Wannan takarda tana bincika hanyar aiki don haɓaka hulɗar mutum-robot (HRI) ta hanyar sadarwar hankali mara magana. Babban jigon shine ana iya ƙara karɓar fasaha ta hanyar sanya hulɗa ta zama mai sauƙin fahimta da kuma daidaitawa da hankali. Maimakon fuskokin android masu sarƙaƙiya da tsada, binciken ya binciki ingancin nunin RGB-LED mai ƙarancin ƙuduri don isar da hankali guda huɗu na asali: farin ciki, fushi, baƙin ciki, da tsoro. Binciken ya tabbatar ko tsarin launi da haske mai ƙarfi za a iya gane su da aminci ta hanyar masu lura da ɗan adam a matsayin takamaiman yanayin hankali, yana ba da madadin mai tsada don robot ɗin da ke da ƙayyadaddun bayyanar.
An tsara binciken don gwada alaƙar tsakanin tsarin haske da aka tsara da hankalin da aka fahimta.
Dangane da aikin tushe a cikin lissafin hankali da ilimin halayyar launi (misali, [11]), masu bincike sun tsara hankali guda huɗu na asali zuwa launuka na farko:
Bayan launi mai tsayayye, sigogi masu ƙarfi sun kasance masu mahimmanci. An ayyana tsarin ta hanyar:
An nuna wa mahalarta ɗan adam jerin tsarin haske da nunin LED ya samar. Ga kowane tsari, an nemi su gane hankalin da aka yi niyya daga zaɓuɓɓuka huɗu ko kuma su nuna "ba a sani ba." Binciken mai yiwuwa ya auna daidaito (ƙimar gane), lokacin amsawa, da kuma tattara ra'ayoyin mahalarta kan sauƙin fahimtar kowane tsari.
Nunin ya ƙunshi grid na LEDs na RGB, yana ba da cikakken sarrafa launi ga kowane pixel. Al'amarin "ƙarancin ƙuduri" yana nufin grid ɗin da ya isa (misali, 8x8 ko 16x16) don zama madaidaici amma yana iya nuna siffofi masu sauƙi, gradients, ko tsarin sharewa, daban da allo na fuska mai ma'ana.
An tsara microcontroller (kamar Arduino ko Raspberry Pi) don samar da tsarin hankali da aka riga aka ayyana. Sigogin sarrafawa da aka aika zuwa direban LED sun haɗa da ƙimar RGB ($R, G, B \in [0, 255]$) ga kowane LED da umarnin lokaci don ƙarfi.
Takardar ta ba da rahoton cewa wasu daga cikin hankali na asali da aka yi la'akari da su ana iya gane su ta hanyar masu lura da ɗan adam a ƙimar da ta fi girma sosai (25%). An nuna cewa hankali kamar fushi (Ja, saurin ƙyallen ido) da baƙin ciki (Shuɗi, jinkirin shuɗewa) mai yiwuwa sun sami ƙimar gane mafi girma saboda ƙaƙƙarfan alaƙar launi da tunani.
Za a yi amfani da nazarin ƙididdiga (misali, gwajin Chi-square) don tabbatar da cewa ƙimar gane ba bazuwa ba ne. Matrix rikicewa mai yiwuwa ya bayyana takamaiman kuskuren rarrabuwa, misali, "tsoro" yana rikicewa da "fushi" idan duka sun yi amfani da tsarin mita mai girma.
Sharhin mahalarta ya ba da mahallin bayan daidaiton danye, yana nuna waɗanne tsarin suka ji "na halitta" ko "mai tsauri," yana ba da labarin gyare-gyare ga taswirar hankali-zuwa-tsari.
Manyan fa'idodin tsarin sune ƙarancin farashi, ƙarancin amfani da wutar lantarki, babban ƙarfi, da sassauƙar ƙira. Ana iya haɗa shi cikin robot ɗin kowane nau'i, daga hannayen masana'antu zuwa robot ɗin zamantakewa mai sauƙi, ba tare da tasirin kwarin da wani lokaci ke da alaƙa da fuskoki na gaske ba.
Iyaka sun haɗa da ƙarancin ƙamus na hankali (hankali na asali kawai), yuwuwar bambancin al'adu a cikin fassarar launi, da yanayin m da ke buƙatar wasu koyan mai amfani idan aka kwatanta da gane fuska na asali.
Wannan aikin ya yi daidai da amma ya sauƙaƙa binciken da ya gabata kamar na Geminoid F [6] ko KOBIAN [10]. Yana musayar bayyanar fuska mai cikakken bayyanawa don duniya da aiki, kama da falsafar da ke bayan "ƙayyadaddun bayyanar" robot [4, 7, 8].
Fahimtar Cibiya: Wannan bincike ba game da ƙirƙirar robot masu hankali ba ne; yana game da injiniyan iyawar zamantakewa. Nunin LED wata wayo ce, "madaidaiciyar" "mahaɗa" wacce ke amfani da dabarun ɗan adam da suka riga sun wanzu (launi=hankali, saurin ƙyallen ido=ƙarfi) don sanya yanayin na'ura ya zama mai iya karantawa. Wani nau'i ne na ƙirar sadarwa tsakanin nau'ikan, inda "nau'in" su ne wakilai na wucin gadi. Ainihin gudummawar ita ce tabbatar da cewa ko da alamun gani marasa ƙarfi, lokacin da aka tsara su da kyau, za su iya haifar da abubuwan da suka dace da hankali—wani binciken da ke da babban tasiri ga HRI mai girma, mai ƙarancin farashi.
