一種新型的欠驅(qū)動(dòng)仿擬人機(jī)械手爪的設(shè)計(jì)-肌電假肢手傳動(dòng)和執(zhí)行部分含12張CAD圖
一種新型的欠驅(qū)動(dòng)仿擬人機(jī)械手爪的設(shè)計(jì)-肌電假肢手傳動(dòng)和執(zhí)行部分含12張CAD圖,一種,新型,驅(qū)動(dòng),擬人,機(jī)械,手爪,設(shè)計(jì),假肢,傳動(dòng),以及,執(zhí)行,履行,部分,部份,12,十二,cad
綜述
肌電假手的研究現(xiàn)狀
人手是人類賴以生存和勞動(dòng)的最復(fù)雜，最精細(xì)的工具。人類可以通過自己的手與外界環(huán)境接觸并從環(huán)境中取得信息。物體的紋理，溫度及人手運(yùn)動(dòng)的速度，力量幾乎同時(shí)潛意識(shí)地傳入人的大腦。人手不僅僅是絕妙的環(huán)境操作者，而且是人類交流的強(qiáng)有力的工具。人手不僅僅可以增強(qiáng)語言的含義及感情色彩，有時(shí)甚至可以代替語言，例如撫摸和擁抱。由于人手的重要性，因此手的缺失不僅僅使人喪失了重要功能而且遭受心理傷害。
假肢是人缺損肢體的代替物，用以彌補(bǔ)缺損肢體的形狀與功能。理想的假手應(yīng)該在形狀和功能上與真手一樣。它不僅僅能替代人手的感覺和運(yùn)動(dòng)功能，而且還要像人手一樣具有優(yōu)美的外形。但目前技術(shù)還遠(yuǎn)遠(yuǎn)達(dá)不到這個(gè)理想要求。
我們這里介紹的肌電假手是利用肌電作為控制信號(hào)的動(dòng)力假肢，是一種生物電控制的典型的“人—機(jī)”系統(tǒng)。其工作原理是利用殘肢者手臂上的殘端肌肉中檢測(cè)出的肌電位變化作為假手動(dòng)作的控制信號(hào)，控制假手動(dòng)作，從而代替人軀體上失去的手臂。與其他方式控制的假手比起來具有很多的優(yōu)越性，因而受到患者的青睞，擁有廣闊的市場(chǎng)，也成為上肢假肢研究中的一個(gè)熱點(diǎn)。
一、國(guó)外的研究現(xiàn)狀
1919年，Borchard等最早發(fā)明了用電能作為外來動(dòng)力驅(qū)動(dòng)的機(jī)械假手，打開電源開關(guān)，食指和中指就能推動(dòng)拇指，但由于當(dāng)時(shí)缺乏合適的控制系統(tǒng)以及未能解決可提供電源的電池，這種電子假手從未得到實(shí)際應(yīng)用。雖然有報(bào)道說在1945年Erlangen開發(fā)出一種電子假手，但這條消息未得到證實(shí)。后來一種被稱為Vaduz的假手得到應(yīng)用，并在1949年申請(qǐng)了專利。這種電子假手是應(yīng)用殘肢殘存的肌肉收縮或松弛時(shí)殘肢周徑的不同作為控制信息源的。
直到肌電控制假肢出現(xiàn)以后，電子假手的發(fā)展才受到重視。1945年德國(guó)的Reihold Reiter對(duì)肌電控制理論進(jìn)行了基礎(chǔ)研究，并發(fā)表了肌電控制假肢的實(shí)驗(yàn)研究結(jié)果。1948年Reihold Reiter研制成功世界上第一只肌電假手，這只假手被安裝在長(zhǎng)凳的一端，電子系統(tǒng)中有許多真空管，用一塊肌電控制它的張開和閉合，是第一個(gè)單點(diǎn)控制的范例。Reiter的假手在1959年的布魯塞爾世界博覽會(huì)上引起轟動(dòng)。從此引發(fā)了世界范圍的電子假肢研究熱潮。
在過去的幾十年中，科技的進(jìn)步極大影響了上肢假肢產(chǎn)業(yè)。隨著微電子技術(shù)的進(jìn)步及微型計(jì)算機(jī)的出現(xiàn)，肌電控制假手逐漸成熟并得到了廣泛應(yīng)用。經(jīng)過1957—1960年的發(fā)展，通過將電子假手內(nèi)的晶體管，電池與電子元件的分別安裝，并通過電纜連接，使假手變的十分輕便。1960年，Kobrinsk等設(shè)計(jì)的肌電假手在蘇聯(lián)第一次應(yīng)用于臨床，1965年HSchmidl在法蘭克福的聯(lián)邦骨科技術(shù)職業(yè)學(xué)校研制出第一只這正實(shí)用的肌電控制假手。二十世紀(jì)六十年代，德國(guó)OttoBock公司致力于開發(fā)一種符合機(jī)械和美容要求的電子假手，該公司于1965年生產(chǎn)出了包括三個(gè)部分的肌電假手：1 具有拇指，食指，中指三個(gè)手指，并能完成抓握動(dòng)作的機(jī)械地盤；2 應(yīng)用軟塑料做成的具有手形狀的內(nèi)手；3 具有真手外觀的美容手套。1978年Herberts等報(bào)道了多功能假手的研究成果，利用EMG幅度識(shí)別和截肢者的幻覺，用表面電極檢出肌電信息，控制三個(gè)自由度電子假手完成指伸，指屈，旋前，旋后，腕伸，腕屈六個(gè)動(dòng)作，準(zhǔn)確率為百分之五十七。80年代Denning等用新的方法識(shí)別肌電信息，控制三個(gè)自由度假肢準(zhǔn)確率提高到百分之七十二。1992年，由德國(guó)Otto Bock公司和奧地利Viennatone助聽器公司合作生產(chǎn)第一代經(jīng)橈肌電系統(tǒng)可以在市場(chǎng)上買到。從此以后，電子元件的質(zhì)量和可用的電子元件的類型逐漸增加。目前，肌電控制假手是殘肢患者應(yīng)用最為廣泛的假手，假手的尺寸可以滿足從嬰兒到成人。不同尺寸的電子肘關(guān)節(jié)，腕關(guān)節(jié)旋轉(zhuǎn)器也可以買到。隨著電池技術(shù)的發(fā)展，肌電假手變得越來越輕，電池可以使用一整天而不需充電。
