Ama-intersection asemadolobheni kudala ancike kumalambu okukhanya kwezimoto anesikhathi esinqunyiwe alandela izinhlelo eziqinile, kodwa ijubane elisheshayo lokukhula kwamadolobha liphoqa amadolobha ukuthi acindezele kabusha le ndlela. Izitayela zentuthuko yesikhathi esizayo zezimpawu zokuhamba zikhomba ngokuqinisekile ezinhlelweni eziguquguqukayo, eziqhutshwa yidatha ezingaphendula ezimweni zokuhamba ezishintshayo ngesikhathi sangempela. Njengoba izindawo zamadolobha amakhulu zikhula, ukuminyana, ukukhishwa kwezinto ezingcolisayo, kanye nokukhathazeka ngokuphepha kuba kubi kakhulu, futhi izindlela zendabuko zokulawula izimpawu ziyahluleka ukuhambisana nakho. Lokhu kudale isidingo esinamandla sezixazululo zohlelo lwezokuhamba oluhlakaniphile ezingathuthukisa ukuhamba, zinciphise izikhathi zokulinda, futhi zithuthukise ukuphepha emgwaqeni. Kulesi sihloko, sihlola ukuthi amalambu okuhamba ahlakaniphile ashintsha kanjani, ubuchwepheshe obubalulekile obusekela lolu shintsho, kanye neqhaza lezingxenyekazi zekhompyutha ezishumekiwe ezithuthukile — njengebhodi eliyinhloko le-FET3399-C lakwa-Shandong Penghui — ekwenzeni lezi zinto ezintsha zibe yiqiniso.

Izibani zomgwaqo ezihlakaniphile zimelela ushintsho lwe-paradigm olusuka kumadivayisi angawodwa, ahlelwe kusengaphambili kuya kuma-node anenethiwekhi, aqaphela umongo axhumana nezimoto, ingqalasizinda, nezinkundla zamafu. Esikhundleni sokumane zijikeleze kubomvu, kuphuzi, nokuluhlaza, lezi zimpawu zesizukulwane esilandelayo zihlaziya idatha yethrafikhi ebukhoma, zibikezele amaphethini okugcwala, futhi zilungise isikhathi sazo ngokuguquguqukayo. Ziyitshe legumbi lamasu esimanje okuhamba kwedolobha, zisiza amadolobha ukunciphisa izikhathi zokuhamba, ukwehlisa ukusetshenziswa kukaphethiloli, nokunciphisa ukukhishwa kwegesi. Njengoba sibheka indawo ebanzi, kuyacaca ukuthi ikusasa lokulawulwa kwezimpawu zethrafikhi alihlukaniseki ne-Internet of Things (IoT), ubuhlakani bokwenziwa (AI), nezinkundla ezinamandla zekhompyutha ezishumekwe zingakwazi ukuphatha ama-algorithms ayinkimbinkimbi emaphethelweni. Izigaba ezilandelayo zihlukanisa lokhu kuguquka, ubuchwepheshe obunika amandla, nezisombululo zehadiwe ezenza ukuhlangana kwezimpawu zomgwaqo ezihlakaniphile kube yiqiniso namuhla.

Ama-traffic light ajwayelekile asebenze ngendlela elula esekelwe kusikhathi iminyaka eminingi, anobude bomjikelezo obungaguquki obulungiswa ngesandla ngezikhathi zamahora aphezulu noma imicimbi ekhethekile. Nakuba izinhlelo ezinjalo zithembekile futhi zingabizi, ngokwemvelo azikwazi ukubona izimo zangempela zomgwaqo — umgwaqo ocishe ungenamuntu uthola isikhathi esilinganayo sokukhanya okuluhlaza njengomgwaqo ominyene ogcwele izimoto. Lokhu kungasebenzi kahle kubangela ukuma okungadingekile, ukumosha uphethiloli okwengeziwe, nokucasuka kwabashayeli. Isinyathelo sokuqala sokuthuthuka safika ngamasignali asebenzayo asebenzisa izinzwa ze-inductive loop ezigxunyekwe ngaphansi komgwaqo ukuze zizwe izimoto ezisondelayo. Lezi zinhlelo zingandisa izikhathi zokukhanya okuluhlaza uma izimoto zitholakala, kodwa zisashoda ekuhlakanipheni kokuxhumanisa phakathi kwezimpambano eziningi noma ukubona kusengaphambili ukwanda kwezimoto.

