Amamodeli amakhulu olimi (LLMS) asekela amacala ahlukahlukene okusetshenziswa, ukusuka ekufingwini kokuqukethwe kuya emandleni okubonisana ngemisebenzi eyinkimbinkimbi. Isihloko esisodwa esijabulisayo sithatha i-ai ekhiqizayo emhlabeni wenyama ngokuyisebenzisa kumarobhothi kanye ne-hardware yomzimba.
Ugqugquzelwe yilokhu, sakha umdlalo we-AWS Re: Qamba abakhi abangama-2024 Fair besebenzisa i-Amazon Bedrock, ama-Strands Agents, AWS IOTNOD, ne-Amazon DynamoDB. Umgomo wethu bekuwukukhombisa ukuthi ama-LLM angabonisa kanjani ngecebo lomdlalo, imisebenzi eyinkimbinkimbi, futhi alawule amarobhothi omzimba ngesikhathi sangempela.
I-Robotic-Tac-Toe ngumdlalo osebenzayo lapho amarobhothi amabili omzimba ahamba ebhodini le-tic-tac-toe, ngokunyakaza kwe-gameplay kanye namarobhothi 'archestrated yi-LLMS. Abadlali bangakwazi ukulawula amarobhothi besebenzisa imiyalo yolimi lwemvelo, ibaqondise ukubeka izimpawu zabo ebhodini lomdlalo. Kulokhu okuthunyelwe, sihlola amasu wobunjiniyela kanye nobunjiniyela obusetshenziselwa ukubonisana ngomdlalo we-tic-tac-toe futhi unqume isu elingcono kakhulu lomdlalo kanye nohlelo lokuhamba lwabadlali abalandelayo.
Isipiliyoni esisebenzayo
I-Robotic-Tac-uzwane ikhombisa ukusebenzisana okunembile phakathi kwabantu, amarobhothi kanye ne-AI. Ababambiqhaza bangafinyelela kwi-portal yomdlalo ngokuskena ikhodi ye-QR, bese ukhetha kusuka kuzindlela eziningi:
- Umdlali we-player vs – Inselelo umphikisi womuntu
- Umdlali vs llm – Hlola amakhono akho ngokumelene ne-AI-Powered LLM
- Llm vs. llm – Bukela amamodeli amabili e-AI ahlelekile futhi ancintisane ngokuzimela
Lapho umdlali ekhetha iseli eliqondiwe, amarobhothi amabili, abekwe eceleni kwebhodi le-tic-tac-toe, aphendule imiyalo ngokwenza ukunyakaza okunembile ukubeka ama-X noma o. Ividiyo elandelayo ikhombisa lokhu kusebenza.
Ukubuka konke
I-Robotic-Tac-Toe ifaka ukuhlanganiswa okungenamthungo kwezinsizakalo ze-AWS, kunciphisa isidingo sokulandelanisa kwangaphambili. Esikhundleni salokho, i-AI yakha imiyalo echazayo ngesikhathi sangempela. Umdwebo olandelayo uchaza ukwakhiwa okwakhiwe kwi-AWS IOT Core, enika amandla ukuxhumana phakathi kwamarobhothi alawulwa yi-raspberry PI nefu.
Isixazululo sisebenzisa izinsizakalo ezilandelayo ezilandelayo:
I-Hardware ne-Software
- Ukusethwa ngokomzimba kwephrojekthi kufaka iBhodi ye-Tic-Tac-Toe eshumeke ngezinkomba ze-LED ukugqamisa ukubekwa kwe-X no-O.
- Amarobhothi amabili (amamodeli wamathoyizi aguquliwe) asebenza ngokusebenzisa izilawuli ze-raspberry PI ezifakwe amamojula we-infrared ne-RF.
- Ikhamera ye-raspberry ye-raspberry ye-raspberry inika amandla ukuhlaziywa okususelwa kombono, ithumba isimo seBhodi kanye nokudlulisa idatha ukuze kusetshenzwe kabusha ukucubungula ikhompyutha. Ngokwengeziwe, isilawuli sehabhulethi esizinikezele sisebenza njengedivaysi ye-IOT exhuma ku-AWS IOT Core, ethuthukisa ukusebenzisana okuhle kwe-gameplay.
- Ohlangothini lwesoftware, ama-AWS Lambda aphatha ngokunxenxa umphathi I-STRANDS Agent, ye-Core Game Logic no-OrcheStation.
- Amandla okubuka amakhompyutha, anikwe amandla yi-OpenCV, ahlaziye ukwakheka kweBhodi kanye nokuhamba okuqondile kwamarobhothi. Ama-ejenti we-Amazon Bedrock agents Orchestrate imisebenzi ukukhiqiza amasu wokunyakaza namasu wegeyimu.