二、國(guó)內(nèi)的研究現(xiàn)狀
我國(guó)從60年代初開始探討應(yīng)用肌電控制假手。60年代中期研制出單自由度肌電控制前臂假手，1979年研制成功三自由度肌電控制前臂假肢。據(jù)不完全統(tǒng)計(jì)，我國(guó)殘肢者正在使用的各類肌電控制假手已達(dá)數(shù)萬只。事實(shí)上，我國(guó)的假手方面的研究與國(guó)外相比還存在較大的距離，從事這一領(lǐng)域研究和開發(fā)的大學(xué)和研究機(jī)構(gòu)相對(duì)較少，相關(guān)產(chǎn)業(yè)比較落后。各假手生產(chǎn)廠家及康復(fù)中心目前的產(chǎn)品主要是裝飾性假手和機(jī)械牽引式假手。因此開發(fā)和研制國(guó)產(chǎn)的智能假手有廣泛的應(yīng)用前景和社會(huì)效益。我國(guó)一些大學(xué)和研究所積極地從事假手方面的研究工作。胡天培等人采用再造“指”作為控制信號(hào)源，實(shí)現(xiàn)了電子假手的準(zhǔn)確控制。
目前，已經(jīng)開發(fā)出來的商品化的性能比較完善的假手仍然是單自由度、開環(huán)控制系統(tǒng)的肌電假手。盡管科研人員對(duì)假手作了很多研究，但是大部分成果實(shí)際上仍然處于實(shí)驗(yàn)室階段，距離商品化、實(shí)用化還有很大的距離。
三、肌電假手的發(fā)展趨勢(shì)
隨著康復(fù)工程、機(jī)械、電子學(xué)的發(fā)展，肌電假手的研究也逐漸深入，向著更加智能化發(fā)展。上面介紹的傳統(tǒng)肌電假手采用開環(huán)控制，最大弊端是缺少感覺功能，始終以恒力、恒速來進(jìn)行物體抓取。由于控制系統(tǒng)比較簡(jiǎn)單，對(duì)于剛性物體的抓舉還是實(shí)用的。但對(duì)于易變形（如薄膜工件、柔軟材料等）、易碎物體（如雞蛋、玻璃制品等）就可能造成變形。破碎或者對(duì)表面造成損傷。還有環(huán)境的變化，如一個(gè)外力突然作用在被握物體上，被握物體可能脫落。智能假手技術(shù)是用工程方法模擬人的感覺和智能行為，利用傳感器獲取外部信息，形成閉環(huán)控制，其關(guān)鍵技術(shù)在于傳感器和控制系統(tǒng)。利用安裝在假手手指前端的觸、滑覺傳感器，來對(duì)握物狀態(tài)進(jìn)行實(shí)時(shí)識(shí)別。假手的握力控制可采用模糊控制技術(shù)，使假手能根據(jù)物體的滑動(dòng)速度隨時(shí)自適應(yīng)握力調(diào)整，保證以最小握力實(shí)現(xiàn)無滑移抓取控制。
假手握物過程擬人性好，響應(yīng)快，握物穩(wěn)定，可靠。但帶有觸滑覺肌電假手是一種新的假肢模式，其完善和產(chǎn)品化還需要更多的學(xué)者和工程技術(shù)人員的不斷努力。
參考文獻(xiàn)
1 鄭修軍主編. 肌電假手的研究現(xiàn)狀. 中國(guó)康復(fù)醫(yī)學(xué)雜志. 2003
2 周勝軍，白智鵬主編. 淺析肌電假肢人造假肢控制原理和結(jié)構(gòu). 1997
3 蔡立羽主編. 肌電信號(hào)分析方法的研究及進(jìn)展. 中國(guó)醫(yī)療器械雜志. 1999年23卷4期
4 雷敏主編. 肌電假肢控制中的表面肌電信號(hào)的研究進(jìn)展與展望. 中國(guó)醫(yī)療器械雜志. 2001年25卷第3期
肌電假手的調(diào)研報(bào)告
肌電控制假手是利用肌電作為控制信號(hào)的動(dòng)力假肢，是一種生物電控制的典型的“人—機(jī)”系統(tǒng)。其工作原理是利用殘肢者手臂上的殘端肌肉中檢測(cè)出的肌電位變化作為假手動(dòng)作的控制信號(hào)，控制假手動(dòng)作，從而代替人軀體上失去的手臂。與其它方式控制的假手比起來具有很多的優(yōu)越性，因而受到患者的青睞，擁有廣闊的市場(chǎng)，也成為上肢假肢研究中的一個(gè)熱點(diǎn)。
調(diào)查過程中，發(fā)現(xiàn)市場(chǎng)上除了有我們現(xiàn)在介紹的肌電控制假手之外，還有其它很多種類的假手：開關(guān)控制（壓控式）電動(dòng)假肢；索控式機(jī)械假肢；骨骼式裝飾假肢；混合式上臂假肢；肘離斷混合假肢等等。從產(chǎn)品的結(jié)構(gòu)上，絕大部分的假手都應(yīng)具備：假手的機(jī)械結(jié)構(gòu)、傳感器、驅(qū)動(dòng)器和控制系統(tǒng)，只有各個(gè)組成部分有序的工作，才能讓假手完成相應(yīng)的動(dòng)作。下面就此次市場(chǎng)調(diào)查，對(duì)假手的這些組成部分進(jìn)行簡(jiǎn)要的分析。
1、假手的機(jī)械設(shè)計(jì)