Ukushintshwa kwezibonisi zomgwaqo ezihlakaniphile ngempela kwaqala ngokuhlanganiswa kwamanethiwekhi okuxhumana nezikhungo zokuphatha umgwaqo ezimaphakathi. Ngokuxhumanisa izibonisi ngezintambo ze-fiber optic noma ngezixhumi zeselula, amadolobha akwazi ukusebenzisa imigwaqo ehlelekile eguqula izikhathi zokushintsha kwemibala ukuze kwakheke "amagagasi aluhlaza" ahamba kahle. Kodwa-ke, inguquko yangempela yeza ngokufaka amandla e-Edge computing ngqo kukhabhinethi yesilawuli sezibonisi. Izilawuli zezibonisi zomgwaqo zanamuhla manje sezikwazi ukusebenzisa amamodeli e-AI endaweni, zicubungule okokufaka kwamakhamera ne-radar, futhi zixhumane nezinye izingqalasizinda ngaphandle kokuncika ngokuphelele kuseva ekude. Lokhu kuguquka kusuka kumadivayisi angasebenzi, ahlukanisiwe kuya kuma-node asebenzayo, axhunyiwe, futhi ahlakaniphile yikho okuchaza inguquko yamanje ekuphathweni komgwaqo. I-Shandong Penghui izibeke phambili kulo shintsho ngokuhlinzeka ngamabhodi ayisisekelo afakwe ngaphakathi anikeza lawa madivayisi amandla okucubungula nokuxhumana.

Imindeni emithathu yobuchwepheshe exhumene iqhakambisa ukuguqulwa kwezimpawu zomgwaqo: ubuhlakani bokwenziwa (AI), i-Internet of Things (IoT), kanye nokuhlaziywa kwedatha enkulu. I-AI iletha amandla okubikezela nokuzivumelanisa ekulawuleni izimpawu zomgwaqo, ivumela izinhlelo ukuthi zifunde emaphethini omlando futhi zilungiselele izinto ezingavamile zesikhathi sangempela njengezingozi, imicimbi ekhethekile, noma ushintsho lwesimo sezulu olungazelelwe. Amamodeli okufunda ngomshini angathuthukisa ukuhlukaniswa komjikelezo, ukuhlehla, nokulandelana kwezigaba ngempumelelo enkulu kunanoma yimuphi umqondo ohlelwe ngaphambilini. Isibonelo, ama-algorithms okufunda ngokuqinisa angalingisa izimo zomgwaqo eziyizinkulungwane ngemizuzwana ukuthola uhlelo lokugcina isikhathi olusebenza kahle kakhulu empambanweni ethile, bese lusebenzisa ngokushesha.

I-Internet of Things inikeza umgogodla wezinzwa walezi zinhlelo ezihlakaniphile. Izinzwa ze-IoT — okuhlanganisa amakhamera, amayunithi e-radar, amalophu e-inductive, kanye namamojuli okuxhumana phakathi kwemoto nengqalasizinda (V2I) — zihlala zisakaza idatha mayelana nenani lezimoto, isivinini, ukugcwala, ngisho nokuba khona kwabahamba ngezinyawo. Ukuhlangana okukodwa okuhlakaniphile kungakhiqiza amathuba edatha angama-terabyte ngonyaka, futhi inselelo isekucubunguleni leyo datha emaphethelweni ngesikhathi esincane kakhulu. Yilapho izindlela zedatha enkulu zingena khona, zivumela onjiniyela bezokuhamba ukuthi bathole amaphethini esikhathi eside, baqinisekise ukusebenza kwemodeli, futhi balungise ngokuqhubekayo ama-algorithms. Ndawonye, i-AI, i-IoT, nedatha enkulu zakha umjikelezo omuhle: idatha eningi yenza amamodeli e-AI angcono, futhi amamodeli e-AI angcono asebenzisa ngempumelelo ingqalasizinda, ethuthukisa ukuhamba ezindaweni zasemadolobheni kwawo wonke umuntu.

