulate and systemize the expert’s knowledge and finally
to test the ES. A Sports Technologist is to sports is what a
knowledge engineer is to the AI.
A User utilizes the finished ES, participates in making requests, and can also participate in testing and writing the ES
documentation. In sports, users are usually a coach and a
sportsman.
ES can be applied to various tasks in sports:
Potreba za ekspertnim sistemima (Es) se povećavala kako se
ubrzavao tempo života ljudi, a sa druge strane skratilo se
„gubljenje“ vremena u čekanju da dođu ekpserti određenih
profila i poku�������������������������������������������
{������������������������������������������
aju da re���������������������������������
{��������������������������������
e ili pomognu oko određenih problema. Es je računarski program kojim se emulira re{avanje
problema na način na koji to čini ekspert. Da bi neki program
mogao da se nazove Es mora da sadrži ekspertsko znanje iz
neke oblasti i da omogućava automatizovano rezonovanje.
Es se razvija kori{ćenjem odgovarajućih softverskih alata
(shell). U razvoju Es učestvuju ekspert, inženjer znanja i
korisnik (Slika 4).
Ekspert (lat. expertus) je stručnjak, poznavalac, majstor u
čemu. Ko je ekspert sportu? To je najče{će istaknuti profesor
– naučni radnik iz redova međunarodno priznatih domaćih
i inostranih univerzitetskih nastavnika. Njegova uloga je da:
„pozajmljuje“ (daje, prenosi) znanje i pomaže pri proveri
(testiranju) znanja Es. Problemi su ako je ekspert: nedostupan,
nekomunikativan, sklon tome da ističe očigledno i ne može
da se seti svega.
Inženjer znanja (engl. knowledge engineer) je čovek koji vodi
intervju sa ekspertom i od njega prikuplja („izvlači“) znanje.
Zatim vr{i izbor odgovarajućih tehnika za predstavljanje znanja,
vr{i izbor odgovarajućih tehnika za zaključivanje, vr{i izbor
razvojnog alata, formalizuje, formuli{e i sređuje ekspertovo
znanje i na kraju testira Es. Ono {to predstavlja inženjer znanja
u Vi, to je tehnolog sporta u sportu. Profil takvog čoveka je
značajan za sport, jer je bitno da postoji neko ko je obrazovan
i da može da prenosi znanja iz nauke/teorije u praksu sporta.
Korisnik koristi gotov Es, učestvuje u formiranju zahteva, a
može da učestvuje u testiranju i pisanju dokumentacije za
Es. U sportu su korisnici Es najče{će trener i sportista.
Es mogu da se primene na različite zadatke u sportu:
•Stvaranje podsetnika i generisanje alarma. U tzv. ril-tajm
(real-time) situacijama, Es povezan sa monitorom može
24
Stefanović, \.: SPORT I VE[TAČKA INTELIGENCIJA
Zbornik radova 2010, 22-33
•Creating reminders and generating alerts. In so-called
real-time situations, ES connected to a monitor may warn
about changes in the player’s condition. For example, in
the beginning of preparations of a football team, the
coach gave to the players a task to move around the field in the aerobic zone of load intensity with specified
TE-TA tasks. If one of football players is not within the
required range of pulse load, the coach will remind him
to increase or decrease his motion speed.
•Assistance in establishing diagnoses. When the case of a
sportsman is complex, rare or the person establishing a
diagnosis simply does not have enough experience, ES
may help in getting to the most probable diagnosis based
on the information.
•Analysis and planning of training. ES can either search for
contradictions, errors and failures in the existing training
plan and program, or it may use the analysis of the specific sportsman’s condition for determining a treatment,
based on the adopted training instructions.
•Agents for finding information. Software agents may be
sent to search and deliver information, e.g. on the Internet. Agents have knowledge on user’s preferences and
needs, and they may also have some knowledge on sports,
in order to be able to evaluate the relevance and usefulness of their findings.
• Recognition and interpretation of images. Many images
in the form of curves (e.g. Pulse load) can be automatically interpreted, including standard and more complex
images.
Details about a sportsman are inputs, and outputs represent
the diagnosis – recommended training methods, means and
loads during a training. ES were not created with a view to
replacing completely professionals, coaches, but rather as
big support. This is the way in which Omega Wave Sport
Technology System (OW) works. This system carries out the
analysis of a sportsman / team in respect of body and mental loads for the purpose of the optimization of the training
process. OW analyzes the electrical activity in the heart and
slow brain waves in order to create an “internal image” of
how the sportsman body functions. Thereat, OW does this
very fast, in a non-stressful and non-invasive way. For the
first time ever, sportsmen, coaches, physiologists and scientists can monitor cycles of stress and recovery of their players
on daily basis. With OW, it is nowadays possible to follow,
INTRODUCTORY LECTURE
UVODNO PREDAVANJE
Slika 4: Učesnici u razvoju i korišćenju ES u sportu
Figure 4: Participants in the development and utilization of ES in sports
da upozori na promene stanja sportiste. Npr. trener je na
početku priprema fudbalske ekipe zadao da se svi igrači
kreću po terenu u aerobnoj zoni intenziteta opterećenja
sa definisanim TE-TA zadacima. Ukoliko se neko od
fudbalera ne nalazi u zadatom opsegu pulsnog opterećenja, trener ga opomene da poveća ili smanji brzinu
kretanja.
•Asistencija u postavljanju dijagnoza. Kada je slučaj sportiste kompleksan, redak ili osoba koja postavlja dijagnozu, jednostavno, nema dovoljno iskustva, Es može pomoći
da se dođe do najverovatnije dijagnoze na osnovu podataka.
•Analiza i planiranje treninga. Es može, ili da traži protivrečnosti, gre�������������������������������������������
{������������������������������������������
ke i propuste u postojećem planu i programu treninga, ili može da koristi analizu specifičnog stanja
sportiste za određivanje tretmana koja je bazirana na
prihvaćenim trenažnim uputstvima.
•Agenti za pronalaženje informacija. Softverski agenti
mogu biti poslati da traže i da dostavljaju, donose informacije, npr. sa Interneta. Agenti sadrže znanje o korisnikovim preferencijama i potrebama, i mogu takođe da
imaju i znanje iz sporta, kako bi bili u mogućnosti da
procene važnost i korisnost onoga {to su na{li.
•Prepoznavanje i interpretacija slika. Mnoge slike u obliku
krivulja (npr. pulsnog opterećenja) mogu da budu automatski interpretirane, od standardnih do kompleksnijih
slika.
Podaci o sportisti su ulazi, a izlazi su dijagnoza – preporučeni trenažni metodi, sredstva i opterećenja na treningu.
Es nisu kreirani sa namerom da potpuno zamene stručna
lica, trenere, već kao velika ispomoć. Tako funkcioni{e npr.