市場(chǎng)上的假手有三指的，四指的，也有五指的，為了能更好的模仿真人手的功能，五指假手的研究的比較多。五指包括四個(gè)手指和一個(gè)拇指，形狀如人手。一般地，手里裝有3個(gè)電機(jī)。其中兩個(gè)電機(jī)分別驅(qū)動(dòng)拇指和食指，另外3個(gè)手指是聯(lián)動(dòng)的，由一個(gè)電機(jī)驅(qū)動(dòng)。
手指可采用模塊化設(shè)計(jì)，4個(gè)手指具有相同的結(jié)構(gòu)。食指有3個(gè)可活動(dòng)的關(guān)節(jié)，分別是基關(guān)節(jié)、中關(guān)節(jié)和遠(yuǎn)關(guān)節(jié)。電機(jī)通過減速器驅(qū)動(dòng)基關(guān)節(jié)轉(zhuǎn)動(dòng)，并通過欠驅(qū)動(dòng)連桿將驅(qū)動(dòng)力矩傳遞給中關(guān)節(jié)。中關(guān)節(jié)和遠(yuǎn)關(guān)節(jié)通過連桿機(jī)構(gòu)實(shí)現(xiàn)耦合運(yùn)動(dòng)，沒有電機(jī)驅(qū)動(dòng)，是被動(dòng)運(yùn)動(dòng)的。中關(guān)節(jié)安裝了扭簧。各個(gè)關(guān)節(jié)處的機(jī)械限位限制關(guān)節(jié)的運(yùn)動(dòng)范圍。中指、無名指和小指的結(jié)構(gòu)和食指相同，只是驅(qū)動(dòng)部分不同。這3個(gè)手指由一個(gè)電機(jī)驅(qū)動(dòng)，它們的基關(guān)節(jié)通過彈簧支架安裝在同一個(gè)驅(qū)動(dòng)軸上。
拇指的基關(guān)節(jié)運(yùn)動(dòng)是兩個(gè)自由度，即側(cè)擺與張合。為了減少機(jī)構(gòu)的復(fù)雜性和重量，假手的拇指可設(shè)計(jì)為只有一個(gè)自由度。使用球鉸鏈的傳動(dòng)機(jī)構(gòu)，使拇指在從初始位置到最終閉合位置的運(yùn)動(dòng)軌跡形成錐面的形狀，與拇指自然的運(yùn)動(dòng)類似。
4個(gè)手指與拇指相對(duì)運(yùn)動(dòng)可以實(shí)現(xiàn)強(qiáng)力抓握，如果指尖首先與物體接觸可以實(shí)現(xiàn)捏取。整個(gè)手有3個(gè)自由度，食指和拇指可以獨(dú)立運(yùn)動(dòng)，其余3個(gè)手指共用一個(gè)自由度。自由度的增加大大提高了假手的靈活性。
2、傳感器
傳感器用于提供手指位置、力矩等信息。在每個(gè)手指的基關(guān)節(jié)和指尖上分別安裝了基于應(yīng)變片的一維力矩傳感器。拇指、食指和三指聯(lián)動(dòng)的基關(guān)節(jié)上安裝位置傳感器。位置傳感器基于霍爾效應(yīng)，用來測(cè)量基關(guān)節(jié)的轉(zhuǎn)動(dòng)角度?；魻杺鞲衅鞯闹饕獌?yōu)點(diǎn)是體積小，而且它與測(cè)量部分是非接觸的，避免了摩擦產(chǎn)生。
目前較為先進(jìn)的還有觸覺傳感器與滑覺傳感器聯(lián)合使用。觸覺傳感器主要用來檢測(cè)物體與手指是否接觸，以通知滑覺傳感器開始工作或停止工作。觸覺傳感器的種類是多種多樣的，有點(diǎn)式、棒式、緩沖器式、平板式、環(huán)式。常用的滑覺傳感器是轉(zhuǎn)輪式滑覺傳感器。
3、驅(qū)動(dòng)器
驅(qū)動(dòng)器的性能對(duì)于人手的設(shè)計(jì)非常重要，手的重量，功耗，抓取力的大小，噪聲等很多指標(biāo)都與驅(qū)動(dòng)器有關(guān)。市面上采用的驅(qū)動(dòng)器也有很多種，PORTECAP的17N78—216E直流電機(jī)就是其中的一種，該電機(jī)轉(zhuǎn)速可達(dá)8500r/min,連續(xù)轉(zhuǎn)矩5.69mN·m，堵轉(zhuǎn)力矩12.5mN·m。電機(jī)與假手上的行星齒輪減速器安裝在一起，后端安裝磁碼盤。查閱相關(guān)的資料，得知，這樣的電機(jī)能滿足手指的抓握要求。
4、控制系統(tǒng)
傳統(tǒng)的多自由度假手控制系統(tǒng)一般采用單片機(jī)或計(jì)算機(jī)作為主控制器，其功耗和體積較大，處理速度較慢，接口功能有限，不適合于假手的嵌入式、低功耗應(yīng)用。另一方面，假手的控制器需要處理的數(shù)據(jù)量大，包括肌電信號(hào)的采集處理、傳感器數(shù)據(jù)采集、濾波、電機(jī)控制等任務(wù)。
假手的控制器應(yīng)具有自動(dòng)控制的功能，這樣可以減少使用者控制假手的難度，為了提高數(shù)據(jù)處理能力，控制器采用了雙處理器結(jié)構(gòu)，兩個(gè)DSP分別在兩塊電路板上。一塊電路板是假手運(yùn)動(dòng)控制板，用于假手本體的控制，一塊是EMG信號(hào)處理板，用于肌電信號(hào)的識(shí)別。假手本體的控制包括手指的位置控制和力控制以及各個(gè)手指的協(xié)調(diào)動(dòng)作控制。假手運(yùn)動(dòng)控制板上包含了微處理器、電機(jī)驅(qū)動(dòng)器。微處理器采用了TI的TMS320F2810.這也是最常用的一種微處理器。EMG信號(hào)處理板的功能是對(duì)肌電信號(hào)進(jìn)行采集、預(yù)處理和實(shí)現(xiàn)模式分類算法，識(shí)別后將控制命令通過接口發(fā)送給假手運(yùn)動(dòng)控制板，控制假手的動(dòng)作。
