I-Mingshang ADAS: Izinhlelo Zokusiza Umshayeli Ezithuthukile Zokushayela Okuzimele Okuphephile

Kwadalwa ngo 07.15

I-ADAS kaMingshang: Izinhlelo Ezithuthukile Zokusiza Abashayeli Zokushayela Okuphephile Okuzimele

Izimoto zanamuhla ziguquka ngokushesha zisuka ezintweni ezilula zokuhamba zibe yizinkundla ezihlakaniphile zokuhamba, futhi enhliziyweni yalolu shintsho kukhona umqondo wokushayela kwe-ADAS. Incazelo yegama elithi ADAS, elimele i-Advanced Driver-Assistance Systems, ibhekisela eqoqweni lobuchwepheshe be-elekthronikhi nesofthiwe eklanyelwe ukukhulisa ukuphepha kwezimoto nokusiza umshayeli enqubweni yokushayela. Yonke imoto ye-ADAS yanamuhla incike ekuhlanganisweni okuyinkimbinkimbi kwezinzwa, amakhamera, i-radar, nobuhlakani bokwenziwa ukuze ibone indawo ezungezile futhi yenze izinqumo zesikhathi sangempela. I-Mingshang ivele njengamandla ahole phambili kulo mkhakha, ihlinzeka ngezixazululo ezibanzi ze-ADAS ezinika amandla abakhiqizi bezimoto nabahlinzeki bezinga lokuqala ukuthi bakhe izinhlelo zokushayela ezizimele eziphephile, ezihlakaniphile, nezithembekile kakhulu. Ngokuhlanganisa ama-algorithms asezingeni eliphezulu obuhlakani bokwenziwa nezinkundla zamahardware ezisezingeni lezimboni, inkampani ihlinzeka nge-ecosystem ephelele ehlanganisa konke kusukela ekuqoqweni kwedatha nokulingiswa kwe-neural kuya kumayunithi okucubungula akhulisekayo nokuqinisekiswa kokuphepha kokusebenza. Lesi sihloko sinikeza ukuhlola okujulile, okugxile enhlanganisweni, kwemikhiqizo eyinhloko ye-Mingshang, izinto ezintsha zobuchwepheshe, ubambiswano lwamasu, imininingwane yemboni, kanye nezigigaba zakamuva, kubonisa ukuthi kungani inkampani ibekwe njengomnikazi wamandla obalulekile ekushintsheni komhlaba wonke kusiya emazingeni aphezulu okushayela okuzenzakalelayo.

Isingeniso: Ukubuka konke kwezixazululo ze-ADAS zikaMingshang

Imboni yezimoto zomhlaba wonke ibhekene nokushintsha okukhulu okungakaze kubonwe, njengoba abalawuli, abathengi, nabahlinzeki bezobuchwepheshe bebambisana ukuze bafinyelele amazinga aphezulu okuphepha nokunethezeka okukhulu ekushayeleni. Kulesi simo, ukubaluleka kokushayela kwe-ADAS akunakweqiwa, njengoba izinhlelo zanamuhla manje zisebenzisa imisebenzi ebucayi efana nokubhuleka okuzenzakalelayo okuphuthumayo, usizo lokugcina umzila, ukulawula ukuhamba ngejubane eliguquguqukayo, nokuqashelwa kwezimpawu zomgwaqo. I-Mingshang ibhekana nalezi zinselele ngendlela ehlelekile nephothiwe, inikeza abakhiqizi bezimoto amathuluzi abawadingayo ukuze basheshise imijikelezo yokuthuthukisa kuyilapho behlisa izindleko nobunkimbinkimbi. Iphothifoliyo yenkampani yakhelwe ngokukhethekile izidingo eziguqukayo zokushayela okuzimele kwe-Level 2+ kuya ku-Level 4, igxile ezixazululweni ezilungele ukukhiqizwa ezisukela ezimotweni zabantu ezisezingeni lokungena kuya ezimikhunjini zezohwebo ezisezingeni eliphezulu. Ngokunikeza isakhiwo sobuchwepheshe esihlangene esihlanganisa ukuphathwa kwedatha, ukulingisa, ukucutshungulwa kwemizwa, nokuhlanganiswa kwemoto, i-Mingshang inika ozakwethu amandla okuletha izici zokuphepha ezithuthukisiwe emakethe ngokushesha nangokuzethemba okukhulu. Leli su elibanzi alithuthukisi nje kuphela ukuphepha kwazo zonke izimoto ze-ADAS emgwaqweni, kodwa futhi linciphisa umthwalo wobunjiniyela obuvame ukuhlotshaniswa nokuhlanganisa izingxenye ze-software ne-hardware ezihlukene ezivela kubathengisi abaningi.

