Ngomhlaka-9 Juni 2026, uMnyango Wezimboni Nobuchwepheshe Bolwazi kanye neKhomishini Yokuphatha Nokuphatha Izimpahla Eziphethwe Nguhulumeni yoMkhandlu Kahulumeni bakhiphe isaziso sokwethula ngokusemthethweni i-2026 Special Action for Real-World Training of Humanoid Robots and Embodied Intelligence. Lesi saziso sithi, ekupheleni kuka-2026, imikhiqizo yobuhlakani efakiwe njengamarobhothi asebenza njengabantu kufanele iqedele ukuqinisekiswa kwesicelo kanye nokuthunyelwa okuvamile ezimweni ezikhethiwe ezimele futhi iqhubekele ekusetshenzisweni kokusebenza. Ihlose futhi ukuthuthukisa izimo zesicelo ezingaphezu kwe-100 zenani eliphezulu kanye nokwakha ikhono lokuthumela imikhiqizo esikalini samayunithi ayizinkulungwane eziyishumi. Lesi senzo esikhethekile sigxile ezimweni ezibalulekile emikhakheni yezimboni, yezinjongo ezikhethekile, kanye nezinsizakalo, ngenkathi kukhuthazwa izikhala zokuqeqesha zomhlaba wangempela, izinhlangano zezicelo ezintsha, ukuthuthuka kwamakhono omsebenzi, kanye nokuqinisekiswa kokuthunyelwa.
Ngesikhathi esifanayo, intuthuko yokuzivikela emngceleni ehlakaniphile yaseShayina ithuthuka ngesivinini esingakaze sibonwe. NgoMashi kulo nyaka, irobhothi eliqaphayo elimpunga elisiliva lafakwa eKarakoram Plateau endaweni ephakeme ngamamitha angu-5,300 eceleni komngcele weShayina neNdiya. Ngohambo oluzimele kanye nokuqapha okujikelezayo kwamadigri angu-360, lachazwa abezindaba baseNdiya ngokuthi “umlindi wesayensi.” Lesi yisibonelo esicacile sokuvikelwa komngcele okunikwe ubuchwepheshe eShayina okusuka emqondweni kuya ekusetshenzisweni kwangempela. Yakhelwe izindawo ezikude kakhulu zomngcele njengokuphakama okuphezulu, ukubanda okukhulu, umoya-mpilo ophansi, kanye nendawo eyinkimbinkimbi, irobhothi lingasebenza ngokuzinzile ngamamitha angu-5,300 ngaphezu kolwandle, lapho amazinga omoya-mpilo engaba ngu-60% kuphela walawo asethafeni, futhi lingamelana namazinga okushisa asukela ku--40°C kuya ku-55°C. Echwebeni lase-Erenhot e-Inner Mongolia, irobhothi eliqaphayo elihlakaniphile elakhiwe ngokuzimela yisiteshi sokuhlola umngcele sendawo seliqalile ukusebenza, okwenza kube lula ukushintsha kusuka ekuhlolweni okulandela isitimela ngesandla kuya ekuhlolweni okuhlakaniphile kwezitimela zabagibeli bamazwe ngamazwe. E-Guangxi, imoto yokuqapha ehlakaniphile enekhono lokushayela elizimele le-L4 isiqalile ukusebenza eDongxing, echwebeni laseFangcheng, ihlinzeka ngokuqapha okuzimele amahora angama-24 ngosuku, ukuqashelwa okunembile, kanye nesixwayiso ngesikhathi sangempela. Kusukela emingceleni esenyakatho kuya eningizimu, kusukela emathafeni kuya emachwebeni, “abalindi bensimbi” abangakhathali manje basemsebenzini ubusuku nemini.