以上就是對(duì)假手各個(gè)部分的介紹和簡(jiǎn)要分析。
肌電控制假手作為一個(gè)真正的“人—機(jī)”系統(tǒng)，其信號(hào)提取簡(jiǎn)單，控制準(zhǔn)確率高，具有廣闊的應(yīng)用前景和市場(chǎng)。但目前這樣的假手仍然處于研究、完善階段，市場(chǎng)上出現(xiàn)的成品還不能滿足殘肢者的要求，主要體現(xiàn)在結(jié)構(gòu)簡(jiǎn)單笨拙，自由度少，外形難看，控制系統(tǒng)簡(jiǎn)單等，因此，要真正研制生產(chǎn)出與人手一樣功能的肌電假手還需要技術(shù)人員的不斷努力。
Design and Experiments on a Novel
Biomechatronic Hand
Abstract : An “ideal” upper limb prosthesis should be perceived as part of the natural body by the amputee and should replicate sensory-motor capabilities of the amputated limb. However, such an ideal “cybernetic” prosthesis is still far from reality: current prosthetic hands are simple grippers with one or two degrees of freedom, which barely restore the capability of the thumb-index pinch. This paper describes the design and fabrication of a novel prosthetic hand based on a “biomechatronic” and cybernetic approach. Our approach is aimed at providing “natural” sensory-motor co-ordination to the amputee, by integrating biomimetic mechanisms, sensors, actuators and control, and by interfacing the hand with the peripheral nervous system.
1.Introduction
The development of an upper limb prosthesis that can be felt as a part of the body by the amputee [1], and that can substitute the amputated limb by closely replicating its sensory-motor capabilities [2], is far to become reality. In fact, current commercial prosthetic hands are unable to provide enough grasping functionality and to provide sensory-motor information to the user. One of the main problems of the current available devices is the lack of degrees of freedom (DOFs).
Commercially available prosthetic devices, such as Otto Bock SensorHand?, as well as multifunctional hand designs [3,4,5,6,7,8,9] are far from providing the manipulation capabilities of the human hand [10]. This is due to many different reasons. For example, in prosthetic hands active bending is restricted to two or three joints, which are actuated by a single motor drive acting simultaneously on the metacarpo-phalangeal (MP) joints of the thumb, of the index and of the middle finger, while other joints can bend only passively.
