isiZulu
Ikhanda Lekusasa Lamamojuli Ekhamera Ne-LiDAR Ezimotweni Ezizihambelayo
Kwadalwa ngo 01.21
Ingxoxo mayelana namamojula ekhamera uma kuqhathaniswa ne-LiDAR ezimotweni ezizihambelayo isikhathi eside ibhekwa njengempi lapho oyedwa kuphela ozowina: u-Elon Musk uyakuchitha i-LiDAR njeng "intambo ebiza kakhulu," kanti i-Waymo ne-Huawei batshala izigidigidi ekuzweleni okusekelwe ku-laser ukuletha ukuzihambela ngokuphepha. Kodwa njengoba imboni yokuzihambela ingena endaweni ebalulekile ngo-2025, indaba entsha iyavela—eyokuthi lezi zibonelo ezimbili azizona izimbangi kodwa abalingani abadansayo emzamweni wokuzihambela okuthembekile ngempela. Lesi sihloko sihluza ukuthi kanjaniimikhiqizo yokukhanya kanye ne-LiDAR ziyaqhubeka nokuthuthuka, kungani ukuhlanganiswa kwazo kuba okungagwenywa, nokuthi lokhu kusho ukuthini esikhathini esizayo sokuhamba.
Ukuze siqonde ikusasa lawo, kufanele siqale siqaphele amandla ayisisekelo nemikhawulo eyakhelwe ngaphakathi echaza ubuchwepheshe ngabunye. Amakhamera, afana namehlo omuntu, aphumelela ekuthwebuleni ulwazi oluningiliziwe—imibala yezibani zomgwaqo, imigqa yemigwaqo, izimpawu zabahamba ngezinyawo, ngisho nesimo sezibani zamabhuleki zabanye abashayeli. I-LiDAR, ngokungafani, ikhipha imisebe ye-laser ukudala amamephu ayi-3D anembayo endaweni ezungezile, iletha ukuqonda okungenakuqhathaniswa kokujula kanye nokuqonda indawo okungalinganiswa amakhamera kuphela ngezigcawu eziyinkimbinkimbi ze-AI. Iminyaka, lemihluko isebenzele izimiso zobuchwepheshe eziphikisanayo: umbono ocacile osuselwe kusofthiwe ngokumelene nokuhlanganiswa kwezinzwa eziningi okunamandla okweqile.
Ukuvela Kwama-Module Ekhamera: Kusuka kuma-Pixel angu-2D kuya ekuqondeni okuhlakaniphile
Amamojuli ekhamera aseke enza intuthuko enkulu kusukela kumadivayisi alula okuthwebula izithombe kuya kumathuluzi obuhlakani bokubona, aqhutshwa yintuthuko ku-AI kanye ne-computational photography. Indlela ye-Tesla esebenzisa amakhamera kuphela, enikwa amandla yi-FSD V12 system yayo kanye nedatha yokushayela yangempela engaphezu kwezinkulungwane eziyizigidi eziyi-100, ikwazile ukufakazela ukuthi amakhamera angakwazi ukubhekana nezimo eziningi zokushayela zansuku zonke lapho ehambisana ne-neural networks ethuthukisiwe kanye ne-BEV (Bird’s Eye View) + Transformer architectures. Inzuzo enkulu yale ndlela ilele ekukaleni: isethi yamakamera ayi-8 ibiza ngaphansi kuka-$500, okuyingxenye encane yentengo yezinhlelo zokuqala ze-LiDAR, okwenza kube nokwenzeka ezimotweni ezithengiswa kakhulu.