Ama-Strends Agents asebenza
Ama-Strys Agents ashintshanisa imisebenzi yabasebenzisi bakho besicelo ngokuxhumana kwama-orchestrating phakathi kwemodeli yesisekelo (FM), imithombo yedatha, izinhlelo zokusebenza zesoftware, nezingxoxo zabasebenzisi.
Umphathi we-Supervisor
I-Superpersor Agent isebenza njenge-Orchestrator ephatha yomibili umenzeli wokuhambisa kanye ne-ejenti yomdlalo, ukuxhumanisa kanye nezinqumo zokuqondisa izinqumo kulo lonke uhlelo. Le nqubo iqukethe lezi zinyathelo ezilandelayo:
- Umenzeli uthola imiyalo ephezulu kakhulu noma imicimbi ye-gameplay (ngokwesibonelo, “umdlali x uhanjiswe ku-2b, akhiqize impendulo ye-robot”) futhi enquma ukuthi yimuphi umenzeli we-ejenti-hambisa akhethekile noma umenzeli womdlalo – kufanele acelwe.
- Umsebenzi we-Supervisor Aw Lambda usebenza njengesilawuli esimaphakathi. Lapho kubangelwa, kuphamba isicelo esingenayo, kuqinisekisa umongo, bese kuthi kuhambisa isicelo kwi-ejenti yemicu efanelekile. Ukulandela umkhondo kunikwe amandla wonke umsebenzi ukuhamba komsebenzi ukuvumela ukuqapha kanye nokulungisa iphutha.
- Kuya ngohlobo lwesicelo:
- Uma kufaka phakathi ukubuyekeza noma ukuhlaziya isimo segeyimu, umphathi we-ejenti wegeyimu, ethola isimo sebhodi futhi akhiqize ukuhamba okulandelayo kwe-AI.
- Uma kubandakanya ukuzulazula kwamarobhothi ngokomzimba, umphathi ucela umenzeli wokuhambisa, okhiqiza imiyalo yokunyakaza kwikhodi yePython.
- I-ejenti ye-Superperisor ihlanganisa izimpendulo ezivela kuma-ejenti aphansi futhi ibalwa ngefomethi ehlukile yokuphuma. Lokhu kuvumela ukuguquguquka kokuthi umphumela umyalo werobhothi, ukunyakaza komdlalo, noma inhlanganisela yakho yomibili.
- Ukusebenzisana, kufaka phakathi izindlela zokuthatha izinqumo nokuphuma kokugcina, ungene ngemvume kubhakede le-S3. Le ndlela yokugawula ihlinzeka ngokulandela ama-ejenti amaningi futhi isekela ukuphatha iphutha ngokubuyisa imiyalezo yephutha ehlelekile lapho kuvela izingqinamba.
Le module ihlinzeka ngesendlalelo sokubusa ngaphezulu kwendawo enikwe amandla i-AI, inika amandla ama-orchestaration acalleble ama-overchest kuwo wonke ama-ejenti. Ngokuqondisa okuqondisayo kanye nezimpendulo ezihlanganayo, i-ejenti ye-Supervisor kusiza ukwenziwa okuthembekile, ukuqapha okwenziwe lula, kanye nesipiliyoni somsebenzisi esithuthukisiwe.
Hambisa umenzeli
I-ejenti yokuhambisa ikhiqiza ikhodi ye-Python ye-Step-by-step. Le nqubo iqukethe lezi zinyathelo ezilandelayo:
- Umenzeli uthola isikhundla sokuqala nesikhathi esiya kuyo kugridi (ngokwesibonelo, “3a kuya ku-4b North”), inquma ukunyakaza okudingekayo, futhi ithumela imiyalo ku-robot efanelekile.
- Umsebenzi we-LLM Navigator AWS Lambda wenza imiyalo yokunyakaza yamarobhothi esebenzisa ama-Strands Agents. Lapho kubangele, ithola isicelo esine-ID yeseshini kanye nombhalo wokufaka ochaza isikhundla sokuqala se-robot nendawo oya kuyo. Umsebenzi ube esekhipha umenzeli we-strands, ukuthumela isicelo kanye nokulandelela kunikwe amandla ukuvumela ukulungisa iphutha.
- Impendulo evela ku-ejenti iqukethe imiyalo yokunyakaza efana nokujika nokuqhubekela phambili ngamasentimitha.
