Riba:
- Ƙarfin fahimtar bambance-bambance tsakanin PCG na algorithmic da PCG na wucin gadi na wucin gadi da kuma samar da abun ciki tare da tsarin matasan (ƙirar fasaha na fasaha na fasaha, ƙaddamar da algorithm).
- Ƙarfin don samar da ikon sarrafa wasa da bambance-bambance ta hanyar samar da jerin mahimmin ƙuntatawa ga samarwa
- Ikon sarrafa ingantattun kasada a cikin samar da tsari, kamar ma'anar maimaitawa, cika mara ma'ana da rikodi na maimaitawa.
A game's appetite for content is insatiable: the player plays a level once and finishes it, but you spend days producing it. Ƙirƙirar abun ciki na tsari (PCG) Ita ce samar da abun ciki ta atomatik ko ta atomatik (taswira, matakin, sanya abokan gaba, abu, manufa, har ma da sassan labarin) ba da hannu ɗaya bayan ɗaya ba, amma tare da dokoki da algorithms. PCG Classic algorithm ne (ayyukan surutu, rugujewar aikin igiyar ruwa, janareta-koridor). AI yana ƙara sabon Layer zuwa PCG: ƙarni na ma'ana, abun ciki na mahallin-tambayoyi, bayanin wuri, comps abokan gaba, da ƙirar matakan da suka dace da jigo da ƙira, ba kawai ba da gangan ba.
A cikin wannan rukunin, zaku koyi yadda ake amfani da AI a cikin PCG; Za ku koyi yadda ake haɗa PCG na algorithm tare da AI, yadda ake samar da abun ciki wanda za'a iya tantancewa da sake fitarwa (ƙaddarawa), da sarrafa inganci.
Duniya PCG guda biyu: algorithmic da generative
Mu fayyace bambancin. Algorithmic PCG shine canonical kuma ana iya sakewa dangane da iri (lambar farawa wanda ke ƙayyade samarwa); Iri ɗaya yana haifar da taswira iri ɗaya. Mai sauri, arha, mai iya faɗi; amma yana iya zama "marasa rai". Tare da Generative AI, PCG yana fahimtar mahallin da jigo; Yana biyan buƙatu masu ma'ana kamar "jigon kogon kankara, mai tsananin tarko, zanen gidan kurkuku mai ɗaki 3." Amma fitowar AI ba ta kai tsaye sake sakewa ba kuma baya bada garantin iya wasa. Hanyar da ta fi karfi ita ce matasan: AI yana haifar da ƙira mai girma da jigo (rubutun nema, manufar ɗaki, jigon maƙiyi), algorithm yana zubar da wannan ƙira a cikin lissafin lissafi da daidaituwa.
Matakan kwarara mataki-mataki:
- Ƙayyade ƙa'idodin samarwa da ƙuntatawa (jigo, wahala, girman, dole-has).
- Samo babban jigo daga AI (maƙasudin ɗaki, kwarara, jigo, dabaru na nema).
- Nan take tare da algorithm/injini (geometry, layout, balance).
- Verify playability (is there an exit, is it solvable, is there no blockage).
- Seed and record (wanda shigarwar da aka samar wanda fitarwa - reproducibility).
Tukwici: Koyaushe ba AI jerin "matsaloli": "kowane matakin dole ne ya sami mafita", "aƙalla ɗaki mai aminci", "babu abokan gaba da za su fara da". PCG mara ƙarfi yana haifar da matakan da ba za a iya warwarewa ko rashin adalci ba.
Aiki da samar da labari
Mafi kyawun yanki na AI na PCG shine nema da tsara labari. AI na iya yin suturar kwarangwal na manufa (tafi, nemo, debo, kare) tare da jigo da hali; zai iya haifar da bambancin don ayyukan gefe; Zai iya rubuta kwatancen wurare. Amma tarkon: AI yana kula da samar da "ayyukan tattarawa" guda ɗaya. Nemi ayyuka iri-iri da zaɓi mai ma'ana; bukatar kowace manufa don gabatar da shawara ga mai kunnawa.
Tsanaki: Babban haɗari a cikin ayyukan tsari shine jin maimaitawa da faɗuwar ma'ana. AI na iya samar da ayyuka 100, amma 100 kuma zai iya "kashe wolf 10." Quality ya dogara da iri-iri da mahallin, ba yawa ba; sanya kowane rukuni da aka samar ta hanyar sarrafa ɗan adam.