Ifilosofi kaMingshang igxile enkolelweni yokuthi inkululeko yangempela ayitholakali ngaphandle kwemigudu yedatha eqinile, izindawo zokulingisa ezingokoqobo, kanye nezakhiwo zokuphepha eziqinisekiswayo. Izixazululo zenkampani ziklanywe kusukela ekuqaleni ukuze zihlangabezane nezidingo eziqinile zokuthuthukiswa kwesofthiwe yezimoto, okuhlanganisa nokuhambisana namazinga omhlaba wonke afana ne-ISO 26262 yokuphepha kokusebenza. Ukwengeza, uMingshang ugcizelela kakhulu ukuguquguquka, evumela amakhasimende ukuthi alungise ipulatifomu ngokuya ngezinhlobo zezinzwa ezithile, ukusebenza kwemoto, kanye nezimo zokushayela zendawo. Lokhu kuguquguquka kubaluleke kakhulu emakethe lapho izinhlaka zokulawula nokulungela kwengqalasizinda kwehluka kakhulu ezindaweni ezahlukene. Ngokuhlanganisa ubuchwepheshe bomkhakha nokuqonda okujulile kwencazelo ye-ADAS nezinselelo zayo zokusebenza, uMingshang uzibeke njengozakwethu othembekile kuma-OEM nabahlinzeki bezinga lokuqala emhlabeni wonke.

Imikhiqizo Eyinhloko: Inhlekelele Ye-Full-Stack Yokushayela Okuzimele

I-Mingshang Drive: Isitaki Sokushayela Esizenzakalelayo

I-Mingshang Drive iyisitaki esiyinhloko se-software yokushayela ezenzakalelayo yenkampani, eklanyelwe ukuphatha yonke inqubo ye-perception-planning-control ngesikhathi sangempela. Le nkundla ye-modular ihlanganisa ama-neural network ajulile ukuze ithole izinto, ukuhlukaniswa kwe-semantic, nokubikezela kwendlela, konke kwenziwe kahle ukuze kusebenze kahle kwi-hardware efakwe kuyo. Lesi sitaki sisekela izinhlobo eziningi zezinzwa, okuhlanganisa amakhamera, i-LiDAR, i-radar, nezinzwa ze-ultrasonic, futhi sihlanganisa okukhiphayo zazo kube yimodeli yemvelo emunye. Omunye womehluko obalulekile we-Mingshang Drive yikhono layo lokusebenza ezimweni eziningi zemvelo, kusukela ekukhanyeni kwemini okugqamile kuya emvuleni enkulu nasezimweni zokukhanya okuphansi, okwenza ifanele ukusetshenziswa ezimakethe ezihlukahlukene zomhlaba. Uhlaka lwakhiwe phezu kohlaka lwe-microservices oluvumela izingxenye ngazinye ukuthi zibuyekezwe noma zishintshwe ngaphandle kokuthinta isimiso sonke, okunikeza ama-OEM ithuba lokuthuthukisa ngokushesha njengoba ubuchwepheshe buvela. Yonke imoto ye-ADAS enikwa amandla yi-Mingshang Drive ihlomula ekuthuthukisweni okuqhubekayo kwe-over-the-air, iqinisekisa ukuthi izici zokuphepha nokuthuthukiswa kokusebenza zifinyelela ezimotweni isikhathi eside ngemva kokuba ziphumile efektri.

I-Mingshang Data: Ukulungiswa Okunembile Nokuhlaziywa Kwemininingwane

Idatha iwumgogodla wanoma yiluphi uhlelo lwe-ADAS olusebenzisa i-AI, futhi i-Mingshang Data inikeza inkundla ebanzi yokuqoqa, ukuhlela, ukumaka, nokuphatha amasethi amakhulu edatha adingekayo ukuze kuqeqeshwe amamodeli aqinile okubona. Inkundla inikeza izindlela ezizenzakalelayo zokumaka ezisebenzisa ukufunda okusebenzayo ukuze kubekwe phambili amasampula anolwazi oluningi, kunciphisa kakhulu umsebenzi wokumaka ngesandla kuyilapho kuthuthukisa ukunemba kwemodeli. I-Mingshang Data isekela izinhlobo eziningi zokumaka, okuhlanganisa amabhokisi okumboza, amapholigoni, amaphoyinti abalulekile, ama-3D cuboids, nama-semantic segmentation masks, konke ngezindlela zokuqinisekisa ikhwalithi eziqinisekisa ukungaguquguquki emaqenjini amakhulu. Inkundla iphinde ibe namathuluzi akhelwe ngaphakathi okulawula izinguqulo zedatha, ukuhlunga ubumfihlo, nokukhiqiza idatha yokwenziwa, okuvumela onjiniyela ukuthi bakhe amasethi okufundisa alinganiselayo ahlanganisa izimo eziyingozi ngokuphephile nangokonga. Ngokusebenzisa i-Mingshang Data, amaqembu okuthuthukisa anganciphisa isikhathi esisuka kumarekhodi ezinzwa eziluhlaza kuya kudatha yokuqeqesha elungele ukukhiqizwa ngaphezu kuka-50%, okusheshisa ngokuqondile ukukhishwa kwezici ezintsha zokushayela ze-ADAS.