The way to overcome all these problems is to develop a “cybernetic” prosthesis following a biomechatronic approach, i.e. by designing a mechatronic system inspired by the biological world. A cybernetic prosthesis must solve the following problems of the commercial prostheses：
1. the reduced grasping capabilities;
2. the noncosmetic appearance;
3. the lack of sensory information given to the amputee;
4. the lack of a “natural” command interface.
The first and the second problems can be solved by increasing the number of active and passive DOFs; this can be achieved by embedding a higher number of actuators in the hand structure and designing coupled joints.
The third and forth problems can be addressed by developing a “natural” interface between the Peripheral Nervous System (PNS) and the artificial device (i.e., a “natural” Neural Interface (NI)) to record and stimulate the PNS in a selective way. This can be useful in order to make possible the ENG-based control of the prosthesis (to solve the forth problem) and to give back some sensory feedback to the amputee by stimulating in an appropriate way his/her afferent nerves solving the third problem of the commercially-available hand prostheses.
Current research activities at Scuola Superiore Sant’Anna aimed at the development of a biomechatronic prosthetic hand controlled through a “natural” NI are presented in this paper. In particular, preliminary results obtained in processing ENG signals from afferent nerves are illustrated and analyzed.
2. Design of the biomechatronic hand
The main requirements to be considered since the very beginning of a prosthetic hand design are the following: cosmetics, controllability, noiselessness, lightness and low energy consumption. These requirements can be fulfilled by implementing an integrated design approach aimed at embedding different functions (mechanisms, actuation, sensing and control) within a housing closely replicating the shape, size and appearance of the human hand. This approach can be synthesized by the term: “biomechatronic” design [11].
2.1. Architecture of the biomechatronic hand
The biomechatronic hand will be equipped with three actuator systems to provide a tripod grasping: two identical finger actuator systems and one thumb actuator system (see Figure 1)
Figure 1.Architecture of the biomechatronic hand.