Eqhutshwa yizinqubomgomo ezisekelayo nezidingo zangempela zokusebenza, amarobhothi okuhlola ukuqapha imingcele ashintsha ngokushesha kusukela ekujikelezeni okuzenzakalelayo kuya ekwenzeni izinqumo okuzenzakalelayo. Njengoba amamodeli e-AI eqeqeshwe ngamacala ezehlakalo zemingcele, izici zendawo, kanye namaphethini okusebenza, amarobhothi athola kancane kancane ikhono lokuhlukanisa izakhamuzi, abasebenzi abahlomile, kanye nemfuyo, ngenkathi ehlonza ngokunembile ukuziphatha okuyingozi kakhulu njengokubuthana okusolisayo noma ukusetha imishini. Lokhu kubeka izidingo ezidinga kakhulu kumlawuli oyinhloko: ukubala i-AI enamandla ukuze kutholakale ukuqonda okujulile ngesikhathi sangempela, ukutholwa okuvumelanisiwe kusuka kuzinzwa eziningi, isethi ecebile yezixhumi zokuxhumana zezimboni, kanye nokuzivumelanisa nezimo okuqinile kwemvelo. Empeleni, umlawuli kumele ahlinzeke ngokusebenza okuphezulu kwe-AI, asekele okokufaka okuningi kwekhamera kokubona okune-panoramic okungu-360-degree, futhi anikeze izixhumi zokuxhumana zezinga lemoto njenge-CAN FD kanye ne-RS232/485 ukuze axhumane nezinhlelo ze-chassis kanye nezinzwa.
Ingemuva Lecala
Emingceleni emide yomhlaba, imisebenzi yokuqapha ibhekene nezinselele eziningi, okuhlanganisa izindawo ezinzima, indawo eyinkimbinkimbi, abasebenzi abalinganiselwe, kanye nengozi ephezulu. Ukuqapha okuvamile okwenziwa ngesandla akugcini nje ngokungasebenzi kahle; ezimweni ezimbi kakhulu njengokuphakama okuphezulu, ukubanda okukhulu, kanye nokwehluka okukhulu kwezinga lokushisa, ukukhuthazela ngokomzimba nokuphepha kwabasebenzi nakho kuyavivinywa kakhulu. Ngokuthuthuka okusheshayo kobuhlakani obubonakalayo, amarobhothi ahlakaniphile okuhlola ukuqapha umngcele ahlanganisa ukuzulazula okuzenzakalelayo, ukuqonda kwemvelo, ukuqashelwa kwe-AI, kanye nokuxhumana ngesikhathi sangempela sekuba imishini ebalulekile yokufaka esikhundleni somsebenzi wezandla emizileni enengozi ephezulu kanye nokwenza kube lula ukuqapha umngcele kuzo zonke izimo zezulu.
Umhlinzeki ohamba phambili wasekhaya wezixazululo zamarobhothi akhethekile ubhekane nezinselele ezilandelayo ngenkathi ethuthukisa uhlelo lokulawula oluyinhloko lwerobhothi lakhe lokuhlola imingcele elihlakaniphile lesizukulwane esilandelayo:
Inselele Yamakhasimende
1. Isiphetho se-AI esisebenza kahle kakhulu: Irobhothi lokuhlola i-border patrol lidinga ukucubungula ukusakazwa kwevidiyo okuningi kwe-HD ngesikhathi sangempela futhi lenze imisebenzi yokucabanga yokufunda okujulile njengokuqashelwa kwabasebenzi nezimoto kanye nokuhlaziywa kokuziphatha okungavamile. Lokhu kudinga ukusebenza okuqinile kokubala i-AI ngenkathi kugcinwa ukusetshenziswa kwamandla ngaphakathi kwebanga elifanele lepulatifomu yeselula esebenzisa ibhethri.
2. Ukutholwa kwamakhamera amaningi okuvumelanisiwe: Irobhothi idinga ukuhlanganisa amakhamera amaningi e-HD namakhamera okushisa e-infrared ukuze kufezwe umbono wemvelo ophelele ngamadigri angu-360. Isilawuli kumele sisekele izixhumi eziningi zekhamera ye-GMSL2 kanye nokuvumelanisa isikhathi se-GNSS ukuqinisekisa ukuhambisana okunembile kohlaka kuyo yonke imifudlana yesithombe.
3. Ukusekelwa kwezinqubo zokuxhumana zezimoto: Amarobhothi okuhlola ukuqapha imingcele avame ukwakhiwa kumapulatifomu e-chassis anamasondo noma alandelelwe. Izinhlelo zawo zokushayela izimoto zixhumana ngebhasi le-CAN, kuyilapho izinzwa ezifana ne-LiDAR, i-IMU, kanye ne-GNSS/BeiDou nazo zidinga ukuxhunywa. Ngakho-ke, isilawuli kumele sisekele ngokwemvelo izixhumi zokuxhumana zezimboni ezifana ne-CAN FD kanye ne-RS232/485.
