【溫馨提示】 dwg后綴的文件為CAD圖，可編輯，無水印，高清圖，，壓縮包內(nèi)文檔可直接點(diǎn)開預(yù)覽，需要原稿請(qǐng)自助充值下載，請(qǐng)見壓縮包內(nèi)的文件，所見才能所得，下載可得到【資源目錄】下的所有文件哦--有疑問可咨詢QQ:1304139763 或 414951605
中國地質(zhì)大學(xué)長(zhǎng)城學(xué)院
本科畢業(yè)設(shè)計(jì)外文資料翻譯
系 別： 工程技術(shù)系
專 業(yè)： 機(jī)械設(shè)計(jì)制造及其自動(dòng)化
姓 名： 趙少帥
學(xué) 號(hào)： 05208310
2011年 12 月 31 日
外文資料翻譯譯文
一種室內(nèi)境導(dǎo)航條件下自主智能移動(dòng)機(jī)器人系統(tǒng)的研究
摘 要
這種為室內(nèi)境導(dǎo)航條件下設(shè)計(jì)生產(chǎn)的自主移動(dòng)機(jī)器人系統(tǒng)是一個(gè)不完整的差速傳動(dòng)系統(tǒng)，它有兩個(gè)安裝在同一軸上通過兩個(gè)PID控制的電機(jī)驅(qū)動(dòng)的驅(qū)動(dòng)輪和兩個(gè)分別安裝在前部和后部的腳輪。它裝備了大量的傳感器，比如紅外探測(cè)器，超聲波傳感器，激光線發(fā)生器和數(shù)碼相機(jī)，通過這些傳感器來感知它的周圍循環(huán)境。它的計(jì)算源是由多任務(wù)多程序處理功能的多處理器組成的同步運(yùn)行系統(tǒng)。為了實(shí)現(xiàn)復(fù)雜的任務(wù)，這種移動(dòng)機(jī)器人采用了混合的控制結(jié)構(gòu)。它在智能設(shè)計(jì)中心設(shè)計(jì)完成，在香港城市大學(xué)制造調(diào)試。
關(guān)鍵字：移動(dòng)機(jī)器人；智能控制；傳感器；導(dǎo)航
1 簡(jiǎn)介
隨著自主移動(dòng)機(jī)器人在工廠和服務(wù)業(yè)中的應(yīng)用越來越廣泛。一些在設(shè)計(jì)，感覺，控制和導(dǎo)航等領(lǐng)域的投資已經(jīng)實(shí)施。對(duì)真實(shí)世界的反應(yīng)，探測(cè)循環(huán)境，沒有沖突的執(zhí)行已經(jīng)設(shè)計(jì)好的計(jì)劃以及完成想要其完成的任務(wù)是智能移動(dòng)機(jī)器人的主要要求。作為人類，我們能夠很簡(jiǎn)單的完成這些動(dòng)作，但是對(duì)于機(jī)器人來說，將是非常的困難。一個(gè)自主移動(dòng)機(jī)器人應(yīng)該運(yùn)用不同的傳感器來感知外部循環(huán)境以及轉(zhuǎn)換和認(rèn)識(shí)感知信息以在執(zhí)行任務(wù)時(shí)能夠通過適當(dāng)?shù)乃惴ㄖ贫ò踩囊苿?dòng)路徑。許多不同的傳感器被安裝在移動(dòng)機(jī)器人上用來避障，位置確定，移動(dòng)感知，導(dǎo)航和內(nèi)部控制方面，這些傳感器有射程傳感器，光傳感器，力傳感器，聲傳感器，軸編碼器，陀螺儀。許多人用紅外和超聲波傳感器搜尋在都達(dá)目的地的障礙。激光距離探測(cè)器也在移動(dòng)機(jī)器人混亂空間中的部長(zhǎng)方式中得到應(yīng)用。數(shù)碼相機(jī)在移動(dòng)機(jī)器人視覺系統(tǒng)中也得到了應(yīng)用。
盡管有很多不同種類的傳感器可以利用，但是傳感器并不意味著就是獲得。傳感器從外界循環(huán)境獲得的信息必須認(rèn)真地處理認(rèn)識(shí)才能夠被完全地利用到指導(dǎo)移動(dòng)機(jī)器人對(duì)真實(shí)世界不斷變化的動(dòng)作中。國內(nèi)或者國外，引導(dǎo)移動(dòng)機(jī)器人的方向以便讓它最終到達(dá)目的地和接近的過程中避障是對(duì)移動(dòng)機(jī)器人最基本的基本要求。一些實(shí)現(xiàn)這些任務(wù)流行的方法包括邊緣檢測(cè)，網(wǎng)格分析和潛在范圍法。由于這些方法的局限性，模糊邏輯和人工神經(jīng)網(wǎng)絡(luò)已經(jīng)被應(yīng)用到移動(dòng)機(jī)器人定位和導(dǎo)航的信息整合中，以及移動(dòng)機(jī)器人的運(yùn)動(dòng)控制中。應(yīng)用模糊邏輯我們避免了一些傳統(tǒng)人工智能控制系統(tǒng)中規(guī)劃階段與工作模式方法相關(guān)的計(jì)算瓶頸。雖然模糊控制方法是強(qiáng)大的，但是它不能適應(yīng)隨著時(shí)間改變參數(shù)的情況。這樣神經(jīng)網(wǎng)絡(luò)的學(xué)習(xí)功能被應(yīng)用定義模糊控制輸入輸出的數(shù)據(jù)框架。神經(jīng)網(wǎng)絡(luò)中潛入模糊規(guī)則和知識(shí)庫以便通過一種反向傳播算法進(jìn)行訓(xùn)練，卡爾曼濾波技術(shù)對(duì)多傳感器信息集成和本地化增量跟蹤是很有效的方法。