I-Mingshang Sim: Ukulingisa Okungokoqobo Kokuhlola

I-Mingshang Sim iyinkundla yokulingisa enembile kakhulu eklanyelwe ukuqinisekisa ama-algorithms e-ADAS kanye nokushayela okuzimele endaweni ephephile, ephindaphindekayo, futhi ekhulisekayo. Le nkundla ihlanganisa izindlela ezihlanganisiwe zokudweba ezihlanganisa izithombe ezisekelwe kwi-physics kanye nezinkambu ze-neural radiance, zikhiqize izigcawu ezingokoqobo ezingenakwehlukaniswa nokuqoshwa komhlaba wangempela. Lokhu kunika amandla onjiniyela ukuthi bahlole izitaki zabo ezigidini zamamayela abonakalayo, kuhlanganise nezigcawu ezingavamile ezifana nokugxuma kwabahamba ngezinyawo, izindawo zokwakha, kanye nezimo zezulu ezingezinhle ezingaba yingozi noma ezingenzeki kalula ukuphinda zenziwe ezinkundleni zokuhlola ezibonakalayo. I-Mingshang Sim iphinde ihlangane nayo yonke i-ecosystem ye-Mingshang, ivumele ukudluliswa okungenazihibe kwezigcawu nezilinganiso phakathi kokulingisa nokuhlola komhlaba wangempela. Le nkundla isekela ukucushwa kwe-hardware-in-the-loop (HIL) kanye ne-software-in-the-loop (SIL), okwenza kube ithuluzi elibalulekile kuzo zombili izinqubo zokuthuthukisa nokuqinisekisa. Nge-Mingshang Sim, indlela esuka emcabangweni iye ekusetshenzisweni iba esheshayo, ephephile, futhi ingabizi kakhulu.

I-Mingshang Ware: I-IP ye-Neural Processing Unit

Ukuze kulethwe ukusebenza kwekhompyutha okudingwa izinhlelo zanamuhla ze-ADAS, i-Mingshang inikeza i-Mingshang Ware, umgogodla wobuhlakani bokucubungula (NPU) oyikhiqizo elingakalwa. I-Mingshang Ware yakhelwe ukusheshisa ukufunda okujulile ngokusebenza kahle kwamandla, ifinyelela kuze kube ngu-20 TOPS nge-watt ngayinye emisebenzini evamile yezimoto. Ubuchwepheshe be-NPU busekela inhlobonhlobo yama-neural network operators, okuhlanganisa i-convolution, i-pooling, izingqimba eziphindaphindayo, namabhulokhi e-transformer, okwenza kube nokuvumelana ngokwanele ukuphatha izakhiwo zamamodeli amanje navesikhathi esizayo. I-Mingshang Ware ingahlanganiswa njenge-accelerator ezimele noma njengengxenye yedizayini enkulu ye-system-on-chip (SoC), futhi iza nesethi ephelele yamathuluzi okuthuthukisa isofthiwe elula ukusetshenziswa kwamamodeli. Ngokuhlinzeka ngokusebenzisana kwe-hardware nesofthiwe kusukela ezingeni le-IP kuye phezulu, i-Mingshang inika amandla abakhi bama-chip nabahlinzeki be-Tier 1 ukuthi bakhe ama-processor aqondene nomkhakha athuthukisa ukusebenza kwe-ADAS ekushayeleni kuyilapho kunciphisa ukusetshenziswa kwamandla nendawo ye-die.