The finger actuator system is based on two micro-actuators, which drives the MP and the PIP joints respectively; for cosmetic reasons, both actuators are fully integrated in the hand structure: the first in the palm and the second within the proximal phalange. The DIP joint is passively driven by a four bars link connected to the PIP joint. The thumb is equipped with two active DOFs in the MP joint and one driven passive DOF in the IP joint.
The grasping task is divided in two subsequent phases in which the two different actuator systems are active:
1) reaching and shape-adapting phases;
2) grasping phase with thumb opposition.
In fact, in phase 1) the first actuator system allows the finger to adapt to the morphological characteristics of the grasped object by means of a low output torque motor. In phase two, the thumb actuator system provides a power opposition force, useful to manage critical grasps, especially in case of heavy or slippery objects.
It is important to point out that the most critical problem of the proposed configuration is related to the strength required to micro-actuators to withstand the high load applied during the grasping phase.
In order to demonstrate the feasibility of the described biomechatronic approach, we started by developing one finger (index or middle).
2.2. Design and development of the finger prototype
As outlined above, the two DOF finger prototype is designed by reproducing, as closely as possible, the size and kinematics of a human finger. It consists of three phalanges and of palm housing, which is the part of the palm needed to house the proximal actuator (see Figure 2).
Figure 2. Finger design.
In order to match the size of a human finger, two micro-motors are incorporated, respectively, in the palm and in the proximal phalange. The actuator system is based on Smoovy~ (RMB, Eckweg, CH) micro-drivers (5 mm diameter) linear actuators based on DC brushless motors.
The output force resulting from motor activation is sufficient to move the phalanges for achieving adaptive grasp. In addition, the shell housing provides mechanical resistance of the shaft to both axial and radial loads. This turns out to be essential during grasping tasks, where loads, derived from the thumb opposition, act both on the actuator system and on the whole finger structure.
A first prototype of the finger was fabricated using the Fused Deposition Modeling [FDM] process (see Figure 3). This process allows the fabrication in a single process of three-dimensional objects, made out of acrylonitrile/butadiene/styrene [ABS] resin, directly from CAD-generated solid models.
Figure 3. Finger prototype.
2.3. Fingertip force characterization
A first set of experimental tests has been performed in order to evaluate the force that the finger is able to exert on an external object [12]. To this aim we have measured the force resulting when the finger is pressing directly on a force sensor, corresponding to different configurations of the joints.
Two “pressing” tasks were identified in order to evaluate separately and independently force obtained by the two actuators incorporated in the finger:
TASK 1: the pushing action was exerted only by the distal actuator.
TASK 2: the pushing action was exerted only by the proximal actuator.
Corresponding to each task, two subtasks were identified according to the position of the non-active joint (extended, flexed). The different values of joint rotation angles corresponding to each subtask are illustrated in Figure 4.
Figure 4. Positions of finger joints for each task.The active joint for each task and position is
indicated by a small circle.
During the force characterization the fingertip pushed on the force sensor. The Z force component was recorded, the X and Y outputs of the load cell were monitored. This was obtained by adjusting the finger position for obtaining a force parallel to the Z-axis of the load cell. A first set of experimental tests was done on the finger prototype, with the aim of evaluating how much force the finger is able to apply on an object.
2.4. Results and discussion
Ten tests were performed for each subtask. The obtained results are illustrated in Figure 5. These force values are comparable with force exerted by “natural” human finger during fine manipulation, thus demonstrating the feasibility of the biomechatronic approach, at least for this class of manipulation tasks. The output force resulting from motor activation is sufficient to move the phalanges for achieving adaptive grasp [13].
Figure 5. Experimental results.
3. Development of an intelligent Neural Interface (NI)
3.1. Microelectrodes Array Fabrication
Different microelectrode arrays on silicon substrates (named "dice") were designed and fabricated using various microfabrication technologies. Three different dice designs were fabricated with different dimensions, electrode size, and through-holes dimensions. The designs mainly utilized for in vivo experiments were the so called “Active Die 1” and “Active Die 2”. A photograph of Active Die 2 is shown in Figure 6. Some dice did not incorporate electrodes (“passive dice”) and were used for control purposes. In order to achieve mechanical robustness, each silicon die was mounted in a titanuim ring, fabricated by laser machining purposely to host the die and the electrical connections.
Figure 6
3.2. Electrophysiological Results
3.2.1. Electrophysiological Set-up and Methods
In vivo tests were performed on adult female New Zealand rabbits. For control purposes a first set of experiments was performed using first empty guidance channels and then NI based on "passive" dice (i.e. dice with through-holes but no electrodes) with no interconnects. A second set of experiments was performed using complete NI incorporating active dice. The same surgical procedure was adopted for the two sets of experiments.