為了控制移動(dòng)機(jī)器人有大量的動(dòng)作，布魯克斯提出了包容結(jié)構(gòu)，它可以描述任務(wù)實(shí)現(xiàn)行為層的機(jī)械判別。這種判別是一種可以決定那個(gè)動(dòng)作應(yīng)該在大量的行為同時(shí)沖突作用時(shí)應(yīng)該進(jìn)行的方法。阿金提出的模式基礎(chǔ)反應(yīng)控制思想是將動(dòng)作分解成叫做電機(jī)模式的并發(fā)步驟。每一種模式代表一種基本行為；這樣模式的集合提供了潛在的移動(dòng)機(jī)器人控制動(dòng)作知識(shí)庫。每一個(gè)基本的動(dòng)作行為產(chǎn)生一個(gè)基于傳感器數(shù)據(jù)的速度向量，結(jié)果被傳送到機(jī)器人的控制中。朗應(yīng)用基本的機(jī)器結(jié)構(gòu)實(shí)現(xiàn)基本動(dòng)作之間的轉(zhuǎn)換，基于這種結(jié)構(gòu)的機(jī)械裝置，基本動(dòng)作就和理論上的一樣，推理和研究可以被協(xié)調(diào)完成一些復(fù)雜的任務(wù)。這篇文章介紹了我們自主移動(dòng)機(jī)器人的實(shí)現(xiàn)。以上提到的方面在這篇文章中得到了各方面的討論。
2 移動(dòng)機(jī)器人系統(tǒng)的綜述
一個(gè)自主移動(dòng)機(jī)器人基本的應(yīng)該有探測(cè)它周圍循環(huán)境，處理數(shù)據(jù)，以及判斷和移動(dòng)其自身的的能力。我們的移動(dòng)機(jī)器人集合了多種子系統(tǒng)：機(jī)械和驅(qū)動(dòng)子系統(tǒng)，感知子系統(tǒng)，計(jì)算源子系統(tǒng)制作決定子系統(tǒng)以及電源供給供給系統(tǒng)。
這種機(jī)械形狀和驅(qū)動(dòng)方式可以說是在移動(dòng)機(jī)器人的設(shè)計(jì)中尚屬首創(chuàng)。一個(gè)機(jī)器人的外觀可以決定它有多么的抗沖擊，直流伺服電機(jī)或者步進(jìn)電機(jī)是兩種常用的制動(dòng)器。機(jī)器人的外形將影響到組建的配置，美觀，甚至機(jī)器人的移動(dòng)方式。一種失敗的外形將使機(jī)器人在雜亂的屋子里更容易掉到陷阱里或者不能在路窄的地方發(fā)現(xiàn)他需要的路徑。我們選擇了八面體的外形，這讓他有矩形和圓形的雙重優(yōu)點(diǎn)，并且克服了它們的缺點(diǎn)。這種八面體外形的機(jī)器人容易制造，內(nèi)部的組件容易布局，容易通過窄的路徑，在角落和有物體的地方旋轉(zhuǎn)，并且外表看起來美觀。
這種子系統(tǒng)的概念可以實(shí)現(xiàn)從外部循環(huán)境得到多種數(shù)據(jù)的任務(wù)，包括機(jī)器人到障礙物的距離，地標(biāo)等等。紅外和超聲波距離傳感器，激光距離探測(cè)器和數(shù)碼相機(jī)被利用裝載在這種移動(dòng)機(jī)器人中以對(duì)周圍循環(huán)境產(chǎn)生感知。這些傳感器被一些同時(shí)工作可以計(jì)劃分配任務(wù)的微處理器獨(dú)立控制，而且被當(dāng)前利用的程序進(jìn)程調(diào)度。目前，紅外和超聲波距離傳感器，激光距離探測(cè)器被用來發(fā)現(xiàn)障礙物并且測(cè)量循環(huán)境中障礙物到機(jī)器人的距離，數(shù)碼相機(jī)被用作定位和導(dǎo)航。
判斷子系統(tǒng)是智能移動(dòng)機(jī)器人在識(shí)別利用感知子系統(tǒng)獲得的信息中最重要的部分。他通過一些智能控制算法獲得合理的結(jié)果，并且引導(dǎo)移動(dòng)機(jī)器人的動(dòng)作。在我們的移動(dòng)機(jī)器人系統(tǒng)中，智能的實(shí)現(xiàn)是基于行為主義和古典規(guī)劃原則。判斷系統(tǒng)由兩個(gè)層面組成：基于智慧苦的大量任務(wù)規(guī)劃和工作循環(huán)境圖活動(dòng)的控制以對(duì)應(yīng)動(dòng)態(tài)的真實(shí)世界。判斷系統(tǒng)的動(dòng)作任務(wù)被分解成機(jī)器人建立要?jiǎng)幼骷?jí)別實(shí)現(xiàn)的任務(wù)。模糊理論在一些基本的動(dòng)作中得到了應(yīng)用。動(dòng)態(tài)機(jī)的機(jī)械理念被應(yīng)用以協(xié)調(diào)不同的動(dòng)作。
