幫助臥床不起者日常生活的移動(dòng)機(jī)器人系統(tǒng)外文文獻(xiàn)翻譯、中英文翻譯

幫助臥床不起者日常生活的移動(dòng)機(jī)器人系統(tǒng)外文文獻(xiàn)翻譯、中英文翻譯,幫助,匡助,臥床不起,日常生活,移動(dòng),挪動(dòng),機(jī)器人,系統(tǒng),外文,文獻(xiàn),翻譯,中英文 MOBILE ROBOT SYSTEM TO AID THE DAILY LIFE FOR BEDRIDDEN PERSONSTakashi KOMEDA, Hiroaki MATSUOKA, Yasuhiro OGAWA, Mitsuyoshi FUJII, Tateki UCHIDA, Masao MIYAGI, Hiroyuki KOYAMA and Hiroyasu FUNAKUBOSHIBAURA INSTITUTE OF TECHNOLOGY FUKASAKU307, OMIYA, 330, JAPANABSTRACT People who care for bedridden patients have great burdens in mind and body. The number of old-aged people in the population is increasing in Japan, the problem of care for bedridden patients will become an increasingly social problem. We are trying to solve this problem by developing a small mobile robot system for bedridden patients. The purpose of this system is to pick up and to bring a small object putting it somewhere inside the room semi-automatically. This mobile robot consists of a manipulator, visual sensor unit and mobile unit. This system gives information about the surrounding to the patient through the camera and monitor. When the patient directs an target object on the monitor, the system measures a 3-dimensional location of it, and the mobile unit approaches the object and the manipulator picks up and carries it back to the patient. This system is controlled by the image information and also by human through an interface based on images. 1. INTRODUCTION We generally have great burdens in mind and body to help bedridden patients. The population of old aged increasing today, this helping work will become a social problem. We are trying to solve this problem by developing a small mobile robot systeml. The purpose of this system is to bring the target object putting it somewhere inside the room semiautomatic ally. Of course, there are many kinds of works involved in helping bedridden persons, for example changing clothes, giving the medicine and cleaning the bed, etc With this system. however, we think that even if the machine achieves only one type of work instead of a human, the burdens on the helper should decrease in mind and body.2. SYSTEM CONSTRUCTIONFig. 1 shows the overview of the mobile robot. This mobile robot consists of a manipulator, a visual sensor unit and a mobile unit. This system gives information about the surrounding to the patient through the camera and the monitor. When the patient directs a target object on the monitor. the system measures a 3-dimensional location of it. The mobile unit then approaches the target object and the manipulator picks up and carrying it back to the patient. The si7e of the mobile unit is width 600 mm, length 800 mm. height 550 mm, and it has the space to load the controller and batteries. The mobile unit is driven by two DC servo motors and two wheels. Each motor has a pulse encoder for position control. The length of the manipulator is about 1255mm and it has 4 degrees of freedom and a hand. Its weight capacity is about 3 kgf. The actuator of each joint is DC servo motor with a pulse encoder to control its position. A pair of strain gages is put on to the wrist part of the manipulator. The weight of a target object can therefore be measured and used in the feedback loop of the controller.3.VISUAL SENSORThe visual sensor system consists of a CCD camera and an image processor which is controlled by a personal computer. The camera is attached on the wrist part of the manipulator. This design makes it possible to move the camera and manipulator simultaneously. Then, the system gives 0-7803-3280-6/96/%5.00 1996 IEEE - 2789 - surrounding information to the operator through the camera and the monitor. If a target object is directed, it measure a 3- dimensional location of it. The target objects for this mobile robot system arc things for daily using which have different size and shape. It is difficult to treat an image processing to recognize such things and “me its location in real time. To solve this problem, we put things for daily using on a tray which has a cylindrical fixed size grip. This grip is the target object of the visual sensor. Therefore, we can easily measure the size of the target object on the image plane as the number of pixels of the CCD camera, and calculate the distance to the object using a simple formula(Fig.2). The treatment to measure a 3- dimensional distance using the image and a following visual tracking of the target object can be done in real time.Monitor screen No.2 is the image of after O.Ssec of monitor screen No.1. Monitor screen No.1 shows the center and feature points of the target object. Monitor screen No.2 shows that the visual sensor finds new center and feature points by scanning the image. After O.Ssec, the visual sensor will find another new center and feature points.4. VISUAL TRACKINGThe mobile unit is controlled by a personal computer, a control board and motor drivers. The computer calculates the number of drive pulses of each drive motor from the location of the target object, and sends the information to the control