- Le miyalo iyacutshungulwa futhi ungene ngemvume kubhakede le-S3 ngaphansi kwefayela le-CSV. Uma ifayela le-log likhona, okufakiwe okusha kufakiwe. Ngaphandle kwalokho, kudalwe ifayela elisha.
- Umsebenzi ubuyisela impendulo ye-json equkethe imiyalo ekhiqizwayo nesikhathi esithathiwe sokukhipha isicelo. Uma kwenzeka iphutha, umlayezo wephutha elihleliwe uyabuyiselwa.
Le module ihlinzeka ngokuzulazula okusebenzayo nokunakulandelwa kwamarobhothi ngokusebenzisa ukukhiqizwa kokufundiswa okunamandla kwe-AI ngenkathi kugcinwa indlela yokugawula okuphambayo yokuqapha kanye nokulungisa iphutha.
Umenzeli womdlalo
I-Agent yomdlalo isebenza njengomphikisi, okwazi ukudlala abasebenzisi babantu. Ukuthuthukisa ukufinyeleleka, abadlali basebenzisa ingosi yewebhu ephathekayo ephathekayo ukuxhumana nomdlalo, okubandakanya iphaneli yomlawuli wokuphatha imidlalo eqhutshwa yi-AI. I-LLM Player uhlelo lokusebenza olungenasisekelo oluhlanganisa ama-AWS Lambda, i-Amazon Dynanom, kanye ne-strands ejenti ukuphatha nokushintsha uhambo. Ithrelela inqubekela phambili yomdlalo ngokugcina umlando we-Mov etafuleni le-Amazon Dynanom, likuvumela ukuthi wakhe kabusha umbuso wamanje webhodi noma nini lapho eceliwe. Inqubo ye-gameplay iqukethe lezi zinyathelo ezilandelayo:
- Lapho umdlali enza ukunyakaza, umphathi we-Supervisor Strands abuyisa lo msebenzi wombuso bese ebiza umsebenzi we-ejenti yemicu ukukhiqiza ukunyakaza okulandelayo. Ukukhetha kwe-ejenti kuncike kumaka womdlali (
‘X’noma‘O’), Qinisekisa ukuthi imodeli efanele isetshenziselwa ukwenza izinqumo. - I-ejenti icubungula ibhodi lomdlalo lamanje njengokufaka futhi libuyisa ukuhambisa okulandelayo okulandelayo ngokusakazwa komcimbi.
- Ukuchitheka komsebenzi wonke kuhlotshiswe yi-Supervisor Strands Agent. Leli gama lithola izicelo ze-API, liqinisekisa okokufaka, libuyisa isimo sebhodi, inxusa imodeli ye-LLM, bese ibuyisa impendulo ehlelekile equkethe isimo somdlalo esibuyekeziwe.
Lolu hlelo luvumela i-gameplay yesikhathi sangempela, i-ai-eqhutshwa yi-gameplay, okwenza ukuthi abadlali bancintisane ngokumelene nomphikisi ohlakaniphile onikezwe amandla yi-LLMS.
Ukuzulazula kweRobot nge-Computer Vision
Kuphrojekthi yethu ye-robotic-tac-toe, umbono wekhompyutha udlala indima ebalulekile ekukhiqizeni ukunyakaza kwamarobhothi okuqondile nokunemba kwe-gameplay. Ake sihambe ngendlela esisebenzise ngayo isixazululo sisebenzisa izinsizakalo ze-AWS kanye namasu wokubona wekhompyutha athuthukile. Ukusetha kwethu kufaka phakathi ikhamera ye-raspberry PI ebekwe ngaphezulu kwebhodi lomdlalo, ngokuqhubeka nokuqapha izikhundla nokunyakaza kwamarobhothi. Ikhamera ithumba izithombe ezilayishwa ngokuzenzakalelayo ku-Amazon S3, zakha isisekelo sepayipi lethu lokucubungula umbono.
Sisebenzisa ukuhlaziywa okuyinhloko kwengxenye (i-PCA) ukuze sithole ngokunembe futhi silandelele ukuqondiswa kwamarobhothi kanye nesikhundla ebhodini lomdlalo. Le ndlela isiza ukunciphisa ubukhulu ngenkathi kugcinwa izici ezibalulekile zokulandela amarobhothi. I-angle ye-Orientation ibalwa ngokususelwa ezingxenyeni eziyinhloko zezici ezibonakalayo zerobhothi.