因?yàn)榇罅康碾娮悠骷?，如距離傳感器，數(shù)碼相機(jī)，畫面擷取器，激光直線發(fā)生器，微處理器，直流電機(jī)和編碼器在移動(dòng)機(jī)器人中得到應(yīng)用，所以電源必須提供不同的幅值的穩(wěn)定電壓以及足夠的電能。和普通的移動(dòng)機(jī)器人電源解決方案一樣，在我們的移動(dòng)機(jī)器人中應(yīng)用了兩個(gè)密封的24V輸出電池組用來給電機(jī)驅(qū)動(dòng)組件和電子器件提供24V,15V,+-12V, +-9V, +-5V的電源。為了轉(zhuǎn)換和調(diào)節(jié)電壓，應(yīng)用DC-DC轉(zhuǎn)換器，因?yàn)樗鼈兊母咝?，低輸出紋波和噪聲，以及全電壓范圍。
三個(gè)主要的處理器是一些大程序和判斷程序可以運(yùn)行的摩托羅拉MC68040單板計(jì)算機(jī)。這種MC68040板并列運(yùn)行用VME總線共用同一內(nèi)存。三個(gè)摩托羅拉MC68HC11的處理器用作紅外和超聲波距離傳感器的低端控制器，它和主處理器通過一定的端口進(jìn)行通信。這種大量處理器系統(tǒng)被分解成一個(gè)分層的和分布式結(jié)構(gòu)以建立快速的信息采集和快速反應(yīng)。一個(gè)實(shí)時(shí)的控制的多處理器多任務(wù)處理操作系統(tǒng)在主處理器上和諧的運(yùn)行以解決多處理多任務(wù)的程序。和諧就是只有當(dāng)運(yùn)行循環(huán)境時(shí)程序被下載到交叉編譯的可執(zhí)行映像執(zhí)行。表一是移動(dòng)機(jī)器人的硬件結(jié)構(gòu)。
3 機(jī)器人的控制
對(duì)于機(jī)器人三個(gè)最基本的驅(qū)動(dòng)系統(tǒng)是輪子，軌道和肢體。輪子驅(qū)動(dòng)的機(jī)器人在機(jī)械結(jié)構(gòu)上比需要復(fù)雜硬件的肢體和軌道機(jī)器人要簡(jiǎn)單，所以我們的機(jī)器人設(shè)計(jì)成輪子驅(qū)動(dòng)的機(jī)器人。對(duì)于輪子驅(qū)動(dòng)的機(jī)器人，合理的驅(qū)動(dòng)布局和方向盤應(yīng)當(dāng)從差分，同步，三輪和汽車方式的驅(qū)動(dòng)裝置中選擇。差分驅(qū)動(dòng)應(yīng)用兩個(gè)腳輪和兩個(gè)安裝在普通軸上的的單獨(dú)驅(qū)動(dòng)輪，這樣就可以讓機(jī)器人直線運(yùn)動(dòng)，走弧線，和轉(zhuǎn)彎。在同步驅(qū)動(dòng)中所有的輪子同時(shí)旋轉(zhuǎn)，三輪驅(qū)動(dòng)包括兩個(gè)驅(qū)動(dòng)輪和一個(gè)方向輪；汽車方式驅(qū)動(dòng)前邊的兩個(gè)輪子就像汽車一樣旋轉(zhuǎn)。很明顯差分驅(qū)動(dòng)在程序和建造方面都是最容易實(shí)現(xiàn)的。然而，移動(dòng)機(jī)器人差分驅(qū)動(dòng)的一個(gè)難題就是如何保證機(jī)器人走直線，尤其是當(dāng)機(jī)器人的兩個(gè)輪子得到不同的驅(qū)動(dòng)的時(shí)候。為了得到一個(gè)理想的路線，電機(jī)的向量必須是動(dòng)態(tài)的控制。在我們的移動(dòng)機(jī)器人系統(tǒng)中，一個(gè)裝備了PID控制的伺服電機(jī)控制器被用到。電源擴(kuò)大器從每個(gè)伺服控制器上擴(kuò)大信號(hào)驅(qū)動(dòng)電機(jī)，輪子上的軸編碼器提供反饋。電機(jī)控制和電子器件的結(jié)構(gòu)框圖如表二所示。在圖表三種展示了給予PID原則的兩輪速度控制方案。
頂循環(huán)用來引導(dǎo)左電機(jī)向量到目標(biāo)位置，低循環(huán)用來引導(dǎo)右電機(jī)向量；積分循環(huán)用來保證機(jī)器人按照期望走直線以及控制機(jī)器人的方向盤。這是一個(gè)簡(jiǎn)單的積分控制，它能夠滿足一般的需求。
4 傳感子系統(tǒng)
傳感器方案是利用傳感器信息反應(yīng)周圍實(shí)時(shí)的動(dòng)態(tài)環(huán)境，和傳統(tǒng)的方案相比，他考慮了所有的環(huán)境變化因素的知識(shí)。感知子系統(tǒng)集成了視覺和臨近覺傳感器以使機(jī)器人快速反應(yīng)，它在機(jī)器人行為判斷處理和移動(dòng)處理中扮演了重要角色。
