InoteveraGaidhi rinotevera
CLIP uye Vision-Mutauro Models
AI Yekutarisa
Gwara reVisual AI
Vision-Language-Action Models yeRobhoti
Vision-language-action (VLA) models inogadzirisa zviito zverobhoti zvichienderana nekuona uye mirairo yemutauro.
Teerera kugwaro iriAI-yakarondedzerwa · ~2 min
Pa peji ino2 min verenga
Pfupiso
Mapepa akataurwa eRT-2, π0, uye Open X anotsanangura zvivakwa zvakasiyana uye ongororo.
Kudzika Kwakadzika
Vision-language-action models inogadzirisa zviito zverobhoti zvichienderana nekuona kwekamera uye mirairo yemutauro. RT-2 inomiririra zviito zverobhoti se-tokens dzakakamurwa, kusanganisira shanduko munzvimbo ye-end-effector uye kutenderera uye mamiriro e-gripper. DeepMind yakagadzirisa zvakare muenzaniso we-vision-language wakange watogadzirirwa uchishandisa data rerobhoti uye data rewebhu, ndokuzoshuma zvinopfuura 6,000 zvekuyedza robhoti. Mienzaniso yayo yekuchinjisa inobuda inobata nezvinhu zvakaedzwa uye mirairo; haigadziri generalization yebasa isina tsarukano. Sisitimu yakasiyana, π0, inowedzera flow-matching action expert inogadzira zviito zvinoenderera mberi. Bepa rinoshuma kudzora pamitengo inosvika 50 Hz yemabasa akaedzwa akadai sekupeta hembe. Open X-Embodiment inounganidza zviratidzo kubva kune 22 marobhoti kutsigira tsvakiridzo yemarobhoti akawanda. Hukuru hweDataset, kumiririrwa kwezviito uye kuchinjisa kwakaratidzwa izvivakwa zvakasiyana, chimwe nechimwe chichida ongororo mune yakatarwa setup.
Kuitwa muNyika yeChokwadi
Mhedzisiro yebepa: RT-2 inomiririra zviito zverobhoti zvakakamurwa se-tokens kune transformer.
Paper result: π0 uses a flow-matching action expert in its reported robot-control setup.
Resource yebepa: Open X-Embodiment inounganidza zviratidzo kubva kune akawanda marobhoti.
Zvinoratidzira: ongorora policy yakadzidziswa pane zvinhu zvitsva chete mukati meprotocol yekuyedza yakachengeteka.
Ramangwana reVision-Language-Action Models yeRobhoti
Ongorora VLA transfer pane target embodiment, zvinhu, uye seti yemirairo. Shuma robhoti, zvinodzorwa, uye zvinobudirira; mhedzisiro kubva kune imwe benchmark haigadziri kutumirwa kwese-chinangwa.
Kuzvishandisa
Nei zvichikosha
- Kumhanya uye chiyero. Visual AI inogona kuita otomatiki yekuongorora, yekuona, uye yekumaka mabasa pachiyero.
- Vaka sarudzo. Zvikwata zvekugadzira zvinogona prototype pfungwa nekukurumidza nekudzokororwa kwemaoko mashoma.
- Team uye workflow. Mashandisirwo anogona kushandisa masaini emifananidzo nemavhidhiyo ayo aimbove akaoma kugadzirisa.
Chenjera
- Kodzero dzemifananidzo uye kubvumirwa kunogona kuve njodzi dzepamutemo kana provenance isina kujeka.
- Kuita kwemuenzaniso kunogona kusiyanisa kupenya, huwandu hwevanhu, uye nharaunda.
- Manyepo enhema anogona kusacherechedzwa kunze kwekunge zvikumbaridzo zvekuvimba zvikatariswa.
Matanho anotevera
- Tsanangura maitiro ekugamuchirwa echokwadi, kurangarira, uye mutengo wekukanganisa.
- Edzai nedata rinoenderana nemamiriro chaiwo ekugadzira.
- Wedzera ongororo yemunhu kune yakaderera-kusavimbika kana yakakwirira-inokanganisa kufanotaura.
- Tevera modhi kudonha uye simbisa mushure mekuchinja kwekamera kana dataset.
Ramba Uchiongorora
Tsamba yemahara
Tora iyo yemazuva ese AI pfupiso
Nhatu dzakasimbiswa dzeAI nyaya mangwanani ega ega pakati pevhiki, dzakanyorwa neChirungu chakajeka. Yemahara zvachose, hapana ads.
Imwe email zuva rega rega revhiki. Bvisa kunyoresa kamwe chete. Hatimbotengese kana kugovera kero yako.
Zviedze iwe pachako
Tora Vision-Language-Action Models for Robotics quiz
Mhinduro yepakarepo pamhinduro yega yega, uye chitupa chinogovaniswa chine ID inobatika kana wapedza kosi.
Tsigira dzidzo yemahara yeAI. AI Understanding ndeye 501(c)(3) isingabatsiri - hapana ads, hapana paywall, zvachose. Ita chipo
Mibvunzo inowanzo bvunzwa
Chii chinonzi Vision-Language-Action Models yeRobhoti?
Vision-language-action (VLA) models inogadzirisa zviito zverobhoti zvichienderana nekuona uye mirairo yemutauro. Mapepa akataurwa eRT-2, π0, uye Open X anotsanangura zvivakwa zvakasiyana uye ongororo.
Ndeapi marudzi matatu ekupinza/zvinobuda anotsanangura Vision-Language-Action (VLA) modhi?
A VLA inotora chiratidzo (mifananidzo) pamwe nechinangwa chemutauro uye zvinobuda zviito (motor command), kubatanidza maonero, kuraira-kutevera, uye kutonga.
Ko Google DeepMind's RT-2 yakamiririra sei zviito zvemarobhoti kuti transformer igozvigadzira?
RT-2 yakasungira chidimbu chega chega kuita zviratidzo (semuenzaniso, 256 mabhini) uye ndokuaburitsa setambo, ichisiya mutauro-modhiyo muchina kufanotaura zviito semazwi.
Chii chinonzi 'emergent transfer' chinorevei mumamiriro ezvinhu eVLAs?
RT-2 inoshandisa “emergent transfer” kune zviitiko zvakaedzwa uko web-scale vision-language pretraining yakabatsira robhoti kupindura kune zvinhu zvitsva kana mirairo. Iyi imhedzisiro yebepa, kwete vimbiso yezvinhu kana mirairo isina tsarukano.
Chii chakakosha mukana wepi-0's diffusion/flow-matching action musoro uchienzaniswa neasing discrete action tokens?
Bepa reπ0 rinoshuma zviito zvinoenderera mberi kubva kune yayo flow-matching action expert pamwero wekudzora we50 Hz mune yakayedzwa setup. Uku ndiko kudzora kwakashumwa, kwete vimbiso yekugona kuri nani pane robhoti rega rega.
Sei iyo Yakavhurika X-Embodiment dataset ine basa kune VLAs?
Open X-Embodiment inounganidza zviratidzo zverobhoti kune akawanda marobhoti kutsigira tsvakiridzo yemarobhoti akawanda. Kusiyana-siyana kweDataset kunogonesa kudzidza; hazvivimbise kuti imwe policy inochinjira kune robhoti rega rega.
Ramba uchidzidza
Mazano anoenderana
Mamwe madhairekitori akasarudzirwa nyaya iyi