Ukusebenzisa lezi zobuchwepheshe ngezinga elikhulu kudinga izinkundla zekhompyutha ezishumekile eziqinile ezikwazi ukuphatha ukucabanga kwamanethiwekhi ezinzwa (neural network inference), ukuhlanganiswa kwezinzwa kwesikhathi sangempela, nokuxhumana okuphephile kwefu ngesikhathi esisodwa. Amathuluzi okusebenzisa asekhaya avamile awafaneleki ngenxa yamazinga okushisa adlulele, ukudlidliza, kanye nesidingo sokuthembeka kwesikhathi eside. Izixazululo ezishumekile zezinga lezimboni, njengalezo ezinikezwa yi-Shandong Penghui, zenzelwe ngqo ukuhlangabezana nalezi zidingo ezinzima. Amabhodi azo ayisisekelo ahlanganisa ama-CPU asebenza kahle kakhulu, amayunithi okucubungula ezinzwa (ama-NPU), kanye nezinhlobonhlobo zezixhumanisi ze-I/O, okuzenza zifanelekele izilawuli zamasignali omgwaqo zesizukulwane esilandelayo okufanele zicubungule ukusakazwa kwamavidiyo futhi ziqalise amamodeli e-AI lapha empambanweni yomgwaqo uqobo.

IShantung Penghui Intelligent Technology Co., Ltd isibe yigama eliphawulekayo emkhakheni wezokuthutha okuhlakaniphile ngokuthuthukisa i-embedded hardware enamandla ezinsizakalweni zengqalasizinda yezokuthutha ehlakaniphile. Omunye wemikhiqizo yabo ehamba phambili i-FET3399-C core board, i-module yekhompyutha ehlangene futhi enomsebenzi ophezulu eyakhelwe ku-Rockchip RK3399 system-on-chip. Le processor ene-six-core (i-dual Cortex-A72 kanye ne-quad Cortex-A53) inikeza ukusebenza okulinganiselayo okudingekayo kukho kokubili i-control logic kanye nemisebenzi ye-AI. I-FET3399-C ayiyona nje i-processor; iyipulatifomu ephelele ye-embedded ehlanganisa kufika ku-4 GB ye-LPDDR4 RAM, isitoreji se-eMMC esisemkhunjini, kanye nezinketho eziningi zokuxhumana ezihlanganisa i-Gigabit Ethernet, i-USB 3.0, i-PCIe, kanye nezixhumanisi eziningi zokubonisa. Ukuze uthole imininingwane eyengeziwe ngalo mkhiqizo, ungavakashela ku-Imikhiqizo ikhasi.

Okwehlukanisa i-FET3399-C ukwakheka kwayo kwezinga lezimboni, okuqinisekisa ukusebenza okuzinzile ezimweni ezinzima ezijwayelekile zamakhabethe asemgwaqeni. Isekela uhla lokushisa olubanzi, ukuhambisana kwe-electromagnetic okuthuthukisiwe, kanye nokusekelwa kwesofthiwe yesikhathi eside kwe-Linux ne-Android. Lokhu kuthembeka kubalulekile kuzilawuli zamasignali omgwaqo okufanele zisebenze ngaphandle kokuphazamiseka iminyaka. Ibhodi futhi ihlanganisa ukusheshisa kwehadiwe okuzinikezelwe kokucabanga kwe-AI, ngenxa ye-Rockchip NPU, okuyivumela ukuthi isebenzise izinhlaka zokuxhumana zokwenziwa kwengqondo ezidumile njenge-TensorFlow ne-Caffe ngempumelelo. Lokhu kusho ukuthi i-FET3399-C eyodwa ingakwazi ukuphatha ngesikhathi esisodwa umqondo wokuhlela isikhathi sesignali, ukuhlaziywa kwevidiyo okuvela kumakhamera amaningi, kanye nokuxhumana neplatifomu yokuphatha emaphakathi — izinga lokuhlanganiswa elenza kube lula ukuklama kwesistimu futhi lehlise izindleko zizonke.

Izixazululo ezishumekwe zaseShandong Penghui zinikeza izinzuzo ezimbalwa ezihlukile ezizenza zifaneleke kahle ikusasa lokulawulwa kwezimpawu zomgwaqo. Okokuqala nokubalulekile ukusebenza okuphezulu: i-RK3399 chip inikeza isikhala esanele sokubala ukuze kuqhutshwe ama-algorithms ayinkimbinkimbi okuthuthukisa ukuhamba kwezimoto kanye namamodeli e-AI ngaphandle kokwethula ukubambezeleka. Lapho ama-milliseconds ebalulekile — njengalapho kulungiswa izigaba zesiginali ukuvimbela ingozi noma ukuvula indlela yemoto yesimo esiphuthumayo — la mandla okucubungula ahumusha ngqo ekuphepheni okuthuthukisiwe nasekusebenzeni kahle. Isakhiwo samakhora ayisithupha siphinde sivumele imisebenzi eminingi ukuthi isebenze ngasikhathi sinye, njengokucubungula izithombe zekhamera ngenkathi kugcinwa umjikelezo wokulawula isiginali wesikhathi sangempela, konke kubhodini elilodwa.