感知子系統(tǒng)的視覺范圍首次在感知系統(tǒng)的設(shè)計(jì)中被考慮到。視覺范圍應(yīng)該視野足夠廣泛已很好的了解機(jī)器人的周圍環(huán)境。多傳感器可以提供單一的傳感器很難取得的信息。大量的傳感器彼此之間形成互補(bǔ)，提供更好的工作環(huán)境信息。全方位的感官能力被裝載到我們的移動(dòng)機(jī)器人中。當(dāng)試圖利用大量的傳感器時(shí)，必須從經(jīng)濟(jì)性和精確性上考慮需要多少種類的傳感器就可以實(shí)現(xiàn)期望的移動(dòng)任務(wù)。
超聲波移動(dòng)檢測(cè)是移動(dòng)機(jī)器人非常有吸引力的檢測(cè)方式，因?yàn)樗鄬?duì)簡(jiǎn)單就可以裝備和控制，價(jià)格實(shí)惠以及能耗小。另外，高頻率可以被用來減小外部環(huán)境的干擾。一種特殊的目標(biāo)使得紅外距離系統(tǒng)有和聲納一樣的功能，即感知障礙物的有無以及到物體距離的大小。對(duì)于發(fā)現(xiàn)小的障礙物，激光探距器可以被應(yīng)用，它可以被標(biāo)向下面的地面以發(fā)現(xiàn)里機(jī)器人比較近的物體。斷定機(jī)器人自己的位置和方向是一個(gè)高難度行為實(shí)現(xiàn)需要的基本行為。為了定位，一種采用軸編碼結(jié)果的航位推算法被應(yīng)用到。這種方法可累加在位置和方向上已經(jīng)犯過的錯(cuò)誤。很多外部傳感器可以被應(yīng)用到位置和方向的確定中。數(shù)碼相機(jī)在實(shí)現(xiàn)這個(gè)功能中也得到了廣泛的應(yīng)用，因?yàn)橐粋€(gè)屋子里面的自然特征就可以當(dāng)做路標(biāo)，比如空調(diào)，電燈等。必須考慮到每一種傳感器都有其內(nèi)在的缺點(diǎn)。對(duì)于紅外波段傳感器，如果目標(biāo)和不同的物體有一個(gè)模糊的界限，顏色等，傳感器將不能準(zhǔn)確的計(jì)算出距離。這些問題中的一些可以通過裝備和升級(jí)傳感器來得到改善。串?dāng)_和反射現(xiàn)象是兩個(gè)主要的超聲波傳感器問題。超聲波傳感器的發(fā)射率，消隱間隔，發(fā)射順序以及時(shí)滯可以被重置以提高其性能。激光測(cè)距系統(tǒng)可能無法檢測(cè)到透明材料或者反光性差的材料物件。
在這項(xiàng)工作中，我們用距離和圖像傳感器作為信息的基本來源。測(cè)距傳感器包括超聲波傳感器和上面提到特征的短和長(zhǎng)距離紅外傳感器。圖像傳感器包括灰度視頻數(shù)碼攝像機(jī)和激光測(cè)距器。在我們的移動(dòng)機(jī)器人上24個(gè)超聲波傳感器被以每十五度角分配安裝成環(huán)形用來360度視野內(nèi)發(fā)現(xiàn)物體。這將是機(jī)器人在混亂的環(huán)境中完成導(dǎo)航，通過周圍環(huán)境中的物體建立聲納地圖。利用聲納地圖，我們可以發(fā)現(xiàn)最大10M最小15CM的障礙物。紅外測(cè)距傳感器采用三角測(cè)量，從一個(gè)紅外線發(fā)射器的位置發(fā)光，并用PSD測(cè)量圖像點(diǎn)的位置。由于這些裝置采用三角測(cè)量，物體的顏色，方向和環(huán)境光對(duì)敏感度而不是精度有更大的影響。由于信號(hào)的傳輸方式用光代替了聲音，我們希望用一個(gè)動(dòng)態(tài)更短的時(shí)間周期來獲取所有紅外傳感器的采集信息。一組16個(gè)短距和十六個(gè)長(zhǎng)距紅外傳感器在機(jī)器人上繞了兩圈安裝著。長(zhǎng)距傳感器的視野長(zhǎng)度從60CM到3M,短距的從10CM到80CM。一個(gè)MC68HC11控制器控制每組傳感器。主處理器的一個(gè)任務(wù)就是通過一系列接口接受數(shù)字距離數(shù)據(jù)。四個(gè)有352×288分辨率和65×50的視角的EIA B/W數(shù)碼相機(jī)，和兩個(gè)機(jī)器人頂部的激光直線發(fā)生器。在670MM處的電源輸出功率是2.6MW。機(jī)器人頂部中心的四部數(shù)碼相機(jī)中的兩部環(huán)顧房間四周以發(fā)現(xiàn)路標(biāo)用來定位，以及建立地圖。另外的兩個(gè)數(shù)碼相機(jī)都和一個(gè)激光發(fā)生器組合用作激光測(cè)距器以發(fā)現(xiàn)目標(biāo)物。激光探測(cè)器被安置在頂部平臺(tái)的邊緣用來掃描地面。
5 控制結(jié)構(gòu)和并列計(jì)算
