Sabuwar Neuroprosthetic shine Nasarar Robotics AI

Masana kimiyya a EPFL (École polytechnique fédérale de Lausanne) a Switzerland sun ba da sanarwar ƙirƙirar duniya ta farko don sarrafa hannun robotic - sabon nau'in neuroprosthetic wanda ke haɗar da ikon ɗan adam tare da aikin wucin gadi (AI) don girman girman robot kuma ya buga binciken su a cikin Satumba 2019 in Fasaha Na'urar Fasaha .

Neuroprosthetics (prosthetics na jijiyoyin jiki) na'urori ne na wucin gadi waɗanda ke motsawa ko haɓaka tsarin juyayi ta hanyar motsawar lantarki don rama raunin da ke shafar ƙwarewar motsi, sani, hangen nesa, ji, sadarwa, ko ƙwarewar ji. Misalan neuroprosthetics sun haɗa da hanyoyin sadarwa na kwakwalwa (BCIs), ƙwaƙƙwaran ƙwaƙwalwar ƙwaƙwalwa, masu motsa jijiyoyin jijiyoyin jini (SCS), shigar da sarrafa mafitsara, cochlear implants, da masu bugun zuciya.

Ana sa ran ƙimar karuwancin ƙafar ƙafa ta duniya zai zarce dalar Amurka biliyan 2.3 nan da shekarar 2025, a cewar alkalumman daga rahoton watan Agusta na shekarar 2019 ta Global Market Insight. A cikin 2018, ƙimar kasuwar duniya ya kai dala biliyan ɗaya bisa wannan rahoto. Kimanin Amurkawa miliyan biyu amintattu ne, kuma akwai sama da 185,000 da ake yankewa a kowace shekara, a cewar Cibiyar Ba da Lamuni ta Kasa. Cutar da jijiyoyin jini ta kai kashi 82 cikin dari na yankewar Amurka a cewar rahoton.

Ana amfani da prosthesis na myoelectric don maye gurbin sassan jikin da aka yanke tare da guntun wucin gadi na waje wanda tsoffin tsoffin mai amfani ke kunna su. Dangane da ƙungiyar bincike na EPFL, na'urorin kasuwanci da ake samu a yau na iya ba masu amfani babban matakin cin gashin kai, amma dexterity babu inda yake da ƙarfi kamar hannun ɗan adam.

“Na’urorin kasuwanci galibi suna amfani da tsarin tashar rikodi biyu don sarrafa matakin‘ yanci guda ɗaya; wato, tashar sEMG guda ɗaya don lanƙwasa da ɗaya don faɗaɗawa, ”in ji masu binciken EPFL a cikin binciken su. "Duk da yake da ilhama, tsarin yana ba da ƙarancin hankali. Mutane suna yin watsi da raƙuman myoelectric a cikin ƙima, a wani ɓangare saboda suna jin cewa matakin sarrafawa bai isa ya cancanci farashi da rikitarwa na waɗannan na'urori ba. ”

Don magance matsalar dexterity tare da myoelectric prostheses, masu bincike na EPFL sun ɗauki dabaru daban-daban don wannan binciken tabbataccen ra'ayi ta hanyar haɗa fannonin kimiyya na neuroengineering, robotics, da hankali na wucin gadi don sarrafa kansa wani ɓangare na umarnin motar don “raba iko. ”

Silvestro Micera, Shugaban Gidauniyar Bertarelli na EPFL a Fassarar Neuroengineering, da Farfesa na Bioelectronics a Scuola Superiore Sant'Anna a Italiya, suna ganin wannan hanyar haɗin gwiwa don sarrafa hannayen robotic na iya inganta tasirin asibiti da amfani don fa'idodi da yawa na dalilan neuroprosthetic kamar kwakwalwa -to-mashin mashin (BMIs) da hannayen bionic.