Different set-up configurations were used depending on the experiment performed. For recording from the intact nerve, from nerve regenerated in empty guidance channels and from nerve through passive dice, the sciatic nerve was stimulated proximally through a pair of electrodes (FHC Hook Electrodes # 06-11-2) with constant current pulses (duration = 0.08-0.1 ms) of different intensities. Responses for both anodic and cathodic stimulations were obtained. The electroneurogram (ENG) analysis was performed by measuring the compound action potential (CAP) from the recording electrodes (placed at a distance of 50 mm from the stimulation site). In animals implanted with the NI the connector was plugged to a standard electrophysiological set-up for nerve recording and stimulation. A mechanical switch allowed each of the 10 neural interface electrodes to be connected to the electronic apparatus.
Single pulses (amplitude range = 40 μA-1 mA, duration = 0.2 ms) were delivered at each of the 10 electrodes of the active interface. When muscle contraction was observed, thus indicating the presence of functional axons, the same channel of the neural interface used for stimulation was connected to the recording apparatus and the spontaneous nerve activity was monitored. Neural activity was also recorded in response to passive leg movements (i.e., stretching).
3.2.2. Computer Analysis of Electrophysiological Signals
Signals from the amplifier were stored on a tape recorder and then played back and digitized on a computer hard disk (using a MIO-16 A/D converter). Suitable programs, based on LabView programming techniques, were used in order to acquire experimental data and synchronize the electrical stimulation. The amplitude and delay of compound action potentials were measured in order to assess the functional recovery of the regenerated nerve.
3.2.3. Stimulation and Recording from Nerve Axons using NI#6
The most interesting results were those obtained on rabbit #inter18 using NI #6. In this specific animal, stimulation and recording were performed after 48 days from implant. At this time nerve regenerated through the NI. No signs of nerve damage and/or device failure were visible. Tissue reaction seemed similar to that found in the control experiments. The flat connecting cable was still intact and it could be easily twisted in order to attach the signal conditioning plug.
The action potential duration is about 1-1.5 ms and its amplitude is about 110 μV. An intense electrical activity with respect to the resting level in response to an imposed leg movement (extension of the leg and foot) is shown in Figure 7.
Figure 7. Recording of electrical activity during a leg/foot movement (A) and after the movement (B). The control signal at rest is shown in (C).
3.2.4. Nerve Stimulation Using NI #6
Nerve stimulation was obtained through the NI #6, as demonstrated by a leg/foot contraction for each pulse delivered to the nerve through the microelectrode array. The current threshold which induced a contraction was of the order of tenth of μA, lower than the current needed to excite the nerve using the control experiment extracellular electrodes. Clear EMG signals were recorded by muscle groups as illustrated in Figure 8.
Figure 8. EMG signal recorded during nerve stimulation.
4. ENG Signal Processing Techniques
In order to verify the feasibility of extracting sensory-motor information from the ENG signals recorded from afferent nerves we carried out some preliminary experiments in collaboration with the Center for Sensory-Motor Interaction (Aalborg University, Aalborg – DK). In this Section the results of these experiments are briefly described.
4.1. Experimental Set-up
Acute experiments were performed using four femal adult New Zealand rabbits (identified by progressive numbers). The Danish Committee for Ethical use of Animals in Research approved all procedures used in the experiments. Two tripolar, whole nerve cuff electrodes were implanted around the tibial and peroneal nerves (which are major branches of the sciatic nerve) in the rabbit's left leg (cuff lengths were approx. 20 mm; the inner diameters were 2 mm and 1.8 mm, respectively). The cuff electrodes were produced according to the procedure described in [14] except that a straight cut was used as a closing method. The sural nerve was cut immediately distal to the peroneal cuff electrode to minimize the recording of unwanted cutaneous afferent activity during the experiments. In Figure 10, a schematic of the implantation sites for the cuff electrodes is presented (similar rabbit preparations have been used in other experiments, see [14]).
The equipment used during the experiments consisted of a computer controlled servomotor used to passively rotate the rabbit's ankle in the sagittal plane.
A support and fixation device equipped with four strain gauges was used as torque transducer (sensitivity 10 Nm/V). An optics-based rotation transducer was used to record the position of the ankle during the movements (sensitivity = 10°/V). The position and torque signals were sampled at 500 Hz. The rabbit was placed on its right side, and the left foot was mounted on a cradle. The knee and ankle joint were fixated during the experiment. An elaborate description of the experimental equipment can be found in [14]. The whole nerve cuff recordings were pre-amplified 200,000 times, bandpass filtered using a 2nd order Butterworth analog filter (500 Hz - 5 kHz), and sampled at fS =10 kHz (12 bit National Intruments A/D board).
