INeuroprosthetic eNtsha yindlela yokuqhekeka kweRobotics

Izazinzulu ze-EPFL (i-olecole polytechnique fédérale de Lausanne) eSwitzerland zibhengeze ukusekwa kwelizwe lokuqala lokulawula ngesandla i-robotic-uhlobo olutsha lwe-neuroprosthetic olumanyanisa ulawulo lomntu kunye ne-intelligence eyenziweyo (i-AI) yokwenza i-robot dexterity enkulu kwaye yapapasha uphando lwabo kwi NgoSeptemba 2019 kwi Ubukrelekrele boMatshini weNdalo .

I-Neuroprosthetics (i-neural prosthetics) zizixhobo ezenziweyo ezikhuthaza okanye ezonyusa inkqubo ye-nervous ngokusebenzisa amandla ombane ukubuyekeza ukusilela okuchaphazela izakhono zemoto, ukuqonda, umbono, ukuva, unxibelelwano, okanye izakhono zemvakalelo. Imizekelo ye-neuroprosthetics ibandakanya ubuchwephesha bobuchwephesha bobuchwephesha (i-BCIs), ukukhuthaza okunzulu kwengqondo, ukuvuselela intambo yomgogodla (i-SCS), ukufakelwa kolawulo lwesinyi, ukufakelwa kwe-cochlear, kunye neentliziyo zentliziyo.

Ixabiso leplani yokufakelwa kwamalungu aphezulu kulindeleke ukuba idlule kwi-2.3 yezigidigidi zeedola ngo-2025, ngokwamanani avela kwingxelo ka-Agasti ka-2019 yi-Global Market Insight. Kwi-2018, ixabiso lentengiso lehlabathi liphela lafikelela kwisigidi esinye seedola ngokusekwe kwingxelo efanayo. Kuqikelelwa ukuba kwizigidi ezibini abantu baseMelika banama-amputees, kwaye kukho ngaphezulu kwe-185,000 yokunqunyulwa kwamalungu okwenziwa minyaka le, ngokweZiko leLizwe leLifa lokuLawulwa kweLimb. Isifo seVascular sibangela iipesenti ezingama-82 zokunqunyulwa kwe-US ngokwengxelo.

I-myoelectric prosthesis isetyenziselwa ukutshintsha amalungu omzimba anqunyanyisiweyo ngelungu langaphandle elenzelwe ukusebenza kwezihlunu zomsebenzisi ezikhoyo. Ngokweqela lophando le-EPFL, izixhobo zorhwebo ezikhoyo namhlanje zinokunika abasebenzisi inqanaba eliphezulu lokuzimela, kodwa ubungqongqo abukho ndawo iphantse yafana nesandla somntu esifanelekileyo.

“Izixhobo zorhwebo zihlala zisebenzisa inkqubo yokurekhodisha amajelo amabini ukulawula inqanaba elinye lenkululeko; Oko kukuthi, ishaneli enye ye-sEMG yokujija kunye nenye yokwandiswa, ”babhala abaphandi be-EPFL kuphando lwabo. Ngelixa linomdla, le nkqubo ibonelela ngobuchule obuncinci. Abantu bayazilahla ii-myoelectric prostheses kumazinga aphezulu, ngokuyinxenye kuba bevakalelwa kukuba inqanaba lolawulo alonelanga ukuba lixabise ixabiso nokuntsonkotha kwezi zixhobo. ”

Ukujongana nengxaki yokuxakeka ngee-myoelectric prostheses, abaphandi be-EPFL bathathe indlela eyahlukeneyo yolu phononongo lobungqina bengqondo ngokudibanisa iinkalo zesayensi ye-neuroengineering, iirobhothi, kunye nobukrelekrele bokuzenzela ukuze benze inxenye yomyalelo wemoto "ekwabelwana ngawo ulawulo. ”

USilvestro Micera, uSihlalo weBertarelli weSiseko se-EPFL kwi-Neuroengineering yeNguqulelo, kunye noNjingalwazi we-Bioelectronics eScuola Superiore Sant'Anna e-Italiya, ujonga le ndlela ekwabelwana ngayo yokulawula izandla zerobhothi inokuphucula iimpembelelo zeklinikhi kunye nokusebenziseka uluhlu olubanzi lweenjongo ze-neuroprosthetic ezinje ngengqondo -Umatshini wokudibanisa (BMIs) kunye nezandla zebionic.