Okwesibili, ukwakhiwa kwezinga lezimboni kwamabhodi ayinhloko e-Penghui kuqinisekisa ukuthi angakwazi ukumelana nezingcindezi zemvelo ezitholakala kumakhabethe omgwaqo, okuhlanganisa ukushisa okwedlulele, amakhaza, umswakama, nokudlidliza. Izinto zikagesi zezinga labathengi zingahluleka ngokushesha ngaphansi kwezimo ezinjalo, okuholela ekumeni kwesikhathi eside okubizayo nasesondweni. Ngokuphambene, i-Shandong Penghui iklama imikhiqizo yayo ukuze isebenze isikhathi eside okungenani iminyaka eminingana ngokusebenza okuqhubekayo, isekelwa ukuhlolwa okuqinile nokulawulwa kwekhwalithi. Okwesithathu, ukusekelwa kwe-AI yemvelo okunikezwa yi-NPU kwenza ukuthi kusetshenziswe i-Edge inference ngaphandle kokuthumela idatha yevidiyo efwini, kugcinwe umkhawulokudonsa futhi kuthuthukiswe ubumfihlo bedatha. Lokhu kubaluleke kakhulu njengoba amadolobha amukela izibani zomgwaqo ezihlakaniphile ezincike ekuboneni kwekhompyutha ukuze kutholwe izimoto, ukubalwa kwabahamba ngezinyawo, nokutholwa kwezigameko. Ukuze ufunde kabanzi mayelana nenkampani engemuva kwalezi zinto ezintsha, vakashelaMayelana nathi ikhasi. Ungahlola futhi uhla oluphelele lwezixazululo ku- Ikhaya ikhasi.

Ikusasa lezimpawu zokuhamba kwemigwaqo ngokungenakuphikwa lihlakaniphile, lixhumene, futhi liyazivumelanisa nezimo. Njengoba amadolobha emhlabeni wonke ebhekene nezinselelo zokuminyana kwezimoto, ukungcola, kanye nokuphepha emigwaqweni, ukushintsha kusuka kuzibali zokudala zokuhamba kwemigwaqo kuya kuzilawuli zezimpawu zokuhamba kwemigwaqo ezisebenza nge-AI nezisekelwe ku-IoT kuyakhula. Lezi zinhlelo zithembisa ukunciphisa izikhathi zokuhamba ngo-20-30%, ukunciphisa ukukhishwa kwegesi ezimotweni kakhulu, nokudala izindawo zokuhlangana eziphephile ngokutholwa kwezingozi ngesikhathi sangempela. Ubuchwepheshe obusiza — i-AI, i-IoT, kanye nedatha enkulu — sekuvele bukhulile, kodwa ukusetshenziswa kwabo ngempumelelo kuncike ekubeni nehadiwe ethembekile, esebenza kahle kakhulu efakwe e-Edge.

I-Shandong Penghui Intelligent Technology Co., Ltd ihlangabezana nalesi sidingo ngemikhiqizo efana nebhodi eliyisisekelo le-FET3399-C, elihlanganisa ukuqina kwezinga lezimboni, ukusheshisa kwe-AI okunamandla, kanye nokuxhumana okuguquguqukayo ngendlela ehlangene. Ngokukhetha izixazululo ezigxilile ezinjalo, abahlanganisi bezinhlelo zokuhamba bangakha izilawuli zesiginali eziqinisekisela ikusasa ezilungele ukusingatha amanani edatha kanye nobunkimbinkimbi be-algorithm yohlelo lwezokuhamba oluhlakaniphile lwakusasa. Ukuthuthuka kokuhamba kwedolobha kuncike ekuqanjweni okusha kuzo zonke izigaba, kusukela kuma-algorithm enza ngcono ukugeleza kwaze kufike ku-hardware ewasebenzisa ngokuthembekile emkhakheni. Ngezinkundla ezigxilile eziqinile, umbono wamanethiwekhi okuhamba aguquguqukayo ngokuphelele, azithuthukisayo ususeduze kakhulu ukuba ube yinto ejwayelekile emigwaqweni yethu.