一個(gè)可以判定策略的控制結(jié)構(gòu)被嵌入用以決定一個(gè)機(jī)器人如何整合不同的資源以實(shí)現(xiàn)一個(gè)期望的任務(wù)。目前兩種截然不同的應(yīng)用結(jié)構(gòu)的行為和功能都良好。行為結(jié)構(gòu)能夠快速的反應(yīng)快速變化的環(huán)境。功能結(jié)構(gòu)利用人工智能技術(shù)如搜索和推理功能來發(fā)現(xiàn)目標(biāo)，然而，相對(duì)來說對(duì)環(huán)境的變化反映比較遲緩。為了能夠組合兩種結(jié)構(gòu)的優(yōu)點(diǎn)，我們利用了混合結(jié)構(gòu)在我們的機(jī)器人上。它包括三個(gè)層面：功能層面，行為層面和任務(wù)層面。
功能層面由一些和大量的控制傳感器，傳感器數(shù)據(jù)譯碼器以及電機(jī)閉環(huán)反饋控制有關(guān)的模塊組成。這個(gè)層面的模塊讓機(jī)器人有一些基本功能。
感知和行動(dòng)的的映射完成是在行為和任務(wù)層。行為層是一個(gè)行為操作者，它操作從任務(wù)層接收到的行為任務(wù)；這個(gè)層面保證了對(duì)變化的外部環(huán)境的快速反應(yīng)。
任務(wù)層包括任務(wù)的制定者和全局的制定者。任務(wù)層的使命就是制定一些任務(wù)并將其實(shí)施。一些直接指導(dǎo)行為層，另外的一些首先被全局規(guī)劃者過濾然后送往執(zhí)行層。基于一個(gè)大地圖全局策劃者應(yīng)用一些經(jīng)典的算法搜索，為了提高機(jī)器人對(duì)外界環(huán)境的的反應(yīng)速度，任務(wù)策劃者和全局策劃者在主機(jī)上運(yùn)行，功能層和行為層在被植入到機(jī)器人上的處理器中運(yùn)行。
總的來講，包括感知，制定和動(dòng)作的機(jī)器人移動(dòng)閉環(huán)控制應(yīng)該花費(fèi)很少的循環(huán)時(shí)間，所以一個(gè)一位處理器為基礎(chǔ)的平行的計(jì)算裝置被植入到我們的移動(dòng)機(jī)器人中。通常我們的事件就在并行的微處理器或者多個(gè)微處理器上運(yùn)行進(jìn)行多任務(wù)多程序處理。廣為人知的多任務(wù)OS向微軟的win95和可以通過給一定的時(shí)間實(shí)現(xiàn)行為循環(huán)的多任務(wù)并行運(yùn)行的UNIX的OS。事實(shí)上，多任務(wù)機(jī)制只是模擬所有的事件同時(shí)運(yùn)行產(chǎn)生的影響。在多處理器上運(yùn)行所有的事件可以實(shí)現(xiàn)真正的并行處理。在我們的移動(dòng)機(jī)器人中，應(yīng)用的和諧OS多任務(wù)和多處理程序被裝備到微處理器中，通過VME總線共享彼此的內(nèi)存和信息。和諧可以和期望的一樣將任務(wù)分配到不同的微處理器中進(jìn)行多任務(wù)并行工作。另外，運(yùn)行MC68H040寫在C上的任務(wù)可以集合控制紅外和遙感傳感器的MC68H11 SBC上的編碼來得到距離數(shù)據(jù) 。這些SBC和MC68H040處理器保持一致。圖五是實(shí)事的一個(gè)數(shù)據(jù)結(jié)構(gòu)的實(shí)例。
一些實(shí)例，如下所述，在機(jī)器人系統(tǒng)上的避免障礙物和一些展開的數(shù)據(jù)填充已經(jīng)被證明它對(duì)工作環(huán)境和形同性能良好的實(shí)時(shí)反映。
6 總結(jié)
我們已經(jīng)描述了一種被測(cè)試用作室內(nèi)自主導(dǎo)航的移動(dòng)機(jī)器人裝置和一些相關(guān)的智能系統(tǒng)理論和技術(shù)研究。這種機(jī)器人裝備了距離傳感器，激光探測(cè)傳感器和視覺系統(tǒng)以察覺周圍環(huán)境。在移動(dòng)機(jī)器人中植入微處理器進(jìn)行數(shù)據(jù)的并行計(jì)算，用以提高它的合理性和快速性能。底層的處理和傳感器控制由低成本的微處理器完成。一個(gè)基于實(shí)時(shí)的任務(wù)操作系統(tǒng)可以支持很多不同的控制結(jié)構(gòu)，這可以讓我們用不同的方法進(jìn)行試驗(yàn)。這些實(shí)驗(yàn)證明了這種移動(dòng)機(jī)器人系統(tǒng)的有效性。這個(gè)平臺(tái)被用作試驗(yàn)和研究傳感器的數(shù)據(jù)融合，區(qū)域填充，反饋控制以及人工智能。
外文原文
The investigation of an autonomous intelligent mobile robot system for indoor environment navigation
Abstract