Omega vejv (OW) sportsko tehnolo{ki sistem (Omega Wave Sport Technology System). Ovaj sistem vr{i dijagnozu
jednog sportiste/tima u vezi telesnog i psihičkog opterećenja radi optimalizacije trenažnog procesa. OW analizira
električnu aktivnost u srcu i sporim moždanim talasima, da
bi stvorio „unutra{nju sliku“ kako telo sportiste funckioni{e.
Pri tome, OW to čini brzo, nestresno i ne-invazivno. Po
prvi put ikada, sportisti, treneri, fiziolozi i naučnici mogu
da nadgledaju (monitor) cikluse stresa i oporavka svojih
sportista na dnevnoj bazi. Sa OW, danas je moguće pratiti, kod svakog sportiste, njegovu adaptaciju na zahteve
treninga, takmičenja, putovanja i ostale stresove. Putem
25
INTRODUCTORY LECTURE
UVODNO PREDAVANJE
Stefanović, \.: SPORTS AND ARTIFICIAL INTELLIGENCE Proceedings 2010, 22-33
for each sportsman, his adaptation to requirements of training, matches, travels and other stressful situations. A proper
managing of stressful and recovery cycles allow coaches to
train the adaptive response from a player, so that he/she
adapts much faster than before. In this way, OW correlates
the sports science/theory with sports practice.
Lately, the application of computer systems in a sport training
is more and more present. Can a computer (ES) replace a coach
in managing the system of sport preparation? There are fitness
centres where its member can simply enter into a computer
what he/she wants to achieve, and then it makes the plan and
program for him/her. As the member goes through various
machines (stations), the computer adjusts the plan and program
to the member’s reactions. What if, someday, achievements
in Information Technologies would be so high that we could
create a single ES which would be homomorphous (of the
same shape), analogue, even isomorph (similar) to a coach?
Could it guide a sportsman to the final success/results at a
competition?
In order to develop such ES, we should find an algorithm (a
series of individually interconnected procedures resulting in
a solution of a problem) which would provide for a coach
to manage the system of preparation of sportsmen. Thereat,
the plan and program, and any corrections thereto, are just
the result of such algorithm, which is aimed to that the desired condition is achieved, set by a model. This is primarily
the matter of determinism (the determinism being here
perceived as a pre-known behaviour of a system, conditioned by its current state), which can be applied to lifestyle
and not only to a sports training. Is a man’s consciousness
(awareness) just one complex algorithm? We firmly believe
that it is not, otherwise there would be no free will. This has
direct implications to the issue man against machine in
managing training. How a machine can be constructive,
creative minded? A machine can learn, adapt, but it can
never be an artist (inventive creator) and have free will like
a sports coach. We should keep in mind that managing a
system of sports preparation is the art, in its large part, as
well as the science is. A machine can never be an artist.
dobrog upravljanja stresnih i oporavljajućih ciklusa, moguće
je trenirati adaptivni odgovor sportiste, tako da se on/ona
adaptira mnogo brže nego pre. Na taj način OW povezuje nauku/teoriju sporta sa sportskom praksom.
U poslednje vreme prisutna je sve veća primena računarskih
sistema u sportskom treningu. Može li računar (Es) da zameni trenera u upravljanju sistemom sportske pripreme?
U nekim fitnes klubovima, član jednostavno u računar
unese {ta želi da postigne, a zatim mu on pravi plan i
program. Kako član prolazi kroz razne ma{ine (stanice),
tako računar koriguje plan i program prema reakciji člana.
Šta ako bi jednog dana, dostignuća u informatičkoj tehnologiji bila tako visoka da se kreira jedan Es koji bi bio homomorfan (istog oblika), analogan, pa čak izomorfan (sličan)
treneru? Da li bi on mogao bolje da vodi sportistu do konačnog uspeha/rezultata na takmičenju?
Da bi se napravio jedan takav Es, morao bi da se pronađe
algoritam (niz pojedinačno povezanih postupaka koji dovode do re{enja nekog problema) kojim trener upravlja
sistemom pripreme sportista. Pri tome su plan i program,
kao i njihovo korigovanje, samo rezultat tog algoritma, koji
ima za cilj da dostigne željeno stanje postavljeno od strane
modela. Ovo je pre svega stvar determinizma (ovde se determinizam shvata kao unapred poznato pona{anje nekog
sistema koje je uslovljeno njegovim trenutnim stanjem),
koji može da se primeni i na život sam, a ne samo na sportski trening. Da li je čovekova svest samo jedan kompleksan
algoritam? Mi čvrsto držimo da nije, jer onda ne bi postojala slobodna volja. Ovo ima direktne implikacije na problem
čovek protiv ma{ine u upravljanju treningom. Kako ma{ina
može da bude kreativna, da ima stvaralački duh? Ma{ina
može da uči, da se adaptira, ali nikad neće moći da bude
umetnik (kreativni stvaralac) i da ima slobodnu volju kao
sportski trener. Mora se zapamtiti da je upravljanje sistemom
sportske pripreme u jednom dobrom delu umetnost, kao
{to je i nauka. A ma{ina nikad ne može da bude umetnik.
Electronic Games
Elektronske igre (Ei) predstavljaju igre kod kojih se koristi
elektronika u cilju formiranja interaktivnog sistema sa kojim
se igra. Danas je najprisutnija forma elektronskih igara u
obliku video igre (Vig) – Ei koje se baziraju na interakciji sa
igračem putem korisničkog interfejsa i formiranju odgovora na video uređajima. Prema postojećim klasifikacijama
postoji preko dvadesetak različitih vrsta Vig od kojih su
najpoznatije: arkadne igre, avanturističke, borilačke, simulacije, edukativne igre i za nas relevantne sportske igre.
Sportske video igre (SVig) su video igre koje simuliraju tradicionalne sportove. Najveći broj sportova danas poseduje i
svoju elektronsku varijantu, bilo da se radi o timskim sportovima, atletici, ili ekstremnim sportovima. SVig predstavljaju psihofizičke i taktičke izazove za igrača i stavljaju na
proveru njegovu preciznost i tačnost. One su uglavnom
bazirane prema složenim modelima kretnih aktivnosti aktuelne grane/discipline sporta, uključujući brzinu, snagu,
ubrzanje, preciznost pa čak i izdržljivost. Igre su sme{tene
u virtualni svet sličan onom iz stvarnog života: igraju se u
okruženju stadiona, sportskih dvorana, ili na otvorenom
prostoru. Modeli SVig su veoma verno razvijeni tako da se
sa uspehom mogu koristiti za inicijalno obrazovanje, ili
obnavljanje znanja vezanog za pravila same igre. Igre mogu
da obuhvataju jedan-na-jedan interakciju sa virtuelnim
protivnikom, strategiju izbora i pripreme u slučaju igara
Electronic games (EG) are games which employ electronics
to create an interactive system with which a player can play.
The most common form of electronic game today is the
video game (Vig) – EG based on the interaction with a player
using a user interface to generate feedback on video devices.
According to the existing classifications there are over twenty different types of Vig, where the most popular are: arcade games, adventure games, fighting games, simulation
games, educational games and, the most relevant for us,
sports games.