Ukucwiliswa Okujulile Kwezobuchwepheshe: Ubuhle Bobunjiniyela Ngemuva Kwenkundla

Ukuhlelwa Kwedatha Okunamandla Kwe-AI Nokuphawula Okuzenzakalelayo

Enye yemisebenzi edinga izinsiza eziningi kakhulu ekuthuthukisweni kwe-ADAS ukwakhiwa kwamasethi edatha anelebulwe kahle. I-Mingshang ibhekana nalesi sezinselele ngokusebenzisa injini yokuhlela idatha enobuhlakani bokwenziwa (AI) ekhetha ngokuzenzakalelayo amasampula abaluleke kakhulu kumarekhodi ezinzwa eziluhlaza. Injini isebenzisa ukulinganisa kokungaqiniseki, ukukhetha okuhlukahlukene, kanye nama-heuristics aqondene nomkhakha ukukhomba izehlakalo ezingavamile nezimo eziyinselele, iqinisekisa ukuthi idatha yokuqeqesha ihlanganisa umsila omude wezimo zokushayela zomhlaba wangempela. Uma idatha isihleliwe, ipayipi lokulebula ngokuzenzakalelayo lisebenzisa inhlanganisela yamamodeli aqeqeshiwe ngaphambilini kanye nokucabanga kwesikhashana ukukhiqiza amalebula okuqala, abese ecwengwa ngenqubo yokuqinisekisa ebandakanya umuntu. Le ndlela inciphisa umzamo wokulebula ngesandla ngamaphesenti angafika ku-80 kuyilapho igcina noma ithuthukisa ukunemba kwamalebula uma iqhathaniswa nezinqubo zokusebenza ezenziwa ngokuphelele ngesandla. Ipayipi lonke liklanyelwe ukuthi likhule kalula kusukela kumaphrojekthi okuhlola ngamagigabhayithi kuya emikhunjini yokukhiqiza ngamaphetabhayithi, okwenza kufanelekele wona womabili amaqembu amancane okuthuthukisa kanye nezinhlangano ezinkulu zezimoto.

Ukurendwa Okuxubile Nokulingisa Kwe-Neural

I-Mingshang Sim isebenzisa ukwakheka kokurenda okuxubile okuhlanganisa ukwenza imifanekiso ngemigqa (rasterization) kanye nezinkambu zemisebe ye-neural (NeRF) ukuze kukhiqizwe izigcawu ezinembile ngokwefiziksi futhi zibukeka zingokoqobo ngokubonakalayo. Ungqimba lwe-rasterization luphatha ukuma okungaguquki, ukukhanya, nezinto zokwakha kusetshenziswa amasu okwenza imifanekiso ngokusekelwe emzimbeni (PBR), kuyilapho ungqimba lwe-neural lwenza amamodeli wokubonakala okuyinkimbinkimbi, ukubonakala ngokungacacile, nezinto ezinyakazayo okunzima ukuzimelela ngezindlela ezijwayelekile. Le ndlela exubile ivumela ukulingiswa ukuthi kutholakale amazinga wezithombe ngesikhathi sangempela kumahardware ajwayelekile ezimoto, kugcinwe ukucaca okubonakalayo okudingekayo ukuze kuqinisekiswe izinzwa nokuhlolwa kokuqonda. Ngaphezu kwalokho, ipulatifomu ihlanganisa isihleli sezigcawu esivumela onjiniyela ukuthi bachaze ukuhlukahluka okunemingcele kwezimoto, isimo sezulu, nokuma komgwaqo, ngokuzenzakalelayo kukhiqizwe izinkulungwane zokucushwa kokuhlola kusukela kusifanekiso esisodwa. Ngokuhlanganisa ukulingiswa kwe-neural nokukhiqizwa kwezigcawu okuzenzakalelayo, i-Mingshang Sim inikeza indawo yokuqinisekisa ephumelela kakhulu kunalezo zindlela zendabuko zokuhlola emzileni wokugijima.

Ukuklama Okuhambisanayo Kwe-Hardware-Software Okukalekayo

Ipulatifomu ye-Mingshang yakhelwe phezu komgomo wokuklama ngokuhambisana kwe-hardware nesoftware, okusho ukuthi isitaki sesoftware kanye nesakhiwo se-compute esingaphansi kuklanywa ndawonye ukuze kukhuliswe ukusebenza kahle kusukela ezingeni le-silicon kuye phezulu. Le ndlela ibonakala ku-Mingshang Ware, lapho i-instruction set architecture (ISA) iqondaniswa ngokucophelela namaphethini okubala wezinhlelo zanamuhla zokubona nokuhlela. Isitaki sesoftware sihlanganisa i-compiler ehlanganisa ngokuzenzakalelayo amagrafu e-neural network kuma-tensor cores e-NPU, ukuhleleka kwememori, kanye nezindlela zokudlulisa idatha, okususa isidingo sokwenza ukuthuthukiswa ngesandla ezimweni eziningi. Ezingeni lesistimu, i-Mingshang ihlinzeka ngezakhiwo eziyisisekelo ezibonisa indlela yokuhlanganisa i-NPU namakhamera, i-radar, nezinye izinzwa, kanye nama-driver nama-middleware afihla imininingwane ye-hardware. Le ndlela yokuklama ngokuhambisana iqinisekisa ukuthi isitaki sonke se-ADAS sisebenza ngokulibaziseka okuncane kanye nokudlulisa idatha okuphezulu, okuvumela yonke imoto ye-ADAS ukuthi iphendule ezingozini ngemizuzwana eyizinkulungwane kunokuba ngemizuzwana.