Tsarin Hankali: Hankalin takardar yana da inganci amma mai ra'ayin mazan jiya. Ya fara ne daga jigon da aka saba da shi cewa hankali yana taimakawa karɓar HRI [2,3], ya zaɓi pallet ɗin hankali mafi sauƙi, kuma ya yi amfani da taswirar mafi sauƙi (ilimin halayyar launi). Gwajin da gaske gwajin amfani ne don wannan taswirar. Kwararar ta rasa damar bincika ƙarin yanayi marasa tabbas ko masu rikitarwa, wanda irin wannan tsarin zai iya haskaka da gaske fiye da kwaikwayon fuskoki.
Ƙarfi & Kurakurai: Ƙarfinsa shine kyakkyawan aiki. Yana ba da mafita mai aiki tare da yuwuwar aikace-aikacen nan take. Laifin shine a cikin ƙarancin burin bincikensa. Ta hanyar mayar da hankali kawai akan daidaiton gane na jihohi huɗu na asali, yana ɗaukar hankali a matsayin sigina mai tsayayye da za a warware, ba wani ɓangare mai ƙarfi na hulɗa ba. Ba ya gwadawa, alal misali, yadda nunin ke shafar amincewar mai amfani, aikin aiki, ko dogon lokacin haɗin gwiwa—ma'auni masu mahimmanci don "karɓa". Idan aka kwatanta da ƙirar ƙira a cikin gine-ginen hankali na lissafi kamar EMA [9] ko sararin PAD, wannan aikin yana aiki a cikin sauƙi na fitarwa.
Fahimta Mai Aiki: Ga manajoji samfurin, wannan tsarin ƙira ne don bayyana hankali na MVP. Aiwatar da sauƙi, haske mai launi a kan na'urar ku ta gaba. Ga masu bincike, mataki na gaba shine matsawa daga gane zuwa tasiri. Kar ka tambayi kawai "wane hankali ne wannan?" amma "shin wannan hankalin yana sa ku yi haɗin gwiwa mafi kyau/da sauri/tare da ƙarin amincewa?" Haɗa wannan nunin tare da samfuran ɗabi'a, kamar waɗanda suka fito daga wakilan ƙarfafawa waɗanda suka dace da ra'ayin mai amfani. Bugu da ƙari, bincika madaukai na hankali biyu. Shin tsarin LED zai iya daidaitawa cikin ainihin lokaci zuwa ra'ayin mai amfani da aka gano ta hanyar kyamara ko murya? Wannan yana canza nunin zuwa tattaunawa.
Za a iya tsara tsarin hankali a matsayin aiki mai canzawa na lokaci ga kowane pixel na LED:
$\vec{C}_{i}(t) = (R_i(t), G_i(t), B_i(t)) = \vec{A}_i \cdot f(\omega_i t + \phi_i)$
inda:
Tsarin "fushi" na iya amfani da: $\vec{A} = (255, 0, 0)$ (ja), $f$ a matsayin babban mitar murabba'in igiyar ruwa, da daidaitaccen $\phi$ a cikin dukkan pixels don tasirin walƙiya ɗaya. Tsarin "baƙin ciki" na iya amfani da: $\vec{A} = (0, 0, 200)$ (shuɗi), $f$ a matsayin ƙananan mitar sine wave, da jinkirin, share canjin lokaci a cikin pixels don kwaikwayon igiyar ruwa mai laushi ko tasirin numfashi.
Bayanin Ginshiƙi (Hasashen bisa ga da'awar takarda): Taswirar ginshiƙi mai taken "Daidaiton Gane Hankali don Tsarin RGB-LED." X-axis yana lissafa hankali huɗu da aka yi niyya: Farin ciki, Fushi, Baƙin ciki, Tsoro. Ga kowane hankali, ginshiƙi biyu suna nuna kashi na gane daidai: ɗaya don nunin LED ɗaya kuma ɗaya don matakin dama (25%). Abubuwan lura masu mahimmanci:
Sandunan kuskure a kan kowane ginshiƙi mai yiwuwa yana nuna bambancin ƙididdiga tsakanin mahalarta. Taswirar layi na biyu zai iya nuna matsakaicin lokacin amsawa, yana nuna saurin gane ga hankali masu daidaito kamar fushi.
Yanayi: Robot na haɗin gwiwa (cobot) a cikin wurin aiki da aka raba yana buƙatar sadarwa da yanayin cikinsa ga abokin aikin ɗan adam don hana haɗari da daidaita haɗin gwiwa.
Aikace-aikacen Tsarin:
Wannan hali yana motsawa bayan gane mai sauƙi don auna tasirin aiki na nunin hankali akan aminci da ingancin haɗin gwiwa.