Sports Video Games (SVig) are video games that simulate the
playing of traditional sports. Most sports have today their
electronic variant, including team sports, athletics and extreme sports. SVig involve physical and tactical challenges for
a player, and test his/her precision and accuracy. The games
are mostly based on models of motion activities and actual
sports branch/discipline, including speed, strength, acceleration, accuracy and even endurance. The games are placed
in the virtual world similar to the real one: they take place
in a stadium, arena or outdoors. SVig models are very realistically developed so that they can be successfully used for
the initial education or for recapitulation of the knowledge
related to the rules of that specific gameplay. Games can
26
Elektronske igre
Zbornik radova 2010, 22-33
include one-on-one interaction with a virtual rival, strategy vezanih za timske sportove, ili simulacije stvarnih situacija
of selection and preparation in case of games related to u sportu i oko sporta.
team sports, or simulation of real situations in sports and Kompanija Nintendo je 2006. izdala Vi sport kolekciju – Vsk
related to sports.
(Wii Sports colection) SVig za Vi (Wii) konzolu. Na ovoj
In 2006, the company Nintendo released Wii Sports Col- platformi igrač mora fizički da pokreće svoj daljinski upravljač
lection – WSC for the Wii console. On this platform a player kako bi simulirao pokret svog avatara (virtuelnog alter ega
has to move physically its remote control in order to simu- samog igrača). Ovim je svet video igara polako iz svere silate the movement of his avatar (virtual alter ego of the mulacije evoluirao ka emulaciji sveta sportske igre. Vsk
player). In this way, the world of video games slowly evolved poseduje mogućnost igranja pet različitih sportova: boksa,
from the area of simulation towards emulation of sports kuglanja, golfa, tenisa i bejzbola. U toku igre, igrač (sportista)
games. WSC has the possibility of playing five different sports: prati progres razvoja svojih ve��������������������������������
{�������������������������������
tina i koristi treninge na razboxing, bowling, golf, tennis and baseball. In the course of ličitim nivoima razvoja.
the game, a player (sportsman) follows the progress of the Iza uspeha sa Vsk Nintendo je 2008. izdao i Vi fit (Wii Fit)
development of his/her skills and makes benefit of trainings koji pruža mogućnost vežbanja aerobika i rekreativnih vežon different levels of the development.
bi. Poseduje namenski razvijenu balans platformu (Wii BaAfter the success of WSC, Nintendo also released Wii Fit in lance Bord) preko koje je obezbeđena jo{ neposrednija
2008, which gives aerobic and recreation exercises. It has a interakcija igrača – sada već rekreativnog sportiste i platforspecially developed balance platform (Wii Balance Board) me.
which allows a more indirect interaction between a player
Slika 5: Kibernetički model upravljanja pripremom sportista bez sportskog trenera
Figure 5: Cybernetic model of managing the preparation of sportsmen without a coach
– now of a recreational sportsman - and the platform.
Neural Networks
Neural networks (NN) can be defined as the manner in which
the nervous system in humans is organized. In a wider sense,
they represent the potential for modeling a synthetic (computer)
network by utilizing the principles of organization of a human
neural network. At the lowest level of complexity, NN serve as
an intermediary between the somatosensory and motor system
of an organism going in both directions. If we look at the physiological processes of creating an impulse in the sensory system
as a whole, we will see that all the mechanisms involved in
those processes are basically of electromagnetic nature, for they
serve to cause the signal to be above the activating threshold of
a neuron. In order for these issues to be understood and properly investigated, it is necessary to acquire the knowledge about
the relevant space of bioelectric activity of a neuron or a larger
cluster of neurons, pertaining to:
•different kinds of sense receptors and the manner of their
activation,
•the generation and spread of a nerve impulse,
Neuronske mreže
Neuronske mreže (Nm) predstavljaju način organizacije
nervnog sistema kod čoveka. [ire značenje predstavlja
mogućnost modelovanja sintetičke (kompjuterske) mreže
kori{ćenjenjem principa organizacije ljudske neuronske
mreže. Na najjednostavnijem nivou složenosti, Nm služe
kao posrednik između somatosenzornog i motoričkog sistema organizma u oba smera. Ukoliko se globalno posmatraju fizički procesi nastanka impulsa u senzornom sistemu,
uočavamo da su svi mehanizmi pri tim procesima u osnovi elektromagnetne prirode, jer isti treba da uzrokuju signal
iznad akcionog praga neurona. Da bi se ova problematika
razumela i pravilno proučavala neophodno je razumeti
relevantni prostor bioelektrične aktivnosti jednog ili veće
grupe neurona koji se odnosi na:
•vrste čulnih receptora i način njihove aktivacije,
•generisanje i prostiranje nervnog impulsa,
•pasivni i aktivni membranski transport jona (Na+, K+,
Cl–),
27
INTRODUCTORY LECTURE
UVODNO PREDAVANJE
Stefanović, \.: SPORT I VE[TAČKA INTELIGENCIJA
INTRODUCTORY LECTURE
UVODNO PREDAVANJE
Stefanović, \.: SPORTS AND ARTIFICIAL INTELLIGENCE Proceedings 2010, 22-33
•passive and active membrane transport of ion (Na+, K+,
Cl–),
•electrical activities in the brain (electroencephalograph,
EEG) and
•magnetic activities in the brain (magnetoencephalograph,
MEG).
•moždane električne aktivnosti (elektroencefalogram, EEG)
i
•moždane magnetne aktivnosti (magnetoencefalogram,
MEG).
The analysis of bioelectric signals themselves makes use of
gamma-function mathematical transformations (from simpler
to very complex ones). The electrical activities in the brain at
the level of NN as well as the ones related to the brain waves
are also relevant to various psychological functions, such as:
recognition, memorizing, learning, thinking, creativity, consciousness, language faculties, etc.
What is the nature of the interaction between an organism and
the environment? Biological systems (organisms) constantly interact with their environment. They continuously exchange
matter and energy with it. This openness of biological systems
enables biological development. However, the organisms interact
with the environment by means of the senses, through which
they constantly receive new information in different forms, which are further converted into nerve impulses. The nerve impulses
are then processed in hierarchical neural networks CNS, where
they are interpreted also as more complex psychological experiences.
The input of information into the nervous system is enabled
by sense receptors, which register sensory stimuli, such as
touch, pain, cold, heat, sound, light, taste, smell, etc. Receptors turn sensory stimuli into nerve impulses – basic mechanisms of conversion. All sense receptors have one feature in
common – receptor potential. The immediate impact of any
specific stimulus activating the receptor is manifested as a
change in potential on the receptor membrane.
When the receptor potential rises above critical level in the
nerve fibre attached to the receptor, it results in the creation
of action potential. In this process, the more the receptor
potential rises above the threshold of action potential, the
higher the frequency of action potential, which is one of the
factors influencing the experience of intensity of a sensory
stimulus in the brain. The other factor is the number of activated receptors above the threshold, which is determined by
the number of sensory nerve fibers transmitting nerve impulses
into the primary sensory zone of the cortex.