Sarrafa da sake haifuwa
A cikin ƙwararrun PCG, dubawa yana da mahimmanci kamar samarwa. Ƙirƙiri wani yanki na sarrafawa wanda ke tabbatar da cewa abubuwan da aka samar ana iya kunnawa, gaskiya kuma akan jigo. Har ila yau, sarrafa samar da ƙayyadaddun ƙayyadaddun bayanai: rikodin abin da shigarwa (iri, da sauri, siga) ya ba da abin da aka fitar don ku iya sake yin shi idan matsala ta taso. Maimakon kunna fitar da AI kai tsaye cikin wasan, sanya shi ta hanyar amincewa da lokacin gyarawa.
Hanya mai ƙarfi don sarrafa ma'auni ita ce saita rajistan kunnawa ta atomatik: rubutun mai inganci yana duba kowane matakin da aka samar kafin sanya shi cikin wasa - ko ana iya samun mafita, ko akwai maƙiyan da za su kashe a farkon, ko maɓallin da ake buƙata yana samun dama. Hakanan zaka iya tsara dabaru na wannan rubutun tare da AI; Amma kar a amince da rubutun ba tare da gwada shi tare da ainihin abubuwan samarwa a cikin injin ba. Ƙwararrun ɗan adam da sarrafawa ta atomatik suna haɗa juna: injin yana amsa tambayar "yana iya yin wasa" a sikelin, ɗan adam ya amsa tambayar "yana jin daɗi da ma'ana" a samfurin.
Ma'auni na PCG da ƙwarewar ɗan wasa
Haɗari ɗaya na PCG shine bambance-bambancen da yawa ya juya zuwa rashin ganewa. Idan komai ya faru ba da gangan ba, mai kunnawa ba zai iya riƙe wani abu ba; Hakanan ana buƙatar ƙirar da aka saba kuma ana iya ƙware. Kyakkyawan PCG yana daidaita ma'auni tsakanin iri-iri da ƙwarewar ilmantarwa: injiniyoyi ya kamata su saba, shimfidu ya zama sabo. Nemi wannan ma'auni a sarari yayin samun fasahar AI-kamar "ci gaba da ƙamus na injiniyoyi, amma bari kowane matakin ya haɗa waɗannan injiniyoyi a cikin wani yanayi na daban." Ta wannan hanyar, mai kunnawa yana ganin sabon abu kuma baya rasa iko.
uku mini lokuta
Case 1 - The matasan m nasara. A cikin wasan rarrafe gidan kurkuku, ƙungiyar ta ba da manufofin ɗakin da jigo ga AI da lissafi ga algorithm. An samar da bambance-bambancen gidan kurkuku 200; Kowannensu ya zama abin wasa, amma jigon ya bambanta. Idan aka kwatanta da tsantsar sigar algorithmic, ''kowane gidan kurkuku yana jin daban'' 'yan wasa ya karu sosai.
Case 2 - Kudin samarwa mara iyaka. Ɗaya daga cikin ɗakin studio yana da AI ya haifar da zane-zane na 50 ba tare da hani ba; Lokacin da aka gwada, 9 ba su da fita, 6 sun ƙunshi wuraren maƙiyi waɗanda suka kashe ɗan wasan da farko. Lokacin da aka ƙara jerin ƙuntatawa "dole ne", ƙimar matakin lahani ya ragu daga 30% zuwa 2%.
Case 3 - Curatorship ya sake karya jin. A cikin bude wasan duniya, AI ta haifar da tambayoyin gefe 300; A cikin danyen sigar sa, kashi 70% na shi “an tara a kawo”. Tare da buƙatar nau'i-nau'i, mai zanen ya sake sakewa kuma ya tsara kowane aiki don haɗawa da yanke shawara / matsala; A cikin martanin mai kunnawa, gunaguni game da "tambayoyin gefe suna da ban sha'awa" sun ragu zuwa kashi ɗaya bisa uku.
Samfura huɗu masu kwafi
1) Matsakaicin matakin iyaka:
Matsayinku: mai tsara matakin.Jigo: [misali. watsar da tashar sararin samaniya]. Wahala: matsakaici. Girman: dakuna 5. Dole ne a sami: fita ɗaya, aƙalla ɗaki mai aminci, babu abokan gaba a cikin ɗakin farawa, ya kamata ya zama mai warwarewa.Tambaya: ba da ɗaki ƙayyadaddun matakan daki; Manufar, jigo, barazana da shawarar da yake gabatarwa ga kowane ɗaki.