The autonomous mobile robotics system designed and implemented for indoor environment navigation is a nonholonomic differential drive system with two driving wheels mounted on the same axis driven by two PID controlled motors and two caster wheels mounted in the front and back respectively. It is furnished with multiple kinds of sensors such as IR detectors ,ultrasonic sensors ,laser line generators and cameras,constituting a perceiving system for exploring its surroundings. Its computation source is a simultaneously running system composed of multiprocessor with multitask and multiprocessing programming. Hybrid control architecture is employed on the rmbile robot to perform complex tasks. The mobile robot system is implemented at the Center for Intelligent Design , Automation and Manufacturfing of City University of Hong Kong.
Key words:mobile robot ; intelligent control ; sensors ; navigation
Introduction
With increasing interest in application of autonomous mobile robots in the factory and in service environments，many investigations have been done in areas such as design，sensing，control and navigation，etc. Autonomousreaction to the real wand，exploring the environment，follownng the planned path wnthout collisions and carrying out desired tasks are the main requirements of intelligent mobile robots. As humans，we can conduct these actions easily. For robots however，it is tremendously difficult. An autonomous mobile robot should make use of various sensors to sense the environment and interpret and organize the sensed information to plan a safe motion path using some appropriate algorithms while executing its tasks. Many different kinds of senors have been utilized on mobile robots，such as range sensors，light sensors，force sensors，sound sensors，shaft encoders，gyro scope s，for obstacle awidance，localizatio n，rmtion sensing，navigation and internal rmnitoring respectively. Many people use infrared and ultrasonic range sensors to detect obstacles in its reaching domain.Laser range finders are also employed in obstacle awidance behavior of mobile robots in cluttered space.Cameras are often introduced into the vision system for mobile robot navigation.