4. Ukudluliswa kwedatha ngesikhathi sangempela nge-bandwidth ephezulu: Ividiyo ye-HD kanye nedatha yenzwa eqoqwe ngesikhathi sokuhamba ngemoto kumele idluliselwe emuva esikhungweni somyalo ngesikhathi sangempela. Ngakho-ke isilawuli sidinga i-bandwidth yenethiwekhi ye-10GbE noma ye-Gigabit ukuqinisekisa ukubambezeleka okuphansi kanye nokudluliswa kwedatha okuthembekile kakhulu.
5. Ukuzivumelanisa nezimo ezibucayi:Amarobhothi okuhlola ukuqapha imingcele kumele asebenze ngokwethembeka ngaphansi kwezimo ezinzima njengokuphakama okuphezulu, ukubanda okukhulu, umehluko omkhulu wokushisa, kanye nokudlidliza okunamandla. Lokhu kubeka izidingo eziphezulu kakhulu zokusebenza kokushisa okubanzi, ukumelana nokudlidliza, kanye nokumelana nokushaqeka.
6. Idizayini encane futhi elula:Isikhala ngaphakathi kwerobhothi sincane kakhulu. Isilawuli kumele sihlale sincane kakhulu ngenkathi sigcina ukusebenza okuphezulu, okusiza ukunciphisa isisindo sonke sohlelo.
Isixazululo se-APQ
Ngemva kwezikhathi eziningi zokukhetha umkhiqizo onzima kanye nokuhlolwa kwendawo, ikhasimende likhethe i-APQ TER30J-C3 Embodied Intelligent Robot Controller njengesikhulumi esiyinhloko sokubala kanye nokuxhumana serobhothi layo elihlakaniphile lokuhlola imingcele. Isixazululo sinikeza izinzuzo ezilandelayo ezibalulekile:
01 275 Amandla ekhompyutha e-TOPS ukuze uthole isiphetho se-AI sesikhathi sangempela:
I-TER30J-C3 ifakwe imojuli yochungechunge lwe-NVIDIA Jetson AGX Orin, enikeza ukusebenza kwe-AI okufika ku-275 TOPS. Ingasebenzisa amamodeli okufunda okujulile njengokutholwa kwezinto, ukuqashelwa kokuziphatha, kanye nokuhlukaniswa kwe-semantic ngqo emaphethelweni, okuvumela ukuhlonza kanye nokuxwayisa ezingeni le-millisecond kubasebenzi, izimoto, kanye nokuziphatha okungajwayelekile. Lokhu kunikeza ukwesekwa okuqinile kwekhompyutha kokwenza izinqumo ezizimele kumarobhothi okuhlola imingcele.
02 8 Izixhumi zekhamera ze-GMSL2 zokubona okune-panoramic:
I-TER30J-C3 iza ejwayelekile enezixhumi zekhamera ze-GMSL2 ezingu-8 futhi isekela ukuvumelanisa isikhathi se-GNSS. Lezi zixhumi zingaxhumanisa amakhamera amaningi abonakalayo e-HD namakhamera okushisa e-infrared ngesikhathi esisodwa, okuvumela ukuqonda okuzungezile okungu-360-degree. Ukutholwa kwedatha yesithombe esineziteshi eziningi kunikeza ukwesekwa kwedatha esezingeni eliphezulu kwemisebenzi ebalulekile njengokuzulazula okuzenzakalelayo okusekelwe ku-SLAM, ukulandelela okuqondiwe, kanye nezaziso zemicimbi engavamile.
03 I-CAN FD kanye namachweba amaningi ochungechunge wokuxhumeka kwe-chassis kanye ne-sensor:
I-TER30J-C3 isekela iziteshi ezingu-2*RS232/RS485, iziteshi ezingu-2*CAN FD, kanye neziteshi ezingu-8*GPIO. Iziteshi ezingu-2*CAN FD zingaxhuma ngqo ohlelweni lokushayela i-chassis motor robot ye-patrol ngaphandle kwemojuli eyengeziwe yokwandisa i-CAN. Amachweba e-RS232/485 angaxhuma izinzwa ezifana ne-LiDAR, i-IMU, kanye nama-encoder. Iziteshi ezingu-8*GPIO zingalungiswa kalula njengezixhumi ze-actuator zokukhipha i-alamu, ukulawula ukukhanya, kanye neminye imisebenzi. Lokhu kulungiselelwa kwesixhumi kuhambisana nezidingo zokuxhumana zamabhasi amaningi zamarobhothi eselula anamasondo nalandelelwa.