2) Ƙirƙirar aiki daban-daban:
Ƙirƙirar tambayoyin gefe guda 10 don wasan mai zuwa: [nau'i/jigo]. Ƙuntatawa: babu wani dole ne kawai ya zama "tattara-da-kawo"; Kowace manufa tana gabatar da mai kunnawa da zaɓi ko matsala. Ga kowane manufa: take, taƙaitaccen jumla ɗaya, shawarar ɗan wasa, lada. Rubuta kowane jimla mai maimaitawa.
3) Lissafin labarun kan layi:
Duba jigon matakin da aka samar a ƙasa: [ manna shaci]. Duba: (1) Shin akwai mafita mai yuwuwa, (2) shin ana iya warwarewa, (3) akwai rashin adalci/kisan farawa, (4) akwai karkacewa daga jigon. Jera matsalolin kuma bayar da shawarar gyara.
4) Rikodin sake haɓakawa:
Ƙirƙirar samfurin rikodi don samarwa mai zuwa: iri da aka yi amfani da su, faɗakarwa, sigogi, taƙaitaccen fitarwa da matsayin yarda. Manufa: don samun damar sake fitar da abun ciki iri ɗaya idan matsala ta faru.
Rauni mai ƙarfi / Ƙarfi mai ƙarfi
Rawanin faɗakarwa:
Samar da ni gidan kurkuku bazuwar.
Babu hani ko jigogi; fitarwar za ta kasance ba za a iya wasa ba ko kuma ta zama gama gari.
Ƙarfi mai ƙarfi:
Matsayinku: mai tsara gidan kurkuku. Jigo: ambaliya haikali.7 dakuna, ƙara wahala. DOLE SAMU: wani nau'in makaniki a kowane ɗaki (madaidaicin tarko, wuyar warwarewa, yaƙi, bincike), ɗakin sirri guda ɗaya, wurin hutawa a cikin ɗaki mai ƙarfi, mai warwarewa. Fitowa: tebur daki-daki (manufa, jigo, barazana, lada). A ƙarshe, jera hatsarori daban-daban.
Taken, takurawa, juzu'in inji da buƙatun haɗari sun sa fitarwar ta shirya don samarwa.
Algorithmic vs. AI PCG tebur
Girman
PCG
PCG tare da AI
matasan
sake haifuwa
Babban (iri)
ƙananan
matsakaici
Ma'ana/jigo
mai rauni
mai karfi
mai karfi
garantin yin wasa
Ana iya gyarawa
Babu
Ana iya gyarawa
gudun
da sauri sosai
sauri
matsakaici
mafi amfani
geometry, balance
Aiki, jigo, rubutu
layin samarwa
Kuskuren gama gari
- Don samarwa ba tare da ƙuntatawa ba. Ba a warware ba, rashin adalci abun ciki yana fitowa.
- Fitowar AI mai gudana ba tare da kulawa ba. Tabbatar da iya wasa dole ne.
- Ba son bambancin. AI yana ƙoƙarin samar da "ɗaukawa" mai ɗaci.
- Ba rikodin reproducibility. Idan matsala ta faru, ba za ku iya dawo da abun ciki ba.
- Yawan rikicewa tare da inganci. Ayyuka na yau da kullun 300 sun fi muni fiye da ayyuka 30 masu kyau.
A takaice
Abubuwan PCG suna gamsar da ci; AI yana ƙara ma'ana da jigo gare shi. Hanya mafi ƙarfi ita ce matasan: AI yana samar da ƙira mai girma, algorithm yana samar da saurin kunnawa. Ba tare da jerin ƙuntatawa ba, bincika iya wasa, buƙatun iri-iri, da rikodin sake haifuwa, PCG yana samar da datti cikin sauri. Bet akan nau'ikan sarrafawa, ba akan lambobi ba.
Aikin aikace-aikace
Zaɓi jigo (misali kogon kankara). A sami matakin ɗaki 5 da aka samar tare da samfurin "Ƙuntataccen matakin ƙayyadaddun ƙayyadaddun ƙayyadaddun ƙayyadaddun ƙayyadaddun ƙayyadaddun ƙayyadaddun ƙayyadaddun ƙayyadaddun matakan", sannan a sami fitarwa iri ɗaya tare da samfurin "jerin duban wasa". Gyara duk wata matsala da aka samu kuma samar da daftarin karshe.
jerin abubuwan dubawa
- [ ] Na ba samar da jerin abubuwan "dole ne" takura.
- [ ] Na bar babban matakin ƙira zuwa AI da ƙaddamarwa ga algorithm.
- [ ] Na duba kowane abun ciki da aka samar don iya wasa.
- [ ] Na bukaci bambance-bambance da zabi mai ma'ana a cikin manufa.
- [ ] Na ajiye iri/sauri/ siga don sake haifuwa.