Although many kinds of sensors are available，sensing doesn’t mean perceiving. Information obtained by sensors from the environment must be carefully processed and organized in order that it can be fully utilized for the mobile robot’s reaction to the dynamically changing real world. Localizing locally or globally，orienting the direction of mobile robot in order to eventually reach the destination and awiding obstacles in the proximity normally are the fundamental elementary capabilities expected of rmbile robots. Some of the more popular methods for achieving these tasks include edge-detection，certainty grids and potential fields. Due to the shortcoming of these methodologies (for example，high burden of computation，local minima，etc.)fuzzy logic and artificial neural networks have been introduced into integrating the sensed information for mobile robot localization and navigation，and rmtion control of mobile robots. Using fuzzy logic we avid some of computational bottlenecks associated wrath the planning phase of the methods based on the model of warkspace employed in traditional artificial intelligence systems.Although the fuzzy control method is robust，it cannot adapt to changing parameters over time.Learning of neural network is then used for defining mapping between input and ouput data of fuzzy control. Surender embeds fuzzy rules and membership knowledge into a neural network for training via a back propagation algorithm. Kalman filter techniques are also efficient approaches for integrating multi-sensor information and tracking in incremental localization.
For the control of mobile robots having multiple behaviors，Brooks proposed the subsumption architecture，which describes the mechanism of arbitration between the layered task-achieving behaviors. Arbitration is a process of deciding which behavior should take precedence over other behaviors when conflicting behaviors are
triggered. Arkin presented the schema-based reactive control idea that decomposes actions into a set of multiple concurrent processes called motor schemas. Each schema represents a general behavior;a collection of such schemas provides the potential family of actions for control of a mobile robot. Fach of basic active behaviors generates a velocity vector based on sensory data，and the result is transmitted to the robot for execution. Lang used state machine architecture to realize the transition armng basic behaviors. In the mechanism based on this architecture，basic behaviors as well as reasoning，inference and search procedures can be integrated to achieve complex tasks. This paper reports the implementation of our autonomous intelligent mobile robot. The aspects menboned above are discussed in the paper，respectively.
A verview of the mobile robot system
An intelligent autonomous rmbile robot basically should have the abilities to exploring its surrounding area，processing data，make decisions as well as move itself. Our rmbile robot system is grouped into several subsysterns:mechanics and driving subsystem，sensing subsystern，computing sources subsystem，decision-making subsystem and power supplingng subsystem.
The mechanical shape and driving type are commonly first taken into consideration while implementing a rmbile robot. A robot’s shape can have a strong impact on how robust it is，and DC serve rmtors or stepOper motors are often the two choices to employ as actuators. The shape of a robot may affect its configurations of components，ae sthetics，and even the movement behaviors of the robot. An improper shape can make robot run a greater risk of being trapped in a cluttered room or of failing to find its way through a narrow space. We choose an octahedral shape that has both advantages of rectangular and circular shapes，and overcomes their drawbacks. The framework of the octahedral shaped robot is easy to make，components inside are easily arrange and can pass through narrow places and rotate wrath corners and nearby objects，and is more aesthetic in appearance.
The perception subsystem accomplishes the task of getting various data from the surroundings，including distance of the robot from obstacles，landmarks，etc.Infrared and ultrasonic range sen}rs，laser rangefinders and cameras are utilized and mounted on the rmbile robot to achieve perception of the environment. These sensors are controlled independently by some synchronously running microprocessors that are arranged wrath distributive manner，and activated by the main processor on which a supervising program runs. At present，infrared and ultranic sensors，laser rangefinders are programmed to detect obstacles and measure distance of the robot from objects in the environment，and cameras are programmed for the purpose of localization and navigation.
The decision-making subsystem is the most important part of an intelligent mobile robot that organizes and utilizes the information obtained from the perception subsystem. It obtains reasonable results by some intelligent control algorithm and guides the rmbile robot. On our mobile robotic system intelligence is realized based on behaviourism and classical planning principles. The decision-making system is composed of twa levels global task planning based on knowledge base and map of working enviro nment，reactive control to deal with the dynamic real world. Reaction tasks in the decision-making system are decomposed into classes of behaviors that the robot exhibits to accomplish the task. Fuzzy logic is used to implement some basic behaviors. A state machine mechanism is applied to coordinate different behaviors.