04 10GbE kanye ne-Gigabit Ethernet ekabili ukuze kuhlanganiswe idatha ngendlela ephumelelayo:
I-TER30J-C3 inikeza imbobo eyodwa ye-10GbE kanye namachweba amabili azimele e-Gigabit Ethernet. Imbobo ye-10GbE ivumela ukukhishwa okusheshayo kwedatha yokuqeqesha yamamodeli amakhulu, ihlinzeka ngezinto eziningi zekhwalithi ephezulu zokuthuthukisa imodeli ye-AI eqhubekayo. Amachweba amabili e-Gigabit angasetshenziswa ngokwehlukana kokufakwa kwedatha ye-LiDAR point-cloud kanye nokuxhumana nesikhungo sokulawula i-backend, okuvumela ukusakazwa kwedatha ukuthi kuhlukaniswe futhi kuncishiswe ukuminyana kwenethiwekhi.
05 Ukwandiswa okungenantambo kokuxhumeka kweselula:
I-TER30J-C3 isekela ukunwetshwa kwe-M.2 Key-B kwamamojula we-4G/5G kanye nokunwetshwa kwe-M.2 Key-E kwamamojula we-WiFi/BT. Ezimweni lapho amanethiwekhi anezintambo kunzima ukuwasebenzisa, ukuxhumana kwe-4G/5G kuqinisekisa ukuxhumana kwesikhathi sangempela phakathi kwerobhothi kanye nesikhungo somyalo, kuyilapho i-WiFi/BT isekela ukulungisa amaphutha endaweni kanye nokuxhumana okufushane.
06 Ishasi eliqinile kakhulu elinomklamo wokushisa okubanzi kanye nowokulwa nokudlidliza ezindaweni ezibucayi:
Iyunithi ephelele ilinganisa kuphela u-118.2mm x 130mm x 57.6mm, ubukhulu obucishe bufane nesundu. I-chassis yayo ye-aluminium alloy kanye nomklamo wayo onama-bracket-mount kuvumela ukuthi ifakwe kalula esikhaleni sangaphakathi esilinganiselwe serobhothi eliqaphayo. Ngokuvumelana nezimo zemvelo, i-TER30J-C3 isekela izinga lokushisa lokusebenza eliphakathi kuka--20°C no-50°C kanye nezinga lokushisa lokugcina eliphakathi kuka--40°C no-80°C, futhi idlule ekuvivinyweni kokudlidliza okungahleliwe kwe-3Grms kanye nokuhlolwa kokushaqeka kwe-15G. Iphinde isekele ukufaka kwe-DC okubanzi okungu-12-28V, okuvumela ukuhambisana ngqo nezinhlelo zebhethri zezimoto ezingama-24V ezivame ukusetshenziswa kumarobhothi okuhlola amarobhothi emingcele, isusa isidingo samamojula engeziwe okuguqulwa kwamandla kanye nokwenza lula umklamo wokunikezwa kwamandla jikelele.
Izimpumelelo Zenani:
Ngokusebenzisa i-APQ TER30J-C3 Embodied Intelligent Robot Controller, umhlinzeki wezixazululo zerobhothi lokuhlola imingcele uphumelele ekufezeni imiphumela elandelayo:
Ukusebenza kahle kokuqagela kwe-AI okuthuthukisiwe kakhulu: Ngama-TOPS angu-275 wamandla ekhompyutha, amamodeli okufunda okujulile angasebenza ngesikhathi sangempela emaphethelweni. Isivinini sokuphendula sokuqashelwa kwezinto kanye nokuhlaziywa kokuziphatha okungavamile siphakeme kaningana kunezixazululo zendabuko, okunikeza isisekelo sekhompyutha samarobhothi ukuze aguquke kusukela ekuqapheni okuzenzakalelayo kuya ekwenzeni izinqumo okuzenzakalelayo.
Uhlelo lokuqonda oluhlanganiswe kakhulu lwemithombo eminingi: Ukuhlanganiswa kwezixhumi zekhamera ze-GMSL2 ezingu-8, ukuvumelanisa isikhathi se-GNSS, iziteshi ezimbili ze-CAN FD, kanye neziteshi ezimbili ze-RS232/485 kwenza kube lula ukufinyelela kwesilawuli esisodwa kumakhamera e-HD, amakhamera okushisa e-infrared, ama-chassis motors, i-LiDAR, i-IMU, namanye amadivayisi. Lokhu kwenza kube lula kakhulu ukwakheka kwesistimu futhi kunciphisa izindleko ze-BOM kanye nobunzima bokuhlanganiswa.