Abaphengululi babhala bathi: "Esinye sezizathu zokuba amalungu afakelwayo asebenzise iikhowudi ezenzelwe ukwahlula endaweni yezilinganisi ezilinganiselweyo kungenxa yokuba abahleli bahleli ngokuqinileyo." "Ukubamba, olu hlobo lolawulo lufanelekile ukuthintela ukuwa ngengozi kodwa kuncame iarhente yomsebenzisi ngokuthintela inani lezikhundla ezinokubakho. Ukuphunyezwa kwethu kolawulo ekwabelwana ngalo kuvumela zombini iarhente yomsebenzisi kunye nokubamba ukomelela. Kwindawo yasimahla, umsebenzisi unegunya elipheleleyo kulawulo lokushukuma kwezandla, ekwavumela ukwakheka kwangaphambili kwemeko yokubamba. ”

Kolu phando, abaphandi be-EPFL bagxile kuyilo lwee-algorithms zesoftware-izixhobo zerobhothi ezazibonelelwa ngamaqela angaphandle zine-Allegro Hand efakwe kwirobhothi ye-KUKA IIWA 7, inkqubo yekhamera ye-OptiTrack kunye ne-TEKSCAN sensors zoxinzelelo.

Izazinzulu ze-EPFL zenze i-kinematic proportional decoder ngokwenza i-multilayer perceptron (MLP) yokufunda indlela yokutolika injongo yomsebenzisi ukuze ayiguqulele ekuhambeni kweminwe kwisandla esenziwe ngumntu. I-multilayer perceptron yinethiwekhi yokutya eyenziwayo eyenziwayo esebenzisa ukubuyela umva. I-MLP yindlela enzulu yokufunda apho ulwazi luqhubela phambili kwicala elinye, ngokuchaseneyo nomjikelo okanye iluphu ngokusebenzisa inethiwekhi ye-neural eyenziweyo.

Ialgorithm iqeqeshiweyo ngedatha yegalelo evela kumsebenzisi esenza uthotho lweentshukumo zezandla. Ngexesha lokudibana ngokukhawuleza, indlela yeLevenberg – Marquardt yayisetyenziselwa ukulungelelanisa ubunzima bomnatha endaweni yethambeka lokuhla. Inkqubo yoqeqesho olupheleleyo yayikhawuleza kwaye yathatha imizuzu engaphantsi kwe-10 kwisifundo ngasinye, isenza ukuba i-algorithm isebenze kumbono wokusetyenziswa kweklinikhi.

"Kulowo unqunyulwe ilungu lomzimba, kunzima kakhulu ukufumana izihlunu ngeendlela ezininzi, ezahlukeneyo ukulawula zonke iindlela ezihamba ngayo iminwe yethu," utshilo uKatie Zhuang kwi-EPFL Translational Neural Engineering Lab, owayengumbhali wokuqala wesifundo sophando. . Into esiyenzayo kukubeka ezi zivamvo kwisiqobo sazo esishiyekileyo, emva koko sizirekhode sizame ukutolika ukuba yintoni intshukumo yentshukumo. Kungenxa yokuba le miqondiso inokuba yingxolo, into esiyifunayo yile algorithm yokufunda umatshini ekhupha umsebenzi onentsingiselo kwezi zihlunu kwaye azitolike zihambe. Kwaye ezi ntshukumo zilawula iminwe nganye yeerobhothi. ”

Ngenxa yokuba uqikelelo loomatshini bokuhamba kweminwe lusenokungachaneki ngeepesenti ezili-100, abaphandi be-EPFL bafaka iirobhothi ezenzekelayo ukuze zenze isandla sokufakelwa kwaye ziqale ngokuzenzekelayo ukuvala into xa sele uqhakamshelwano lokuqala. Ukuba umsebenzisi ufuna ukukhupha into, konke ekufuneka akwenzile kukuzama ukuvula isandla ukuze acime isilawuli seerobhothi, kwaye abuyisele umsebenzisi kulawulo lwesandla.

Ngokuka-Aude Billard okhokela i-EPFL's Algorithms kunye neLebhu yeLebhu, isandla seerobhothi siyakwazi ukusabela ngaphakathi kwe-400 milliseconds. Uxhotyiswe ngoxinzelelo lwenzwa kuyo yonke iminwe, inokusabela kwaye izinzise into ngaphambi kokuba ingqondo ibone ukuba into iyatyibilika, utshilo uBillard.

Ngokusebenzisa ubukrelekrele bokufakelwa kwe-neuroengineering kunye neerobhothi, izazinzulu ze-EPFL zibonise indlela entsha yolawulo ekwabelwana ngalo phakathi komatshini kunye nenjongo yomsebenzisi-inkqubela phambili kwitekhnoloji ye-neuroprosthetic.

Ilungelo lokushicilela © 2019 Cami Rosso Onke amalungelo agciniwe.