Masu binciken sun rubuta cewa "Dalilin da ya sa karuwancin kasuwanci ya fi yin amfani da kayan kwalliya na asali maimakon masu daidaitawa shine saboda masu rarrabewa sun fi ƙarfin kasancewa cikin wani yanayi," in ji masu binciken. “Don ganewa, irin wannan iko yana da kyau don hana faduwar bazata amma yana sadaukar da hukumar mai amfani ta hanyar ƙuntata adadin yuwuwar hannun. Ayyukanmu na sarrafa madaidaiciya yana ba da damar duka hukumomin mai amfani da fahimtar ƙarfi. A cikin sararin samaniya, mai amfani yana da cikakken iko akan motsin hannu, wanda kuma yana ba da damar yin siyayyar pre-preping don fahimta. ”

A cikin wannan binciken, masu bincike na EPFL sun mai da hankali kan ƙirar algorithms na software -kayan aikin robotic wanda ƙungiyoyin waje suka bayar ya ƙunshi Allegro Hand da aka saka akan robot KUKA IIWA 7, tsarin kamara na OptiTrack da firikwensin matsin lamba na TEKSCAN.

Masana kimiyyar EPFL sun ƙirƙiri kodar ƙima mai ƙima ta hanyar ƙirƙirar multilayer perceptron (MLP) don koyan yadda ake fassara nufin mai amfani don fassara shi zuwa motsi na yatsun hannu a hannun wucin gadi. Multicepter perceptron shine cibiyar sadarwa na wucin gadi na wucin gadi wanda ke amfani da jakar baya. MLP hanya ce mai zurfi ta ilmantarwa inda bayanai ke ci gaba a cikin hanya ɗaya, a cikin sake zagayowar ko madauki ta hanyar hanyar sadarwa ta wucin gadi.

An horar da algorithm ta bayanan shigar da bayanai daga mai amfani da ke yin jerin ƙungiyoyin hannu. Don saurin haɗuwa da sauri, an yi amfani da hanyar Levenberg - Marquardt don dacewa da ma'aunin cibiyar sadarwa maimakon zuriyar gradient. Tsarin horo na cikakken samfurin ya kasance da sauri kuma ya ɗauki ƙasa da mintuna 10 ga kowane batutuwa, yana mai da algorithm mai amfani daga hangen amfani na asibiti.

Katie Zhuang a EPFL Translational Neural Engineering Lab, wanda shine marubucin farko na binciken bincike . “Abin da muke yi shi ne mu sanya waɗannan firikwensin a kan kututturen da suka rage, sannan mu yi rikodin su kuma mu yi ƙoƙarin fassara abin da siginar motsi take. Saboda waɗannan siginar na iya zama ɗan ƙaramin hayaniya, abin da muke buƙata shine wannan injin koyo algorithm wanda ke fitar da aiki mai ma'ana daga waɗannan tsokoki kuma yana fassara su cikin motsi. Kuma waɗannan ƙungiyoyi sune ke sarrafa kowane yatsa na hannun robotic. ”

Saboda tsinkayar injin na motsi na yatsan na iya zama bai kai kashi 100 daidai ba, masu binciken EPFL sun haɗa aikin sarrafa robotic don ba da damar hannun wucin gadi kuma don fara rufewa kusa da wani abu da zarar an fara tuntuɓar farko. Idan mai amfani yana son sakin abu, duk abin da zai yi shine ƙoƙarin buɗe hannun don kashe mai sarrafa robotic, da mayar da mai amfani cikin ikon hannun.

A cewar Aude Billard wanda ke jagorantar Algorithms da Laboratory Systems na EPFL, hannun na robotic yana iya amsawa a cikin millise seconds 400. Billard ya ce "Sanye take da firikwensin matsin lamba a duk yatsun hannu, yana iya amsawa da daidaita abu kafin kwakwalwa ta fahimci cewa abin yana zamewa," in ji Billard.

Ta hanyar amfani da hankali na wucin gadi ga neuroengineering da robotics, masanan kimiyyar EPFL sun nuna sabon tsarin kula da juna tsakanin injin da niyyar mai amfani - ci gaba a fasahar neuroprosthetic.

Copyright © 2019 Cami Rosso An adana duk haƙƙoƙi.