Imodyuli yethu ye-OpenCV itholakala futhi yathunyelwa njenge-Amazon SageMaker Endpoint. Icubungula izithombe ezigcinwe e-Amazon S3 ukunquma okulandelayo:
- Ukuma kwe-robot okuqondile ebhodini lomdlalo
- Ama-engeli akhona njengamanje
- Ukuqinisekiswa kokuhamba
Umsebenzi we-AWS Lambda onikezele othobela ukugeleza komsebenzi wokucubungula umbono. Iphatha okulandelayo:
- I-SageMaker Endpoint Ukuncenga
- Ukucutshungulwa kwemiphumela yokuhlaziywa kombono
- Isikhundla sangempela nesikhathi sokuvuselelwa
Lolu hlelo lwe-Computer Vision lusiza ukuzulazula okunembile kwamarobhothi kanye nokulandela ngomkhondo isimo somdlalo, okufaka isandla esimweni se-gameplay esingenamthungo ku-robotic-tac-toe. Inhlanganisela ye-PCA yokutholwa kokuqonda, i-OpenCV yokucutshungulwa kwezithombe, kanye nezinsizakalo ze-AWS zokuhanjiswa kusiza ukudala isisombululo se-computer esinamandla.
Ukugcina
I-Robotic-Tac-Toe ikhombisa ukuthi i-AI, amarobhothi kanye ne-Cloud Computing ingaguqula kanjani ukudala okuhlangenwe nakho okusebenzayo. Le phrojekthi iqokomisa amandla we-AWS IOT, Ukufunda Komshini (ML), kanye ne-AI ekhiqizayo e-Gaming, Ezemfundo, nangale kwalokho. Njengoba ama-robotic aqhutshwa yi-AI aqhubeka nokuvela, i-robotic-tac-toe isebenza njengezwisiseki ngokuzayo kokudlala okuqondayo, imisebenzi exhumanayo.
Hlala ubukele izithuthukisi zesikhathi esizayo, izindlela ezandisiwe ze-gameplay, futhi zisebenzisana kakhulu nokusebenzisana kwe-AI-Powered.
Mayelana nababhali
UGeorges Hamieh Ingabe Umphathi We-Akhawunti Yezobuchwepheshe Ephakeme e-Amazon Web Services, ekhethekile kwidatha ne-AI. Passionate mayelana ne-Innovation neTekhnoloji, ozakwethu namakhasimende ukusheshisa ukuguqulwa kwawo kwedijithali kanye nohambo lokutholwa kwamafu. Isikhulumi sasobala esinolwazi kanye nomeluleki, uGeorges uyakuthambisa ukuthwebula impilo ngokuthwebula izithombe nasekuhloliseni izindawo ezintsha ohambweni lomgwaqo nomndeni wakhe.
U-Mohamed Salah Ingabe ukwakhiwa kwezixazululo eziphezulu e-Amazon Web Services, okusekela amakhasimende e-Middle East naseNyakatho ne-Afrika ekwakheni izixazululo zamafu ezi-scaloble nezobuhlakani. Unothando mayelana ne-generative ai, amawele edijithali, nokusiza izinhlangano ziphenduka ezintsha zibe umthelela. Ngaphandle komsebenzi, u-Mohamed uyakujabulela ukudlala i-playstation, ukwakha amasethi we-LEGO, nokubuka ama-movie nomndeni wakhe.
Saddam hussain Ingabe ukwakhiwa kwezixazululo eziphezulu e-Amazon Web Services, enakekela e-Aerospace, i-AI ekhiqizayo, kanye nezindawo ezintsha zokuzijwayeza. Ukudweba kusuka kuhambo lwe-Amazon.com Pipharing Journel e-AI / ML kanye ne-Conforative AI, Usiza izinhlangano ziqonde izindlela ezifakazelwe kanye nemikhuba emihle kakhulu eye yakhula ezigidini zamakhasimende. Ukugxila kwakhe okuyinhloko kusiza amakhasimende omkhakha womphakathi e-UAE ukuthi asuse ama-AWS, awawaqondise ngokusebenzisa ubuchwepheshe bokutholwa kwamafu we-CustTunt ngenkathi benza amakhono aqhubekayo ngenkathi enza amakhono aqhubekayo.
UDkt Omer DawelEit Ingabe ukwakhiwa kwe-Solutions okuyinhloko ku-AWS. Unentshisekelo mayelana nokubhekana nezinselelo zezobuchwepheshe eziyinkimbinkimbi futhi asebenzisane kakhulu namakhasimende ukuklama nokusebenzisa izisombululo ezinamandla, ezinomthelela omkhulu. U-Omer unamashumi amabili emisebenzi yezezimali, izinhlaka zomphakathi kanye ne-telecoms isipiliyoni kuwo wonke ama-Startsups, amabhizinisi, kanye nokuguqulwa kobuchwepheshe obukhulu.
