Kamfanin DeepMind mallakin Google, wanda ya shahara wajen ci gabansa a fagen fasahar kere-kere da kuma gina hanyoyin sadarwa na jijiyoyi masu iya yin wasannin kwamfuta a matakin dan Adam, ya sanar da gano injin da zai kwaikwayi tsarin tafiyar da jiki MuJoCo (Multi-Joint dynamics with Contact). ). An yi amfani da injin ɗin ne don yin ƙirar ƙirar ƙira da ke hulɗa da muhalli, kuma ana amfani da ita don ƙirar ƙira a cikin haɓakar mutum-mutumi da tsarin bayanan ɗan adam, a matakin kafin aiwatar da fasahar da aka haɓaka ta hanyar na'urar da aka gama.
An rubuta lambar a C/C++ kuma za a buga ta ƙarƙashin lasisin Apache 2.0. Linux, Windows da macOS dandamali suna tallafawa. Ana sa ran kammala aikin buɗaɗɗen tushen duk abubuwan da ke cikin aikin a cikin 2022, bayan haka MuJoCo za ta matsa zuwa wani buɗaɗɗen ƙirar ci gaba wanda ke ba membobin al'umma damar shiga cikin ci gaban.
MuJoCo babban ɗakin karatu ne wanda ke aiwatar da injunan simintin tsari na gaba ɗaya wanda za'a iya amfani dashi a cikin bincike da haɓaka mutummutumi, na'urori masu sarrafa halittu da tsarin koyon injin, da kuma ƙirƙirar zane-zane, raye-raye da wasannin kwamfuta. An inganta injin simintin gyare-gyare don mafi girman aiki kuma yana ba da damar yin amfani da ƙananan matakan abu yayin da yake samar da daidaito mai kyau da kuma wadataccen damar yin kwaikwayo.
An bayyana samfura ta amfani da yaren bayanin yanayin MJCF, wanda ya dogara akan XML kuma an haɗa shi ta amfani da na'ura mai haɓakawa na musamman. Baya ga MJCF, injin yana goyan bayan loda fayiloli a cikin URDF na duniya (Unified Robot Description Format). MuJoCo kuma yana ba da GUI don kallon 3D mai mu'amala na tsarin simintin da samar da sakamakon ta amfani da OpenGL.
Babban fasali:
- Kwaikwayo a cikin haɗin kai gabaɗaya, ban da cin zarafin haɗin gwiwa.
- Juya ƙarfin kuzari, ana iya ganowa ko da a gaban lamba.
- Yin amfani da shirye-shiryen convex don ƙirƙira haɗe-haɗe a cikin ci gaba da lokaci.
- Ikon saita hani daban-daban, gami da taɓawa mai laushi da bushewar gogayya.
- Kwaikwayo na tsarin barbashi, yadudduka, igiyoyi da abubuwa masu laushi.
- Masu kunnawa (actuators), gami da injina, silinda, tsokoki, tendons da injunan crank.
- Masu warwarewa bisa Newton, gradient conjugate da hanyoyin Gauss-Seidel.
- Yiwuwar amfani da pyramidal ko elliptical friction cones.
- Yi amfani da zaɓinku na hanyoyin haɗin lamba na Euler ko Runge-Kutta.
- Multi-threaded discretization da iyaka iyaka.
source: budenet.ru