Ukuphepha Nokuqinisekiswa (ISO 26262)

Ukuphepha akusiyona into ecatshangwa ngemuva kwalokho enqubweni yokuthuthukisa umkhiqizo yakwaMingshang; kuyimfuneko eyisisekelo yomklamo. Izingxenye zesoftware nehardware zenkampani zithuthukiswa ngokuhambisana ne-ISO 26262, izinga lamazwe ngamazwe lokusebenza kokuphepha kwezimoto. I-Mingshang Drive ne-Mingshang Ware zithole isitifiketi se-ASIL (Automotive Safety Integrity Level) kuze kufike ezingeni le-ASIL-D, eliphakeme kakhulu, elihlanganisa ukutholwa kwamaphutha, ukuphathwa kwamaphutha, nezindlela zokuvikela ekwehlulekeni. Umjikelezo wokuthuthukisa uhlanganisa ukuhlaziywa kwezingozi nokuhlolwa kwengozi, ukuhlolwa okuhlelekile kokufaka amaphutha, kanye nokuhlolwa kokuphepha okuzimele esigabeni ngasinye. I-Mingshang futhi ihlinzeka ngemibhalo ephelele yokuphepha, ehlanganisa izincwajana zokuphepha, ukuhlaziywa kwezindlela zokwehluleka nemiphumela (FMEA), kanye nemibiko yokumboza ukuxilongwa, okusiza ama-OEM ukuhlanganisa ipulatifomu ezinqubweni zawo zokuqinisekisa kalula. Lokhu kuzibophezela ekuphepheni kunikeza abakhiqizi bezimoto ukuzethemba ukuthi ukusebenzisa isoftware ye-ADAS yakwaMingshang ngeke kuphazamise isibopho sabo sokuvikela abashayeli, abagibeli, nabahamba ngezinyawo.

Ubambiswano Nokubambisana

Nabakhiqizi Bezinzwa Nama-Chip

Mingshang isungule ubudlelwano bobuhlakani namakhamphani ahamba phambili ekwenzeni izinzwa nama-chip ukuze kuqinisekiswe ukuthi isofthiwe yayo ihlangana kahle nezingxenye zekhompyutha ezingaphansi. Inkampani isebenza eduze nabahlinzeki bamakhamera, ama-LiDAR, nama-radar ukuze ilinganise amamodeli ezinzwa, ilungise izinqubo zokucubungula amasignali, futhi ithuthukise ama-algorithms okuhlanganisa ukuze ahambisane nezinhlobo ezithile zezinzwa. Ngasohlangothini lwe-silicon, uMingshang usebenzisana namakhamphani ama-semiconductor ukuze athumele i-NPU IP yayo nesitaki sesofthiwe ezinhlobonhlobo zama-process node nama-SoC architecture, okwenza amakhasimende akwazi ukukhetha ihadiwe efanelana nezidingo zawo zezindleko nokusebenza kahle. Lobu budlelwano busekelwa amathimba obunjiniyela ahlanganyelwe enza imikhankaso yokuqinisekisa ukusebenza, ahlole ukusebenza kwezinkundla eziyizithenjwa, futhi ashicilele imihlahlandlela yokuhlanganisa enciphisa isikhathi sokungenisa izinhlelo ezintsha zezimoto emakethe. Ngokuhambisana nabaholi bezinzwa nama-chip, uMingshang uqinisekisa ukuthi izixazululo zawo zokushayela ze-ADAS zihlala ziphambili ekusungulweni kwehadiwe ngaphandle kokulahlekelwa ukuguquguquka kwesofthiwe.

Nabakhiqizi Bezimoto kanye Nabahlinzeki Bezinga Lokuqala

Iqhinga likaMingshang lokufinyelela emakethe lincike kakhulu ekusebenzisaneni okujulile nabakhiqizi bezimoto (OEMs) kanye nabahlinzeki bezingxenye zezinga lokuqala (Tier 1 suppliers) abahlanganisa inkundla ezimotweni ezikhiqizwayo. Inkampani isayine izivumelwano eziningi zokuthuthukisa nabakhiqizi bezimoto bomhlaba wonke ukuze ithumela i-Mingshang Drive ezimodelini ezizayo, igxile ezimakethe zase-Asia, eYurophu, naseNyakatho Melika. Lokhu kubambisana ngokuvamile kuqala ngesigaba sokufakazela umqondo (proof-of-concept), lapho ithimba lobunjiniyela likaMingshang lisebenza ndawonye nethimba lezimoto nezokuphepha le-OEM ukuze liguqule isitaki ukuze sihambisane nenkundla yemoto ethile nokwakheka kwezinzwa. Ngemva kokuqinisekiswa, ubudlelwano budlulela esigabeni sokukhiqiza esihlanganisa ilayisensi yevolumu, ukunakekelwa okuqhubekayo, nokuphathwa kokubuyekeza okungenantambo (over-the-air). Ngaphezu kwalokho, uMingshang ubambisana nabahlinzeki bezingxenye zezinga lokuqala ukuze afake i-Mingshang Ware kuma-domain controllers nama-ADAS ECUs, enze amamojula alungele ukuhlanganiswa okwenza lula uchungechunge lokuhlinzeka lwe-OEM. Lolu hlelo lokusebenzisana lusiza uMingshang ukuthi wandise ubuchwepheshe bakhe emikhakheni yezimoto eziningi kuyilapho kugcinwa ukwenziwa ngendlela oyifunayo abakhiqizi bezimoto abakhulu abakudingayo.