Because many kinds of electronic components such as range sensors，cameras，frame grabbers，laser line generators，microprocessors，DC motors，encoders，are employed on the mobile robot，a power source must supply various voltage levels which should are stable and have sufficient power. As the most common solution to power source of mobile robots，two sealed lead acid batteries in series writh 24 V output are employed in our mobile robot for the rmtor drive components and electronic components which require 24 V，15V，士12V，+9V，士5V，variously. For the conversion and regulation of the voltage，swritching DC DC converters are used because of their high efficiency，low output ripple and noise，and wride input voltage range.
Three main processors are Motorola MC68040 based single board computers on which some supervisory programs and decision-making programs run. These MC68040 boards run in parallel and share memory using a VMEbus. Three motorola MC68HC11 based controllers act as the lower level controllers of the infrared and ultranic range senors，which communicate with the main processors through serial ports. The multi-processor system is organized into a hierarchical and distributive structure to implement fast gathering of information and rapid reaction. Harmony， a multiprocessing and multitasking operating system for real-time control，runs on the main processors to implement multiprocessing and multitasking programming. Harmony is a runtime only environment and program executions are performed by downloading crosscompiled executable images into target processors. The hardware architecture of the mobile robot is shown in Fig.
Robots control
For robots，the three rmst comrmn drive systems are wheels，tracks and legs. Wheeled robots are mechanically simpler and easier to construct than legged and tracked systems that generally require more complex and heavier hardware，so our mobile robot is designed as a wheeled robot. For a wheeled robot，appropriate arrangements of driving and steering wheels should be chosen from differential，synchro，tricycle，and automotive type drive mechanisms. Differential drives use twa caster wheels and two driven wheels on a common axis driven independently，which enable the robot to move straight，in an arc and turn in place. All wheels are rotate simultaneously in the synchro drive;tricycle drive includes two driven wheels and one steering wheel;automobile type drive rotates the front twa wheels together like a car. It is obvious that differential drive is the simplest locomotion system for both programming and construction. However，a difficult problem for differentially driven robots is how to make the robot go straight，especially when the motors of the two wheels encounter different loads. To follow a desired path，the rmtor velocity must be controlled dynamically. In our mobile robot system a semv motor controller is used which implements PID control.Ibwer amplifiers that drive the motors amplify the signals from each channel of serwcontroller. Feedback is provided by shaft encoders on the wheels. The block diagram of the motor control electronic components are shown in Fig. 2，and the strategy of two wheel speed control based PID principle is illustrated in Fig.3.
Top loop is for tracking the desired left motor velocity;bottom loop for tracking right motor velocity;Integral loop ensures the robot to go straight as desired and controls the steering of the robot. This is a simple PI control that can satisfy the general requirements.
Sensing subsystem
Sensor based planning makes use of sensor information reflecting the current state of the environment，in contrast to classical planning，which assumes full knowledge of the environment prior to planning. The perceptive subsystem integrates the visual and proximity senors for the reaction of the robot. It plays an important role in the robot behavioral decision-making processes and motion control.
Field of view of perceptive subsystem is the first consideration in the design of the sensing system. Fneld of view should be wide enough with sufficient depth of field to understand well the robot’s surroundings. Multiple sensors can provide information that is difficult to extract from single sensor systems. Multiple sensors are complementary to each other，providing a better understanding of the work environment. Omnidirectional sensory capability is endowed on our mobile robot. When attempting to utilize multiple senors，it must be decided how many different kinds of sensors are to be used in order to achieve the desired motion task，both accurately and economically.
Ultrasonic range sensing is an attractive sensing rmdalityfor mobile robots because it is relatively simple to implement and process，has low cost and energy consumption. In addition，high frequencies can be used to minimize interference from the surrounding environment. A special purpose built infrared ranging system operates similar to sonar，determining the obstacle’s presence or absence and also the distance to an object. For detecting smaller obstacles a laser rangefinder can be used. It can be titled down to the ground to detect the small objects near the robot. Identifying robot self position and orientation is a basic behavior that can be part of high level complex behaviors. For localizing a dead reckoning method is adopted using the output of shaft encoders. This method can have accumulated error on the position and orientation. Many external sensors can be used for identification of position and orientation. Cameras are the most popular sensor for this purpose，because of naturally occurring features of a mom as landmarks，such as air conditioning system，fluorescent lamps，and suspended ceiling frames.
Any type of sensor has inherent disadvantages that need to be taken into consideration. For infrared range senors，if there is a sharply defined boundary on the t