A case from practice
It is important for a sportsman or a sportswoman to possess
highly developed proprioceptive abilities. The manner in
which a human organism perceives shapes is perfectly exploited by the great masters of martial arts (e.g. dr Masaaki
Hacumi from Japan, the head of the organization Bujinkan
or Morihej Ueshiba, the founder of Aikido), but it is very
poorly investigated in scientific circles. It is very unusual that
they manage to overcome several adversaries at a time simultaneously and in a relatively short space of time, while
their movements seem much slower than the reasonably
expected maximum (Figure 6). One of the most demanding
defense scenarios among all “one on one” systems in the Far
Eastern tradition of martial arts is the bare-handed defense
from an assailant armed with a samurai sabre. Analytically
speaking, without going into detail concerning the nature of
the skill, chances of “mere” survival are measured in terms
of thousandths, i.e. they are certainly close to zero even if
they come up against a barely adept swordsman.
In the practice of the great masters the chances are significantly
higher. Namely, according to the tennets of proprioceptive
abilities of a human organism, the slightly surprising order in
28
Analiza samih bioelektričnih signala raspolaže gamom
matematičkih transformacija (od jednostavnijih do veoma
robustnih). Moždane električne aktivnosti na nivou NM i
moždanih talasa značajne su i za psiholo{ke funkcije:
prepoznavanje, memorisanje, učenje, mi{ljenje, kreativnost,
svest, reči...
Kako izgleda interakcija organizma i okoline? Biolo{ki
sistemi (organizmi) nalaze se u neprekidnoj interakciji sa
okolinom. Oni sa njom neprekidno razmenjuju materiju i
energiju. Zbog ove otvorenosti biolo{kih sistema omogućen
je i biolo{ki razvoj. Međutim, organizmi sa okolinom su u
interakciji posredstvom čula, preko kojih neprekidno dobijaju informacije u različitoj formi, koje se dalje konvertuju u nervne impulse. Ti nervni impulsi se potom obrađuju
u hijerarhijskim neuronskim mrežama CNS gde se interpretiraju i u složenije psiholo{ke doživljaje.
Ulaz informacija u nervni sistem omogućavaju čulni receptori, koji registruju čulne draži, kao {to su dodir, bol, hladnoća,
toplota, zvuk, svetlost, ukus, miris itd. Receptori pretvaraju
čulne draži u nervne impulse – osnovni mehanizmi konverzije. Svi čulni receptori imaju jedno zajedničko obeležje –
receptorski potencijal. Neposredni učinak bilo koje specifične draži koja pobuđuje receptor se manifestuje u obliku
promene potencijala na receptorskoj membrani.
Kada receptorski potencijal poraste iznad praga akcionog
potencijala u nervnom vlaknu priključenom receptoru,
počinje pojava akcionih potencijala. Pri tome, učestalost
akcionih potencijala utoliko je vi{a, ukoliko receptorski
potencijal vi{e nadma{uje nivo praga akcionog potencijala,
{to je jedan od faktora koji utiču na doživljaj intenziteta
čulne draži u mozgu. Drugi je broj pobuđenih receptora
iznad praga, {to određuje broj senzornih nervnih vlakana
koji prenose nervne impulse u primarne senzorne zone
korteksa.
Slučaj iz prakse
U sportu je bitno da sportista poseduje na vrlo visokom
nivou propriocepcijske sposobnosti. Način na koji ljudski
organizam opaža oblike praktično je savr{eno iskori{ćen
od strane vrhunskih majstora borilačkih ve{tina (npr. dr
Masaaki Hacumi iz Japana, poglavar organizacije Buđinkan
ili Morihej Ue���������������������������������������������
{��������������������������������������������
iba, osnivač aikidoa) je veoma malo istraživan u naučnim krugovima. Veoma je čudno {to oni uspevaju da savladaju i po nekoliko protivnika simultano u
relativno kratkom vremenskom intervalu, dok im pokreti
izgledaju prilično sporiji od realno očekivanog maksimuma
(Slika 6). Jedan od najtežih scenarija za odbranu u svim
sistemima „jedan na jedan“ u tradiciji istočnjačkih borilačkih ve������������������������������������������������
{�����������������������������������������������
tina je odbrana golim rukama od napadača naoružanog samurajskom sabljom. Analitički gledano, bez detaljnijeg ulaženja u prirodu ve{tine, {anse za „puko“
preživljavanje se mere promilima, tj. svakako su veoma
blizu nuli ukoliko je „preko puta“ iole ve{t mačevalac.
U praksi vrhunskih majstora ova verovatnoća značajno
raste. Naime, prema postulatima propriocepcijskih sposobnosti ljudskog organizma, redosled zapažanja parametara
objekata se kreće pomalo neočekivanim redosledom u tri
faze: I faza – pomeranje objekta, II faza – oblik objekta i
which the parameters of an object are perceived includes three
phases: the first phase – the change of location of an object;
the second phase – the shape of an object; and the third phase
– the colour and relief of an object. In connection with that, we
should note the common fact that the objects that are moving
faster are more quickly observed than the slower ones. In order
to trick the assailant, a master makes rather slow moves, always
keeping his silhouette unchanged in the eyes of the assailant,
and by constant movement reaching a more favourable distance (surely a shorter one in this scenario). Such concept is sometimes used unawares by top dribblers/defense players in collective sports and thus deserves a much deeper analysis as well as
a wider application. One illustrative example is the dribbling of
Dejan Bodiroga – it was always met with controversial comments,
pointing out that he “was not fast enough, but somehow he
Zbornik radova 2010, 22-33
III faza – boja i reljef objekta. S tim u vezi, važi napomena
da objekti koji se brže kreću bivaju pre uočeni od sporijih.
Da bi prevario napadača, majstor pravi pokrete koji su
prilično spori, permanentno održavajući svoju siluetu neizmenjenu u očima napadača, svojim stalnim kretanjem
dostižući povoljniju distancu (svakako kraću pri ovom
scenariju). Ovaj koncept je ponekad nesvesno iskori{ćen
od strane vrhunskih driblera/odbrambenih igrača u kolektivnim sportovima i zaslužuje znatno detaljniju analizu i
{iru primenu. Jedan od tih primera je dribling Dejana
Bodiroge – uvek je postojao kontroverzni komentar da on
„nije dovoljno brz, ali da nekako uvek prođe“, {to ga je
naravno dovelo u rang najboljeg duel igrača svoga vremena.
Slika 6: Primena propriocepcijskih sposobnosti kod samurajske tehnike
Figure 6: The application of proprioceptive abilities in the samurai techniques
always manages to get through”, which naturally put him in the
rank of one of the best opposition players of the time.