Ukusungulwa kwezinto zakamuva kuhlale kwandisa amakhono ekhamera. Amakhamera ezimoto zesimanje manje asebenza ngaphezu kokukhanya okubonakalayo, esebenzisa izithombe ezishisayo ukuthola abahamba ngezinyawo ekukhanyeni okuphansi kanye nezinzwa eziseduze ne-infrared ukusika inkungu elula. Ngokwesofthiwe, ukufunda kwe-"shadow mode" kuvumela izinhlelo ezisekelwe ekhamera ukuthi zithuthuke njalo ngezigidi zezimo zokushayela ezifanayo, ngezibuyekezo ze-OTA zamasonto onke ezithuthukisa ukwenza izinqumo zazo. Kodwa-ke, amakhamera asabhekana nemingcele yomzimba engahlulwa: emvuleni enkulu, eqhweni, noma enkungu eminyene, izinga lawo lokubona liyehle kuze kufike ku-40%, futhi alulamele ukuqonda ubujamo ezindaweni ezingenazo izici ezifana nemigwaqo emikhulu engenalutho noma imihubhe enezindonga ezimhlophe.
Ukuvuselelwa kwe-LiDAR: Ukwehliswa kwezindleko kanye nokuqhubeka okukhulu kokusebenza
I-LiDAR, eyayiyinto yobuchwepheshe eyayigcinelwe izimoto ezisezingeni eliphezulu, isiguquke kakhulu ngenxa yomklamo oqinile kanye nomnotho omkhulu. Ngo-2018, iyunithi eyodwa ye-LiDAR yezimoto yayibiza cishe u-$800; ngo-2025, izinkampani ezifana neRoboSense zikwazile ukwehlisa amanani ngaphansi kuka-$200, kanti kulindeleke ukuthi ngo-2027 amayunithi azobiza ngaphansi kuka-$100. Leli shintsho elikhulu ngentengo libangelwa ukushintsha kusuka ku-LiDAR eyindilinga eyenziwe ngomshini kuya kwezisekelwe ku-solid-state, ezisusa izingxenye ezihambayo, zinciphisa usayizi, futhi zithuthukisa ukwethembeka—izinto ezibalulekile ekukhiqizeni ngobuningi.
Ukwanda kokusebenza nakho kube kuhle kakhulu. I-LiDAR ye-Huawei enamashaneli angu-192 ifinyelela ukulungiswa kwe-angular okungu-0.05°, okuyivumela ukuthi ibone abahamba ngezinyawo amamitha angu-200 kude—ngaphezu kokuphindwe kabili kwebanga elisebenzayo lamakhamera amaningi ezimoto. Ukuhlolwa kwe-Waymo okwenziwa ngempela kubonisa ukuthi i-LiDAR igcina ukuzinza kwedatha okuphakeme izikhathi ezi-3 kunezinhlelo zokubona emanzini emvula enkulu nemvula enkulu, okuxazulula isikhala esikhulu sokuphepha. Nokho, i-LiDAR ayiphelele: ibhekana nezinselelo ezindaweni ezibonakalayo njengezindonga zengilazi namachibi, okungadala izimo ze-"ghost braking", futhi ayikwazi ukuhlukanisa ulwazi olubhalwe ngemibala njengezibani zomgwaqo—okubalulekile ekuhambeni ezindaweni eziminyene zasemadolobheni.
Iphuzu Eliguqukayo: Kungani Ukuhlanganiswa Kushintsha Ukuqhudelana
Inganekwane yenzwa eyodwa "engcono kakhulu" isivele yahluleka ngezinkinga zangempela. Ngo-2024, iTesla eyayihamba nge-FSD V12 eLos Angeles yahluleka ukubona ichibi njengesithiyo, yagcina ibhuleke kakhulu yaze yaba sengcupheni yokushayisana ngemuva—lokhu kuyisici esijwayelekile sezinhlelo ezisebenzisa amakhamera kuphela. Ngokuphambene, izinhlobo zokuqala ezazisebenzisa i-LiDAR kuphela zahlehlisa ukubona izibani ezibomvu zomgwaqo elangeni eliqhakazile, okugqamisa ukwehluleka kobuchwepheshe ukucubungula izimpawu ezibonakalayo. Lezi zenzakalo ziye zakhuthaza ukuguqukela komkhakha ekuhlanganiseni izinzwa, ikakhulukazi "i-early fusion"—indlela ehlanganisa idatha eluhlaza evela kumakhamera ne-LiDAR esigabeni sokuqala sokucubungula, kunokuhlanganisa imiphumela esivele icacisiwe kamuva.