Imibono Yemboni: Izitayela ze-ADAS kanye Nombono we-Mingshang

Imakethe ye-ADAS ikhula ngezinga lokukhula ngonyaka elihlanganisiwe elingaphezu kuka-12%, iqhutshwa yimithetho ephoqelelayo yezici zokuphepha, isidingo sabathengi sokuthola okulula, kanye nokwehla kwezindleko ze-sensor ne-computing hardware. Enye yezinto ezibaluleke kakhulu ukushintsha kusuka ku-ADAS egxile ezicini ukuya ezakhiweni ezihlanganisiwe, ezilawulwa yizindawo, lapho i-SoC eyodwa enamandla iphethe yonke imisebenzi yokubona, yokuhlela, nokwenza. I-Mingshang ikulungele kahle ukuhola lolu shintsho ngoba i-software stack yayo kanye ne-NPU IP yakhelwe ngokukhethekile izakhiwo ezihlanganisiwe, okwenza amakhasimende akwazi ukunciphisa inani lama-ECU emotweni kusuka emashumini ambalwa kuya kwambalwa nje. Okunye ukuthambekela okubalulekile ukusetshenziswa okwandayo kokuthuthukiswa okuqhutshwa yidatha, lapho idatha yangempela yokushayela iqoqwa njalo, ihlungwe, futhi isetshenziselwa ukuthuthukisa ukusebenza kwemodeli ngokuhamba kwesikhathi. I-Mingshang Data ne-Mingshang Sim zakha umjikelezo ovalekile osekelana nale ndlela, okuvumela ama-OEM ukuthi asebenzise amamodeli athuthuka ngemayela ngalinye elishayelwa yizimoto zabo. Umbono wesikhathi eside wenkampani uwukwenza ukuthutha emgwaqweni kungabikho ukufa, ngokufaka zonke izimoto ngezinga eliyisisekelo lekhono lokushayela le-ADAS elingathuthukiswa nge-software ngokuhamba kwesikhathi. Lo mbono uhambisana nezinhlelo zomhlaba wonke ezifana ne-Vision Zero kanye ne-New Car Assessment Program (NCAP) roadmaps, zombili eziphusha ukuthi kufakwe ngempoqo izici ezithuthukile zokuphepha kuzo zonke izimoto ezintsha. Ngokuhlinzeka ngenkundla evulekile, ekhulayo, neqinisekiswayo, i-Mingshang isiza ukwenza ubuchwepheshe be-ADAS bufinyeleleke kumuntu wonke, okwenza ukushayela okuphephile kube yiqiniso hhayi kuphela kwabathengi bezimoto zikanokusho kodwa kuwo wonke umuntu.

Izindaba Nezigcawu Ezibalulekile

Ukumenywa Kwemikhiqizo Nokubuyekezwa

Esonyakeni odlule, iMingshang imemezele izibuyekezo ezimbalwa ezinkulu zomkhiqizo ezandisa amandla epulatifomu yayo. Inguqulo yakamuva yeMingshang Drive ifaka ukusekelwa kwamamodeli okubona asuselwa ku-transformer, abonise ukuthuthuka okukhulu ekunembeni ekutholeni izinto nasekubikezeleni ukunyakaza uma kuqhathaniswa nezakhiwo ze-convolutional. I-Mingshang Sim ibuyekeziwe ngomshini omusha wokudweba we-neural onciphisa igebe lomkhakha phakathi kwedatha yezinzwa yokwenziwa nengempela ngama-40%, okwenza imiphumela yokulingisa ibikezeleke kakhulu ekusebenzeni kwangempela. I-Mingshang Data manje ifaka ukuhlunga okuzenzakalelayo kokuvikeleka okususa ulwazi olungaziveza olufana nobuso nezinombolo zezimoto, okusiza amakhasimende ukuthi ahambisane nemithetho yokuvikelwa kwedatha efana ne-GDPR kanye noMthetho WaseChina Wokuvikelwa Kolwazi Lomuntu Siqu. Lezi zibuyekezo zibonisa ukuzibophezela kwenkampani ekusunguleni okuqhubekayo nokusabela kwayo empendulweni yamakhasimende nasezidingweni zemakethe. I-Mingshang futhi ihlinzeka ngokukhishwa kwezibuyekezo ezijwayelekile kanye nezibuyekezo zokuphepha kumakhasimende ayo akhiqizayo, iqinisekisa ukuthi izinhlelo ezisetshenziswayo zihlala ziphephile futhi zihambisana nomthetho kuso sonke isikhathi sokuphila kwazo.