Umkhawulokudonsa owanele wokuhlanganiswa kwedatha:Ukwakhiwa kwe-10GbE kanye ne-Gigabit Ethernet ekabili kusekela ukukhishwa okusheshayo kwedatha yokuqeqesha yamamodeli amakhulu futhi kuqoqa izinto ezisezingeni eliphezulu ukuze kuqhutshekwe nokwenza ngcono i-algorithm ye-AI. Iphinde ivumele ukuhlukaniswa kokusakaza kwedatha okubushelelezi, iqinisekise izixhumanisi zokuxhumana ezizinzile nezithembekile phezu kwezimo zemingcele ekude.
Ukuzivumelanisa nezimo okunamandla ezindaweni ezinzima: I-chassis ye-aluminium alloy eyinkimbinkimbi kakhulu, ukwakheka kokusebenza kokushisa okubanzi, ukudlidliza kwe-IEC kanye nokumelana nokushaqeka, kanye nokufakwa kwe-voltage ebanzi engu-12-28V kuvumela isilawuli ukuthi sisebenze ngokuqina kumarobhothi okugada emingceleni ngaphansi kokudlidliza kwesikhathi eside, izinguquko zokushisa ezibukhali, kanye nokushintshashintsha kwamandla kagesi. Lokhu kunciphisa kakhulu amazinga okuhluleka endaweni kanye nezindleko zokulungisa.
Njengoba uMnyango Wezimboni Nobuchwepheshe Bolwazi kanye neKhomishini Yokuphatha Izimpahla Eziphethwe Nguhulumeni YoMkhandlu Kahulumeni beqala ngokubambisana uhlelo lokuqeqesha umhlaba wangempela lobuhlakani obubonakalayo, futhi njengoba intuthuko yokuvikela imingcele ehlakaniphile ishesha ezweni lonke, "ukuvikela imingcele ehlakaniphile" yaseShayina kuthuthuka ngokushesha kunanini ngaphambili. Kusukela "kumlindi wesayensi" eKarakoram Plateau kuya kurobhothi lokuqapha elihlakaniphile e-Erenhot, kusukela emotweni yokuqapha yezokuphepha ye-L4 eGuangxi kuya ekuqalisweni okujulile kwamaphrojekthi okuthengwa kwamarobhothi emingcele ezweni lonke, lokhu kushintsha kusuka ekuqapheni okusekelwe kubasebenzi kuya ekulawulweni kwemingcele okunikwe amandla ubuchwepheshe kudala izidingo ezidingekayo zabalawuli abayinhloko emandleni okusebenzisa ikhompyutha, ukuhlanganiswa kwezinzwa, i-bandwidth yokuxhumana, kanye nokuzivumelanisa nezimo kwemvelo.
Njengoba ine-AI computing tops engu-275, ama-interfaces ekhamera angu-8 e-GMSL2, i-10GbE kanye ne-Dual Gigabit Ethernet networking, kanye ne-interfaces yokuxhumana kwezimoto eziphelele kufaka phakathi i-CAN FD kanye ne-RS232/485, i-APQ TER30J-C3 Embodied Intelligent Robot Controller ilungele izidingo eziyinhloko zamarobhothi okuhlola ama-border patrol ezindaweni ezibucayi. Isekela ukuqagela kwe-AI okusebenzayo okuphezulu, ukutholwa kwamavidiyo amaningi avumelanisiwe, ukubuyiselwa kwedatha ngesikhathi sangempela, kanye nokuhlanganiswa kwamadivayisi amaningi. Usayizi wayo oqinile kakhulu, ukusebenza kwezinga lokushisa elibanzi, ukumelana nokudlidliza, kanye nomklamo wezimboni we-wide-voltage kwenza kube "ubuchopho obuhlakaniphile" obuthembekile bamarobhothi ahambayo anamasondo nalandelwayo ekuhlolweni kwemingcele, ukushayela okuzenzakalelayo, kanye nezinhlelo zokusebenza zama-patrol ezihlakaniphile.
Isikhathi sokuthunyelwe: Juni-29-2026