I-algorithm yakamuva ye-early fusion ye-Haomo.AI ibonisa amandla alendlela, yehlisa amaphutha okubona ngo-72% uma iqhathaniswa nezinhlelo ezisebenzisa inzwa eyodwa. Ngokuhambisa amaphikiseli ekhamera namachashaza e-LiDAR ngesikhathi sangempela, uhlelo lusebenzisa amandla omongo wekhamera kanye nokunemba kwesikhala kwe-LiDAR ukudala imodeli ephelele yemvelo. Ngokwesibonelo, e-Shenzhen ngesikhathi samahora okuphuma emsebenzini kusihlwa, i-Huawei ADS 3.0—ehlanganisa i-LiDAR enamashaneli angu-192 namakhamera angu-8—yaphumelela ukubona isithuthuthu esingakhanyisiwe siwela umgwaqo, isimo obekuzoba nzima kunoma iyiphi inzwa yodwa ukusibona.
Iziqhamo Ezivela Eziphakamiso Zokwakha Le Synergy
Izitayela ezintathu ezibalulekile zishintsha ubudlelwano phakathi kwamamojula amakhamera neLiDAR, zenze ukubambisana kwabo kube nomthelela omkhulu:
1. I-4D Millimeter Wave Radar njengesixhumi: I-Continental Group's latest 4D radar ifinyelela ku-0.5° angular resolution ngentengo ye-1/10 ye-LiDAR, isebenza njengengqikithi yokuhlanganyela phakathi kwamakhamera ne-LiDAR. Ithuthukisa ukukala kude ezimweni zezulu eziphakathi nendawo futhi inciphisa ukuncika kwi-LiDAR ezimweni ezingadingi kakhulu, iqhubeka ithuthukisa izinga lokusebenza kwezindleko.
2. Ukuhlanganiswa kwe-V2X Kwandisa Imingcele Yokubona: Inethiwekhi ye-5G-enikwe amandla yi-China vehicle-to-everything (V2X) manje isihlanganisa amakhilomitha angaphezu kuka-100,000 emigwaqweni, ihlinzeka ngedatha yethrafikhi yesikhathi sangempela neyingozi eyengeza izinzwa ezisemotweni. Kuleli hlelo, amakhamera ne-LiDAR zigxila ezindaweni eziseduze, kanti i-V2X igcwalisa izindawo ezingabonakali ngaphandle kobubanzi bezinzwa—kwakha "i-360°+" yokubona.
3. Ukuphathwa Kwezinsiza Zokuhlola Okusekelwe ku-AI: Izinhlelo ezizimele zesikhathi esizayo zizokwenza ngokushesha ukuhlela idatha evela kumakhamera noma i-LiDAR ngokuya ngezimo zokushayela. Ngesikhathi sokukhanya okucacile emigwaqweni, uhlelo lungase luthathe isinqumo sokusebenzisa amakhamera ukuze luthole amandla; ezindaweni ezinamafu ezisemadolobheni, luzoshintsha lube yi-LiDAR ukuze luthole ukunemba. Le ndlela yokuphatha iyandisa ukusebenza kahle ngenkathi igcina ukuphepha.
Izimo Zemboni Nezithombe Zepolitiki
Amacebo abenzi bezimoto abakhombisa kakhulu le ndlela yokuhlanganisa, kude nezikhundla ezinzima. I-BMW ifaka imali kumkhiqizi we-LiDAR i-Luminar kanye ne-Mobileye egxile kumakhamera; i-Volkswagen ibambisana ne-Horizon Robotics ngenkathi igcina izinketho ze-LiDAR. Ngisho ne-Tesla, umfanekiso ophawulekayo wombono ohlanzekile, isiqalile ukucwaninga ukuhlanganiswa kwe-LiDAR kumamodeli ayo e-robotaxi, okukhombisa ushintsho olungaba khona kumasevisi ezizimele ezithengiswayo.