Iminikelo Yocwaningo Nezinkomfa

I-Mingshang igcina umnyango wocwaningo osebenzayo oshicilela amaphepha ezingqungqutheleni nakumajenali asezingeni eliphezulu ngezihloko ezihlobene nokuqonda, ukulingisa, nokuphepha kokusebenza. Izincwadi zakamuva zihlanganise izindawo ezinjengokutholwa kwezinto ngokuqaphela ukungaqiniseki ezinhlelweni ezibucayi zokuphepha, ukusesha ngempumelelo kwe-neural architecture ukuze kufakwe kuma-embedded system, kanye nokuqina kokulwa nobutha bamapayipi okuhlanganisa izinzwa eziningi. Abacwaningi benkampani bethula njalo emicimbini efana ne-IEEE Intelligent Vehicles Symposium, i-International Conference on Computer Vision (ICCV), kanye ne-Automotive Safety & Security Conference, lapho babelana ngolwazi futhi behlanganyela nomphakathi wesayensi obanzi. Ngaphezu kokushicilela, i-Mingshang ixhasa ukubambisana kwezemfundo nezinhlelo zomthombo ovulekile ezithuthukisa ubuchwepheshe be-ADAS nokushayela okuzimele. Lokhu kufakwa kwesandla akugcini nje ngokuqinisa idumela lobuchwepheshe lenkampani kodwa kusiza nasekuheheni amathalente aphezulu afuna ukusebenza ezinkingeni ezinomthelela oqondile ekuphepheni emgwaqweni.

Imibuzo Ejwayelekile (FAQ)

Iyini incazelo ye-ADAS kumongo wobuchwepheshe be-Mingshang?

I-ADAS imele i-Advanced Driver-Assistance Systems, okuyizinhlelo zikagesi nezesofthiwe ezisiza abashayeli ngokuzenzakalelayo imisebenzi ethile yokuphepha njengokubhuleka, ukuqondisa, nokugcina umzila. Kumongo kaMingshang, i-ADAS ibhekisela kuyo yonke inhlekelele yamathuluzi nezinkundla inkampani ehlinzeka ngazo ukuze ivumele abakhi bezimoto ukuthi bathuthukise, bahlole, futhi basebenzise lezi zinhlelo, okuhlanganisa i-perception stack, ukulungiswa kwedatha, ukulingisa, kanye ne-neural processing hardware IP.

I-ADAS driving stack kaMingshang yehluka kanjani kwabancintisana nabo?

Isitaki se-ADAS sika-Mingshang sigqama ngenxa yezakhiwo zaso ezihlukanisekile, ezisekelwe kuma-microservices ezisekela izibuyekezo ezingaphezulu komoya, uhlaka lwayo lokuphepha olusebenzayo olugunyazwe yi-ASIL-D, kanye nokuhlanganiswa kwayo okuqinile nezinkundla zokuqoqa idatha nokulingisa. Le ndlela yokuphela ekugcineni inciphisa ubunkimbinkimbi bokuhlanganisa futhi isheshise isikhathi sokufika emakethe uma iqhathaniswa nokuhlanganisa izixazululo ezivela kubathengisi abaningi.

Ingabe izixazululo ze-ADAS zika-Mingshang zingasetshenziswa endaweni ekhona ye-ADAS yemoto?

Yebo, i-Mingshang Drive yakhelwe ukuthi inganciki kuzinzwa futhi ivumelane nezinkundla ezahlukene, okusho ukuthi ingaguqulelwa ezakhiweni ezikhona zemoto ngoguquko oluncane kwi-hardware. Inkampani ihlinzeka ngemihlahlandlela yokuhlanganisa, imiklamo eyisibonelo, nokusekelwa kobunjiniyela ukusiza abakhiqizi ukuthi badlulise isitaki kumalungiselelo abo athile ezinzwa nezinkundla zokubala.

Yikuphi ukuqinisekiswa u-Mingshang anakho mayelana nokuphepha kokusebenza?

I-Mingshang Drive ne-Mingshang Ware zithole ukuqinisekiswa kwe-ISO 26262 kuze kufike ku-ASIL-D, okuyizinga eliphezulu kakhulu le-Automotive Safety Integrity Level elichazwe yizinga. Lokhu kuqinisekiswa kuhlanganisa ukutholwa kwamaphutha okuhlelekile, ukuphathwa kwamaphutha, nezindlela zokuvikela ukwehluleka, kuqinisekisa ukuthi isofthiwe nehadiwe kuhlangabezana nezidingo eziqinile zokuphepha zezimoto zokukhiqiza.