How does the body move from the standpoint of spacial efficacy
of a motor neuron? Generation of a motor signal comes from
the direction of the cortex, goes down along the spinal cord,
depending on which cluster of motor neurons is responsible for
the innervation of corresponding muscle fibres. Since the transmission speed of nerve impulses is limited, it is obvious that the
time of propagation of a nerve impulse from the cerebral cortex
to the activating neuron is in direct proportion to the distance
of motor end plate from the corresponding spinal vertebra. This
parameter is greatly important for the cultivation of fast movements. By regular body training one should become accustomed
to activating the muscular structure along the spinal cord at the
motor end plate of the furthest innervating neuron when making
any kind of complex movement. For instance, if in volley-ball a
player wants to hit the ball from the initial position with his
hands down, he will first innervate the muscles for lifting the
shoulders, then the upper arms, followed by the forearms, and
only lastly those of the hands and fingers (of course, an adequate reaction of the leg muscles is taken for granted, but again
applying the same principle – moving from the region of the hip
downwards).
Apart from minimizing the time for appropriate reaction, adhering to this principle allows for subsequent adjustements if the
situation changes abruptly. In the above-mentioned example, if
Kako se izvodi pokret sa stanovi{ta prostorne efikasnosti
motoričkog neurona? Generisanje motoričkog signala kreće
se iz pravca korteksa niz kičmeni stub u zavisnosti koja
grupa motoričkih neurona je određena za invervaciju odgovarajućih mi{ićnih vlakana. Po{to je brzina prostiranja
nervnih impulsa ograničena, očigledno je da je vreme propagacije nervnog impulsa od kore velikog mozga do aktivacionog neurona proporcionalna udaljenosti kraja inervacionog neurona od odgovarajućeg kičmenog pr{ljena. Ovaj
parametar je od izuzetnog značaja za kultivisanje brzih
pokreta. Kroz redovan telesni trening trebalo bi stvarati
naviku da se pri svakom složenom pokretu aktiviraju mi{ićne
strukture uz samu kičmu ka kraju najdaljeg invervacionog
neurona. Npr. ukoliko u odbojci igrač želi smečovati loptu
iz početnog položaja spu�����������������������������������
{����������������������������������
tenih ruku, on će najpre inervirati mi{iće za podizanje ramena, potom nadlaktice, zatim
podlaktice i tek na kraju {ake i prstiju (naravno, adekvatna
reakcija mi{ića nogu se podrazumeva, ali uz po{tovanje istog
principa – krećući se od oblasti kuka pa nadole).
Pored minimizovanja vremena pravilne reakcije, po{tovanjem
ovog principa ostavlja se mogućnost naknadnog prilagođavanja ukoliko se situacija naglo promeni. U gore
pomenutom primeru ukoliko je lopta npr. okrznula vrh
mreže, organizam nije jo{ uvek napravio fine pokrete
29
INTRODUCTORY LECTURE
UVODNO PREDAVANJE
Stefanović, \.: SPORT I VE[TAČKA INTELIGENCIJA
INTRODUCTORY LECTURE
UVODNO PREDAVANJE
Stefanović, \.: SPORTS AND ARTIFICIAL INTELLIGENCE Proceedings 2010, 22-33
the ball happened to make contact with the top part of the net,
the organism still had not made fine movements of the fingers,
the hand and maybe even the forearm, but only that of the
shoulders, for example, so the potential for correction of movements is optimal.
prstiju, {aka i možda podlaktice, već samo npr. ramenskog
dela, tako da je mogućnost za korekciju pokreta maksimalizovana.
Ovaj princip je opravdan i sa stanovi���������������������
{��������������������
ta energetske efikasnosti, {to u slučaju dugih i iscrpljujućih mečeva predstavlja
presudan faktor. Naime, ovim se izbegavaju brojne „preterane“ reakcije spoljnih delova ekstremiteta. Golman u
fudbalu npr. pomera najpre ramena pripremajući se za
potencijalnu reakciju. U sledećem trenutku njegov čulni
sistem opaža da dalji pokret nije neophodan, jer lopta ide
preko gola. Pomeranje ostalih mi���������������������������
{��������������������������
ića ruku nije vi����������
{���������
e neophodno. Ukoliko bi se invervirala celokupna ruka, ta bi energija bila uzaludno potro{ena.
U čemu se ogleda su{tina moždane hijerarhijske neuronske
mreže i kognitivne implikacije? Danas preovlađujuća naučna paradigma jeste da se procesiranje informacija na nivou
CNS odigrava posredstvom hijerarhijski organizovanih i
povezanih neuronskih mreža. Npr. vizuelna informacija
prvo se hijerarhijski procesira na nivou mrežnjače (počev
od mreže fotoreceptornih čepića i {tapića, pa do mreže
ganglijskih ćelija), da bi se hijerarhijsko procesiranje nastavilo na nivou primarnih, sekundarnih i tercijarnih senzornih,
ili interpretacijskih područja u kori velikog mozga (od kojih
se svako sastoji od hijerarhije nekoliko neuronskih mreža).
Veze unutar kao i između susednih neuronskih mreža u
ovoj hijerarhiji ostvaruju se posredstvom sinapsi (jedan
neuron može da ostvari oko 40.000 sinaptičkih veza sa
susedima), koje mogu biti eksitatorne ili inhibitorne. Osim
toga, tokom procesa učenja značajnu ulogu u globalnoj
distribuciji (po celoj moždanoj kori) hijerarhijski obrađivanih informacija igraju i moždani talasi.
Sekvencijalni (fon Nojmanovi) računari danas imaju takt
~ 10-10 s, dok je prosečno vreme generisanja akcionog
potencijala neurona ~ 10-3 s. Iako je ovaj odnos brzine
aktivacije pojedinačnih procesirajućih elemenata ~ 107
puta veći kod sekvencijalnih poluprovodničkih računara,
ipak je mozak superioran nad njima kada se radi o nekim
komplikovanim zadacima, kao {������������������������
�������������������������
to su obrada i prepoznavanje slike, orijentacija i kretanje u prostoru promenljivih
karakteristika, razumevanje govora itd. Razlog velikih
mogućnosti mozga leži u paralelnoj obradi informacija.
Osim toga, po{to je broj neurona u mozgu, kao i broj
veza između njih konstantan, znanje je distribuirano po
vezama, a nove informacije se dodaju pode{avanjem jačine veza između neurona. Takođe, određeni delovi informacija se ne nalaze na tačno određenim pozicijama,
nego su distribuirani po regionima u mozgu. Time o{tećenje
neurona, pa čak i grupa neurona, ne utiče u većoj meri
na pogor���������������������������������������������
{��������������������������������������������
anje performansi sistema, dok kod većine sekvencijalnih računara uni{tenje dela procesorske jedinice,
ili dela memorije vodi, ili ka prestanku rada celog sistema,
ili do nepovratnog gubitka informacija.
Za razliku od sekvencijalnih računara, kod kojih centralna
procesorska jedinica kontroli{e sve interne aktivnosti i ima
pristup memoriji, kod mozga je upravljanje lokalno. Pona{anje
svakog neurona u mozgu zavisi samo od njegovog prethodnog znanja i od ulaznog okruženja, pa se može reći da je
izlaz svakog neurona funkcija lokalno dostupne informacije.