Inqubomgomo iphinde ihambela izixazululo eziningi zezinzwa. I-China iyala i-LiDAR ezimotweni ezizihambelayo ze-L3+, kanti i-European NCAP izofaka i-LiDAR ohlelweni lwayo lokukalwa kwezokuphepha luka-2025. I-U.S. NHTSA ihlala ingathathi hlangothi kwezobuchwepheshe kodwa igcizelela "ukuphindaphinda" ezidingweni zokuphepha—ulimi oluncanyelwa ukuhlanganiswa kwezinzwa kunokuthembela ezinsizweni eziyodwa. Lezi zinguquko zokulawula zisheshisa ukwamukelwa kwezinhlaka ezihlanganisa ikhamera ne-LiDAR.
Umbono Wango-2027: Okugxilwe Ekhamereni Ngokuqinisekiswa kwe-LiDAR
Sibheke phambili ku-2027, ikusasa lamamojula ekhamera ne-LiDAR licacile: ukuhlanganiswa okuyigolide okuthi "ikhamera kuqala, i-LiDAR iqinisekisiwe" ukuze kube nokuzimela kwezinga le-L4. Amakhamera azohlala eyisendlalelo esiyinhloko sokubona, esisebenzisa izindleko zawo eziphansi, ukuqonda okuphezulu kwezimo, nokuthuthukiswa okuqhubekayo kwe-AI ukuze kubhekwane nezimo ezingama-90% zokushayela. I-LiDAR izosebenza njengenethi yokuphepha ebalulekile, isebenze ezimweni eziyingozi kakhulu—isimo sezulu esibi, izimpambano eziyinkimbinkimbi, izindawo zokwakha—ukuze inikeze idatha enemininingwane ye-3D evimbela amaphutha amakhulu.
Le synergy ixazulula inkinga eyinhloko yokuhamba okuzenzakalelayo: ukuhlela ukujula kokukhula nokuphepha. Amakhamera avumela ukutholwa kwenqwaba ngokugcina izindleko ziphansi, kanti iLiDAR ibhekana nezimo "ezikhethiwe" ezivimbele ukuzimela okuphelele. Njengoba amanani eLiDAR eqhubeka nokwehla futhi i-AI yekhamera iba yinkimbinkimbi, ukuhlanganiswa kwabo kuzoba yijwayelo kuwo wonke amazinga ezimoto ezizimele—kusukela kumasistimu e-ADAS abathengi kuya kumarobhothi.
Isiphetho: Ngaphezu Kokuncintisana, Ukuya Etrust
Ingxoxo ye-khamera vs. i-LiDAR ayikaze ibe ngempela mayelana nobuchwepheshe obuphakeme—yayihlanganisa ukwakha ukwethenjwa. Ukuze izimoto ezizimele zibe yinto ejwayelekile, kumele zibe nephephile kunezishayeli zabantu, futhi akukho sensor eyodwa engakwenza lokho yodwa. Amakhamera anikeza ubuhlakani bokuhlangenwe nakho kanye nokwandiswa; i-LiDAR iletha ukunemba nokwethembeka. Ikusasa lazo alikho ekuncintisaneni, kodwa ekuphakamiseni omunye nomunye.
Njengoba sihamba ngasezweni lokuhamba ngokuzenzakalelayo, umbuzo ngeke usaba "amakhamera noma i-LiDAR?" kodwa "ilungiselelwe kanjani kangcono?" Impendulo izochaza inkathi elandelayo yezokuthutha—eyodwa lapho ubuchwepheshe busebenza ngokuvumelana ukuletha isithembiso sokuzimela okuphephile, okufinyelelekayo, nokusebenza kahle kubo bonke.
Uxhumane
Sicela uxhumane nathi uhambele
WhatsApp
WeChat