Ingabe i-Mingshang inikeza izingxenye zehadiwe ezifana namakhamera noma izinzwa ze-radar?

I-Mingshang igxile kakhulu kusofthiwe, amapulatifomu edatha, amathuluzi okulingisa, kanye ne-NPU IP kunokukhiqiza izinzwa ezibonakalayo. Nokho, inkampani ibambisana eduze nabakhiqizi bezinzwa nama-chip abahamba phambili ukuze iqinisekise ukuthi isitaki sayo sesofthiwe silungiselelwe ngokugcwele kumakhamera amasha, i-LiDAR, i-radar, nehadiwe ye-SoC etholakala emakethe.

I-Mingshang Data ithuthukisa kanjani ukusebenza kahle kokuthuthukiswa kwe-ADAS?

I-Mingshang Data isebenzisa ukuhlunga okusebenzisa i-AI kanye nokumaka ngokuzenzakalelayo ukunciphisa umsebenzi wokumaka ngesandla ngokufika ku-80% kuyilapho ithuthukisa ikhwalithi yedatha. Injini yokufunda esebenzayo yepulatifomu ibeka phambili amasampula anolwazi oluningi, futhi amathuluzi ayo akhelwe ngaphakathi okulawula izinguqulo, ukuhlunga ubumfihlo, nokukhiqiza idatha yokwenziwa adala amasethi okuqeqesha alinganiselayo ngokushesha kunezindlela zendabuko zokusebenza ngesandla.

Iqhaza elidlalwa ukulingisa enqubweni yokuqinisekisa i-ADAS ye-Mingshang yini?

Ukulingisa kubalulekile ekuhloleni ngokuphepha izigidi zezimo zokushayela, kuhlanganise nezehlakalo ezingavamile neziyingozi ezingeke ziphindwe emzileni wokuhlola ongokoqobo. I-Mingshang Sim ihlinzeka ngezindawo ezinokubonakala okungokoqobo, ezihlanganiswe ngendlela ehlanganisiwe, nokusekelwa kwe-hardware-in-the-loop, okunika amandla onjiniyela ukuqinisekisa ama-algorithms okubona nokuhlela ngendlela ephindaphindekayo nekhuphukayo ngaphambi kokusetshenziswa emgwaqweni.

Ingabe ubuchwepheshe be-Mingshang bufanele izimoto zezohwebo nezimoto ezinkulu?

Yebo, ngokuphelele. Ukwakheka kwe-modular ye-Mingshang Drive nokukhula kwe-Mingshang Ware kuvumela ipulatifomu ukuthi ijwayele amaloli ezohwebo, amaveni, amabhasi, nezinye izimoto ezinkulu. Inkampani isebenze nabaqhubi bezimoto ukuze yenze ngokwezifiso isitaki sezilinganiso ezinkulu zezimoto, izindawo ezahlukene zokubeka izinzwa, kanye nezindawo ezithile zokusebenza ezifana nokushayela emgwaqweni omkhulu nokuhamba edepho.

I-Mingshang ihlala injani ngokuthola izintambo zakamuva ku-ADAS nokushayela okuzimele?

I-Mingshang igcina umnyango wocwaningo osebenzayo oshicilela ezinkomfeni zamazwe ngamazwe nezincwadi zezifundo, ixhasa ukubambisana namanyuvesi, futhi ibamba iqhaza emicimbini yezimboni emhlabeni wonke. Lokhu kuqinisekisa ukuthi imikhiqizo yenkampani ifaka izintambo zakamuva ekuboneni, ekulingiseni, nasekuphepheni, okugcina amakhasimende ayo ephambili kubuchwepheshe be-ADAS.

Ngingathola kuphi ulwazi oluningi ngemikhiqizo ye-Mingshang nokubambisana?

Ungahlola iwebhusayithi esemthethweni ye-Mingshang, uvakashele amakhasi omkhiqizo ukuze uthole imininingwane yobuchwepheshe, ubheke isigaba Sezindaba ukuze uthole izimemezelo nezigcawu ezibalulekile, noma uxhumane nethimba Lokusekelwa ukuze uthole imibuzo eqondile. Ngaphezu kwalokho, labo abanentshisekelo ngezingxenyekazi zekhompyutha ezihlobene njengamakhamera emoto bangathola ikhasi lekhamera yemoto lilusizo ekuqondeni ukuthi ukuhlanganiswa kwe-sensor kusebenza kanjani nesitaki se-ADAS se-Mingshang.

Ngenisa izithombe zakho futhi
sizokhuluma nawe.