The ankle angle of a normal human subject was recorded during quite standing and this signal was used as a template to move the ankle of a rabbit preparation. The whole nerve activity of the tibial and peroneal nerves were recorded as described in [14]. All ENG recordings were rectified and bin integrated during a 9 ms window. The position and torque data were low-pass filtered at fSP = 100 Hz (12th order digital Butterworth filter).
4.2. Fuzzy Models
Three fuzzy models were implemented with characteristics as follows:
1. The Modified FCRM Fuzzy Model is a Takagi-Sugeno (TS) fuzzy system. To obtain the rules directly from the data, a fuzzy clustering algorithm named fuzzy C-regression model (FCRM) is implemented [15].
2. The Adaptive Network-based Fuzzy Inference System (ANFIS) model is a TS fuzzy system implemented as a feed-forward neural network. The principal characteristic of this network is the hybrid learning procedure described in [16].
3. The Dynamic Non-Singleton Fuzzy Logic System (DNSFLS) is a Mamdani fuzzy system, implemented in the framework of recurrent neural networks [17].
4.3. Results of the prediction
To compare the performances of the different fuzzy models, the root mean square (RMS) of the prediction error has been introduced as a figure of merit. In Tables 1 the RMS of the prediction error is presented.
Table 1: RMS of the Prediction Error for the Different Fuzzy Models
Fuzzy model
Training
Traject RMS
Test
Traject RMS
Rule Number
FCRM
0.0216
0.0480
49
ANFIS
0.0047
0.0079
49
DNSFLS
0.0013
0.0057
45
5. Conclusions
This paper show research activities carried out at Scuola Superiore Sant’Anna for the development of a cybernetic prosthesis. We are currently designing of a new prototype of the biomechatronic hand and developing ENG signal processing techniques to characterize the afferent response of the PNS. Moreover, the Consortium of the GRIP EU Project (coordinated by Scuola Superiore Sant’Anna) is currently developing an implantable system to record and stimulate the PNS with a telemetry connection with an external control system. This device can be used in experiments to design the cybernetic prosthesis.
6. Acknowledgements
This work has been supported by a research project entitled “Design and development of innovative components for sensorized prosthetic systems” currently ongoing at the “Applied Research Center on Rehabilitation Engineering” funded by INAIL (National Institute for Insurance of Injured Workers), and originated by a joint initiative promoted by INAIL and by Scuola Superiore Sant’Anna. This work has been partly funded by the GRIP EU Project (“An integrated system for the neuroelectric control of grasp in disabled persons”, ESPRIT LTR #26322).
The authors are also grateful to Mr. Carlo Filippeschi and Mr. Gabriele Favati for their valuable technical assistance. The authors would also thank Mr. Rinaldo Sacchetti for helpful discussions and criticism on the biomechatronic prosthetic hand concept.
References
[1] D. C. Simpson, The Choice of Control System for multimovement prostheses: Extended Physiological Proprioception (EPP), in The Control of Upper-Extremity Prostheses and Orthoses, P. Herberts et al., Eds., 1974.
[2] J. A. Doeringer, N. Hogan, Performance of above elbow body-powered prostheses in visually guided unconstrained motion task, IEEE Trans. Rehab. Eng. 42 (1995), 621-631.
[3] P.J. Agnew, Functional effectiveness of a myoelectric prosthesis compared with a functional split hook prosthesis: a single subject experiment, Prost. & Orth. Int. 5 (1981), 92–96.
[4] S.-E. Baek, S.-H. Lee, J.-H. Chang, Design and control of a robotic finger for prosthetic hands, Proc. Int. Conf .Intelligent Robots and Systems (1999), 113-117.
[5] M. E. Cupo, S. J. Sheredos, Clinical Evaluation of a new, above elbow, body powered prosthetic arm: a final report, J. Rehab. Res. Dev. 35 (1998), 431-446.
[6] R. Doshi, C. Yeh, M. LeBlanc, The design and development of a gloveless endoskeletal prosthetic hand, J. Rehab. Res. Dev. 35 (1998), 388–395.
[7] P. J. Kyberd, O. E. Holland, P. H. Chappel, S. Smith, R. Tregidgo, P. J. Bagwell,and M. Snaith, MARCUS: a two degree of freedom hand prosthesis withhierarchical grip control, IEE
收藏