This principle is also justified from the viewpoint of energy
efficiency, which in the case of prolonged and exhausting
matches represents a crucial factor. Namely, in this way
numerous overreactions of external parts of the extremities
are avoided. For example, the goalkeeper in football moves
his shoulders first when he’s preparing himself for a potential reaction. In the next moment, his sensory system perceives that further movement is unnecessary, because the ball
misses the goal. Activating other arm muscles is no longer
necessary. If the whole arm had been innervated, it would
mean energy wasted.
What is the essence of hierarchical neural networks in the
brain as well as of cognitive implication? Nowadays, the predominant scientific paradigm assumes that information processing at the level of CNS takes place among hierarchically
organized and interconnected neural networks. For instance,
a piece of visual information is firstly processed at the level of
retina (beginning with a network of photoreceptive plugs and
sticks, and ending with a network of ganglion cells), and the
hierarchical processing continues at the level of primary, secondary and tertiary sensory or interpretative regions in the
cerebral cortex (each of which consists of a hierarchy of several neural networks). Connections inside one as well as between neighbouring neural networks in this hierarchy are realized by means of synapses (one neuron can realize about
40,000 synaptic connections with its neighbours), which can
be excitatory or inhibitory. Apart from that, in the learning
process a significant role in the global distribution (all over the
cortex) of hierarchically processed information is played by
brain waves.
Sequential (von Neuman’s) computers now have a tact ~
10-10 s, whereas an average time of generating action potential in a neuron is ~ 10-3 s. Although this proportion
between the speed of activation of particular processing
elements is ~ 107 times greater in sequential semi-conducting computers, the brain is still superior to them when it
comes to some complex tasks, such as image processing and
recognition, orientation and movement in space having
changeable features, speech recognition, etc. The reason for
such great capabilities of the brain lies in parallel information
processing.
Besides, as the number of neurons in the brain, as well as the
number of connections between them is constant, knowledge
is distributed along the connections, and new information is
added by adjusting the strength of connections between neurons.
Furthermore, certain parts of information are not placed in
specific positions, but are distributed in different regions of the
brain. Thus, possible damage in neurons, and even in a cluster
of neurons, does not greatly affect the system’s performance,
while in the majority of sequential computers the destruction
of a part of a processor unit or a part of its memory leads either
to the whole system shutting down or to an irretrievable loss
of information.
Unlike sequential computers, in which a central processor unit
controls all internal activities and is able to access the memory,
the brain is locally controlled. The behaviour of each neuron
in the brain depends solely on its prior knowledge and of the
input environment, so we can conclude that the output of each
neuron is a function of locally available information.
NN as an attempt at modelling the workings of the human
brain have the following positive traits: parallel processing,
30
NM kao poku{aj modeliranja rada ljudskog mozga imaju
sledeća dobra svojstva: paralelan rad, izvr{enje komplikovanih zadataka u relativno kratkom vremenu, distribuiranu
raspodelu informacija, slabu osetljivost na o{tećenja, kao i
Zbornik radova 2010, 22-33
execution of complex tasks in a relatively short space of time,
distribution of information, low sensitivity to damage, as well
as the learning ability, that is, the ability to adapt itself to
changes in the environment and improve its functioning
based on experience. The advantage of the architecture of
hierarchical NN is that functionally specialized neurons in
each layer process only a limited amount of information.
mogućnost učenja, odnosno adaptacije na promene u okruženju i pobolj���������������������������������������������
{��������������������������������������������
anje rada na osnovu iskustva. Prednost arhitekture hijerarhijskih NM je da funkcionalno specijalizovani
neuroni svakog sloja procesiraju samo ograničenu količinu
informacija.
Software agents
Kompjuterska nauka defini{e softverske agente kao delove
softvera koji rade za račun korisnika ili druge softverske
aplikacije u saradnji sa agencijom. Agencija u ovom slučaju
predstavlja, najče�����������������������������������������
{����������������������������������������
će Es koji odlučuje o akcijama i adekvatnosti akcija agenata. Agenti kao softverske komponente
imaju visok nivo samostalnosti, aktiviraju se i deluju samostalno, a ne pod uticajem drugih aplikacija. Izvedeni koncept
softverskog agenta predstavljaju inteligentni agenti koji poseduju karakteristike i sposobnosti Vi, odnosno učenja i
zaključivanja. Smatra se da danas postoje samo četiri tipa
inteligentnih softverskih agenata:
Computer science defines software agents as parts of software that work for the sake of the user or another software
application in cooperation with an agency. The agency in
this case is mostly an ES which decides on actions and the
adequacy of the agent’s actions. The agents as software
components have a high degree of autonomy, they are activated and operated independently, and are not controlled
by other applications. Such concept of a software agent uses
intelligent agents that possess the characteristics and functions of AI, or learning and inferring. It is thought that today
there are as many as four types of intelligent software agents:
•Agents for purchase on electronic stalls travel across the
network accumulating information about goods and services
on offer. These agents work very effectively at the sales of
electronic goods and services, and Amazon.com is the best
example of a successfully implemented technology of this
type of agent. This web site can offer you a list of books or
music based on your current or prior purchases.
•Personal agents are intelligent agents that work on your
behalf.
•Monitoring-proactive agents are used in complex computer
systems for monitoring and reporting on the state of equipment.
These agents also control stocks of spare parts for the
equipment, prices of new parts or new pieces of equipment,
and then they forward the information to its users.
•Data-mining agents use information technologies in order
to discover the rules and patterns in a wealth of information coming from a great variety of sources, and thereby
narrow the set of sources important for the users themselves.
A case from practice...
There are many ways to perform an analysis of sportspeople’s
activities concerning their competition. In order for these
matters to be understood, we are going to give an account of
one case from volley-ball practice.
„Data Volley 2“ (DV-2), is the only scout programme for volley-ball used by the best teams in Italian Championship, as well
as all the best national teams in the world. Like other „company’s
diamonds“, this programme is available in several versions
adaptable to every level.
Why DV-2?
DV-2 is the most popular and the most widely used statistical
scout software in the world. All the bigger clubs and national
teams in the world chose DV-2 as their only instrument for
scouting and analysis of team statistical data because it was
assessed to be extremely professional and comprehensive.
DV-2 is Data project’s response to the coaches’ request for
complete and detailed results in a very short time interval.
Features of DV-2
Extreme comprehensiveness: DV-2, a programme adopted
by the best national and club teams all over the world is the
Softverski agenti
•Agenti za kupovinu na elektronskim tezgama putuju mrežom dopremajući informacije o ponuđenoj robi i servisima. Ovi agenti rade veoma efikasno na prodaji elektronske robe i servisa, a Amazon.com je najbolji primer
uspe������������������������������������������������
{�����������������������������������������������
no implementirane tehnologije ovakvog tipa agenata. Ovaj veb sajt će vam ponuditi listu knjiga ili muzike
na osnovu va{e tekuće i va{ih predhodnih kupovina.
•Personalni agenti su inteligentni agenti koji deluju u va{e
ime.
•Monitoring-proaktivni agenti se koriste u kompleksnim
računarskim sitemima za nadgledanje i izve{tavanje o
stanju opreme. Ovi agenti prate i stok rezervnih delova
za opremu, cene nabavke novih delova ili opreme i ove
informacije prosleđuju korisnicima.
•Dejta majning agenti koriste informacione tehnologije
kako bi na{li pravila i obrasce u obilju informacija koje
potiču od velikog broja raznorodnih izvora i na taj način
suzili skup značajnih izvora za samog korisnika.
Slučaj iz prakse...
Postoji ne mali broj načina za analizu takmičarske aktivnosti sportiste. Da bi se razumela navedena problematika
objasniće se slučaj iz odbojka{ke prakse.
„Data volej 2“ (DV-2), (engl. „Data Volley 2“), je jedini
skautski program za odbojku koji koriste najbolji timovi u
italijanskom �����������������������������������������������
{����������������������������������������������
ampionatu, kao i svi nacionalni najbolji timovi na svetu. Kao svaki „dijamant kompanije“, program je
dostupan u vi{e verzija koje su prilagodljive svim nivoima.
Za{to DV-2?
DV-2 je najpoznatiji i najkori{ćeniji statistički skautski softver
na svetu. Svi veći klubovi i nacionalni timovi na svetu izabrali su DV-2 kao svoj jedini instrument za skautizam i
analizu timskih statističkih podataka zbog njegovog procenjenog nivoa profesionalizma i sveobuhvatnosti. DV-2 je
odgovor Data projekta na zahtev trenera o kompletnim i
detaljnim rezultatima u kratkom vremenskom intervalu.
Mogućnosti DV-2
Ekstremna sveobuhvatnost: DV-2, program usvojen od strane najboljih nacionalnih i klupskih timova sveta jedini je
statistički skautski softver koji omogućava korisniku da odmah
31
INTRODUCTORY LECTURE
UVODNO PREDAVANJE
Stefanović, \.: SPORT I VE[TAČKA INTELIGENCIJA
INTRODUCTORY LECTURE
UVODNO PREDAVANJE
Stefanović, \.: SPORTS AND ARTIFICIAL INTELLIGENCE Proceedings 2010, 22-33
only statistical scout software which enables the user to
immediately get extremely reliable data easily presentable
on the screen, which is greatly helpful for the coach in his
decision-making during the match.
Comprehensive version: DV-2 is available in two versions to
better meet different needs. Basic version: it gives a wide
range of options for scouting and analysis at a very favourable
price. With this version of the programme you can get a visualized statistics for each player, team, rotation, skill, as well as
a graphic display of the action zone as the direction of an attack.
Professional version: The software is amenable to an infinite
number of possibilities for personalization of required information during defining the parameters in phases, ranging
from the choice of discipline to be monitored and at which
level to the visualization of the direction of an attack.
Speaker possibilities for development: It is possible to connect
another computer on the bench to the scout’s one, allowing
those on the side to see the data directly during the match.
Also, it is possible to connect DV-2 score-board to an interactive screen. It is placed in front of the audience as an
entertaining and commercial promotion using visualization
of statistical data going live during matches.
Programme architecture: The philosophy/architecture that
characterizes DV-2, making it the most widely used statistical scout programme in the world, is based on its enormous
conformity to all user capabilities and needs: from national
trainers to beginners, everybody finds that with DV-2 they
are able to scout and analyze whatever they need during a
volley-ball match.
Flexibility: Programme functions can be personalized, so that
they are adjusted to personal capacities of the user: from
keyboard responsiveness to evaluation table of skills, from
compiling the criteria to printing out statistical reports.
Controlled scouting: The analysis of field zone in which the
skills are played out – for each analysis it is possible to visualize four different values to help the coach and different
skills so he could quickly identify parts of the field where the
best and the worst moves are performed.
Analysis of directions: It shows the direction of the serve and
the attack. It is possible to require an overall or detailed analysis
either of the team or of the individual players under certain
circumstances. DV-2 uses different colours to mark the best,
the worst and the mediocre moves.
Work base: It replaces real volley-ball floor surface and enables
the user to study all the statistical data of scout matches. To
this end, there is a possibility of personalizing prospects of
analysis based on personal needs by inputing very complex
formula.
Direction of the attack: It detects the efficacy and the direction of attacks depending on the region of the attack, the
quality of the impact, etc.
Details of the last score: The video screen shows the last score
for each player, and each of their skills following the „trend“
of checking up on every player in every technical aspect.
HTLM match report: A statistical overview of the match is
created in HTLM format for the purpose of publishing it on the
club’s web site or of forwarding it to other users via the network.
usvoji ekstremno pouzdane podatke lake za prikazivanje
koji mogu pomoći treneru u dono{enju odluke tokom meča.
Sveobuhvatna verzija: DV-2 je dostupan u dve verzije radi
boljeg prilagođavanja različitim potrebama. Osnovna verzija: Daje {irok spektar mogućnosti skautinga i analiza po
pristupačnoj ceni. Sa ovom verzijom programa moguća je
vizuelizacija statistika za svakog igrača, tim, rotaciju, ve{tinu,
i grafički prikaz zone akcije kao pravca napada.
Profesionalna verzija: Softver pristaje na beskonačno mnogo
mogućnosti za personalizaciju željenih informacija tokom
definicija u fazama parametara, od izbora koja ve{tina se
skautuje (bira) i na kom nivou, do vizuelizacije pravca napada.
Spiker mogućnosti za razvoj: Moguće je povezati drugi
kompujer na klupi za skautov, dozvoljavajući onima sa
strane da konsultuju podatke direktno tokom meča. Moguće
je povezati DV-2 bodovnu tabelu na interaktivni ekran. On
se nalazi ispred publike kao zabavna i komercijalna promocija sa kori{ćenjem vizualizacije statističkih podataka uživo
tokom mečeva.
Programska arhitektura: Filosofija/arhitektura koja karakteri{e
DV-2, čineći ga najra����������������������������������������
{���������������������������������������
irenijim i kori������������������������
{�����������������������
ćenim statističkim skauting programom na svetu, je njegova kompletna dostupnost
svakim korisničkim sposobnostima i potrebama: od nacionalnog trenera do početnika, svi nalaze da sa DV-2 imaju
mogućnosti skautinga i analize svega {to je potrebno tokom
odbojka{kog meča.
Prilagodljivost: Mogu se personalizovati programske funkcije prilagoćavajući ih sopstvenim kapacitetima: od odgovaranja tastature do procenjujuće tabele ve��������������������
{�������������������
tina, od prikupljanja kriterijuma do {tampanja statističkih izve{taja.
Kontolisani skauting: Analiza zone terena u kojima se ve{tine
iznose – za svaku analizu moguće je vizualizovati četiri
različite vrednosti koje pomažu treneru i za različite ve{tine
da bi brzo identifikovao delove terena gde se najbolji i
najgori potezi odigravaju.
Analiza pravaca: Pokazuje pravac servisa i napada. Moguće
je zahtevati potpunu, ili detaljnu, analizu bilo tima, igrača,
pod određenim uslovima. DV-2 različitim bojama ističe
savr{ene poteze, najlo{ije, kao i one između.
Radna podloga: Zamenjuje pravu odbojka{ku povr{inu sa
koje je moguće proučiti sve statističke podatke skautovanih
mečeva. Zbog toga postoji mogućnost peronalizovanja
prospekata analize zasnovanih na ličnim potrebama ubacujući veoma kompleksne formule.
Pravac napada: Beleži efikasnost i pravac različitih napada
zasnovanih na području napada, kvalitetu prijema itd.
Detalji poslednjeg pogotka: Pokazuje na videu učinke poslednjeg pogotka svakog igrača, ve{tinu po ve{tinu u cilju
„trenda“ da se svaki igrač proveri u svim tehničkim aspektima.
HTLM izve{taj meča: Stvara statistički pregled meča u HTLM
formatu radi objavljivanja na sopstvenom sajtu ili radi prosleđivanja istog drugim korisnicima preko mreže.
Personalizacija ekrana analize: Pamti i {tampa kompleksne
rasporede sastavljene od različitih prozora, a različitim tipovima analiza. Radi beleženja različitih kompleksnih faza
analiza dopu{ta korisniku/treneru izbor personalizovanja
programa za sopstvene potrebe.
Data volej na klupi: Postoji mogućnost konekcije sa drugim
kompjuterom na klupi sa koga se može konsultovati i proučavati podaci tokom meča – najče{će koriste mnogi mu{ki
i ženski nacionalni timovi.
Novi način praćenja bodova: DV-2 semafor program prikazuje zvanične poentere koristeći standardni video projektor,
Personalization of the screen for analysis: It memorizes and prints
out complex schedules comprised of different windows and
containing different types of analysis. With the aim of identifying
different complex phases, the analysis allows the user/coach to
personalize the programme to suit their own needs.
Data Volley on the bench: There is a possibility of connecting
with another computer on the bench, from which one can
consult and study the data during the match – this is mostly
used by male and female national teams.
32
Zbornik radova 2010, 22-33
A new way of tracking scores: DV-2 score-board programme
shows official pointers using a standard video projector,
immediately transferring data from DV-2, as well as the logo,
commercials, replay videos, etc.
odmah prikazujući podatke od DV-2, kao i logo, reklame i
ponovna prikazivanja, itd...
Conclusion
Cilj ovoga rada je bio da se objasni (prikažu mogućnosti i
olak{ice) kako primena ve{tačke inteligencije (računarska
disciplina čiji je zadatak da stvori računare koji mogu da
rezonuju na način sličan ljudskom rezonovanju) pomaže u
sportu. Implikacije neuronskih mreža na tehnologiju sportskog
treninga postoje. Zahvaljujući prihvaćenim principima funcionisanja, neuronske mreže mogu značajno da unaprede
efikasnost izvođenja pokreta sa jedne strane, a sa druge,
efikasnije procesiranje brojnih inputa putem čulnih receptora u organizam. Prilikom izvođenja treninga, može da se
na ovim principima razvije ekspertski sistem koji će se prilagoditi sportisti. To će sa jedne strane da minimizuje broj
živih kontrolnih operatera, a sa druge da optimizuje trening
sportiste, sprečavajući nastanak povreda uz maksimalni
mogući parametarski progres. Elektronske igre i softverski
agenti su sve vi{e prisutni i neophodni u sportu. Praktični
aspekt ovog teorijskog istraživanja se ogleda u sistematizovanju i pro{irivanju znanja iz oblasti nauke/teorije sportskog
treninga koja ima fundamentalni značaj za transer u praksu.
Na taj način se podstiču stvaralačke ideje koje predstavljaju
osnovu razvoja naučne misli iz oblasti sportskog treninga.
The aim of this paper was to explain how an application of
artificial intelligence (computer science discipline whose mission is to create computers which can imitate human reasoning
processes) can be of assistance in sports. The implications of
neural networks on technology of sports training is evident.
Thanks to accepted principles of functioning, neural networks
can on the one hand greatly improve the efficiency of body
movements, and help a more effective processing of numerous input by means of sense receptors in the organism. During
training, one can build an expert system following these
principles, which is to be adjusted to the sportspeople. That
would minimize the number of human control operators, and
optimize the sports training itself, preventing getting injured
with an optimal parameter progress. Electronic games and
software agents are getting more and more present and necessary in sports. A practical aspect of this theoretical investigation is reflected in systematization and expansion of knowledge concerning science/theory of sports training, which
has fundamental significance for transfer into practice. Thus,
creative ideas are fostered, and they represent the basis for
the development of scientific thought in the field of sports
training.
Reference
Amit, D. (1989). Modeling Brain Functions: The World of
Attractor Neural Nets. Cambridge: Cambridge University Press.
Andrew, R., & Adams, E. (2006). Fundamentals of Game
Design. London: Prentice Hall.
Cvetković, D., Ostojić, M., & Ćosić, I. (2008). Sleep Onset
Estimator: Evaluation of Parameters. Vankuver: EMBS.
Hyacinth, S. N. (1996). Software Agents: An Overview.
Knowledge Engineering Review, 11(3), 1–40.
Received: October, 15th 2010
Correspodence to:
\orđe Stefanović, PhD
Faculty of Sports and Pnysical Education
Blagoja Parovića 156
11000 Belgrade
Serbia
Phone: +381 11 35 31 000
E-mail: djordje.stefanovicªdif.bg.ac.rs
Zaključak
Kohonen, T. (1984). Self-Organization and Associative
Memory. Berlin: Springer.
Penrose, R. (1994). Shadows of the Mind. A Search for the
Missing Science of Consciousness. Oxford: Oxford
Univer. Press.
Raković, D. (2008). Osnovi biofizike. Beograd: IASC &
IEFPG.
Stefanović, \. (2006). Teorija i praksa sportskog treninga.
Beograd: Fakultet sporta i fizičkog vaspitanja.
Stefanović, \., Jakovljević, S., & Janković, N. (2010). Tehnologija pripreme sportista. Beograd: Fakultet sporta i
fizičkog vaspitanja.
Primljeno: 15. oktobra 2010. godine
Korespodencija:
dr \orđe Stefanović
Fakultet sporta i fizičkog vaspitanja
Blagoja Parovića 156
11000 Beograd
Srbija
Telefon: +381 11 35 31 000
E-mail: djordje.stefanovicªdif.bg.ac.rs
33
INTRODUCTORY LECTURE
UVODNO PREDAVANJE
Stefanović, \.: SPORT I VE[TAČKA INTELIGENCIJA
Download

sport i veštačka inteligencija sports and artificial intelligence