board. The control board generates DDA pulse distribution to do the synchronous operation between two motors2. Motor drivers observe the feedback pulse to do the rate control and the position control. However, even if we perfectly control two motors of the mobile unit using the feedback pulse, the mobile unit docs not always move exactly because of the difference of the diameter of each drive wheel and the road surface condition. Therefore, we use the visual tracking method to solve this problem. This mobile robot has an image information from CCD camera, and this information renews every 0.5 second. When the mobile unit approaches the target object, the computer calculate the position of the target object on the monitor every 0.5 second. If the position shifts from the center of the monitor, the mobile unit is controlled to change the direction to catch it in the center. This method is able to keep the object in the center of the monitor continuously(Fig.3). The method was confirmed by measure the relation between the center of the target object on the monitor and the angle of mobile robot. Fig.4 shows one of the results when the mobile robot has approached about 1800 mm. This result shows that the center of the target object positioned about 360 pixel point (full pixel is 512) when the mobile robot starts action to approach to the target, and is caught the center of the monitor (256 pixel point) when the mobile robot run about 750 mm. The method is able to keep the center within 5 pixels during 750 mm to 1800 mm running and enough precision for this mobile robot.5. CONTROLLER AND INTERFACEOur mobile robot can bring the target from the undecided position inside the room semi-automatically. However, if the operator wants a specific target, he must communicate with the robot through the computer. For example, if the target is at the right side of the robot, he must point out the arrow on the monitor by the mouse cursor to turn the robot to find the target on the monitor. In addition, he must teach the robot which is the target. Furthermore, this robot is used by the operator who is not professional about robots, and is handicapped.We have therefore developed an interface system between the robot and the operator based on following conceptions. (1) Easy and simple. The operator can control the robot easily like as radio-controlled car and also this input method must be simple.(2) Safety High space is not necessary, and for the softy tic robot can interrupt the movement every time tic robot or operator recognizes a dangerous motion. This robot has the small camera which is attached on the wrist part of manipulator, the operator can recognize the direction of the end cofactor and the sconce through the monitor. But, it is not sufficient to control tic robot with surrounding information through the camera, the operator also necks information about status of the robot. For example if the position of the robot reaches the upper limit of its working space, it is impossible to move the robot even if the operator directs to move UP.Under the configuration of Multi CPU, each task compensate robot motion, and the operator can choose cash task as an object according to his necessity. With this system. the operator can simply choose the robot motion task as an edict from existing several tasks (Fig. 7). These tasks are not always in active, hut always on standby to get the command from the operator. This means that if the operator finds it necessary, he can interrupt the motion to execute another tasks which can escape from a dangerous path.It is also necessary to compensate the safety when the robot finds a danger while the movement, this means not the operator but the robot can recognize the danger. In our system, a pair of strain gages is placed on the wrist part of the manipulator, the robot can measure the load at the wrist part using these strain gages, and rccogni7c the danger when excessiveness force arise at this part. Using the time slice method, we can interrupt the robot motion without influence when robot recognize a danger. This means that this system watches a danger both operator and robot.6. CONCLUSIONS(1)We have developed the mobile robot system which consists of a manipulator, a mobile unit and a visual sensor, and is controlled by image information. (2)We have developed a visual tracking method to correct the motion direction, it is simple way, high response and good accuracy to use this system.(3)We have developed the interface system between our mobile robot and the operator. This interface can observe each task of robot motions and show the graphical motion using the real status information. We tried to check functions of this system, it was simple and easy to operate the robot and also we confirmed that it is useful for safety, In future, we will improve this system more interactive and try to apply to the obstacle avoidance.ACKNOWLEDGMENT We wish to thank the Welfare Equipment Development Center of Japan for support this project.REFERENCES 1. T.Komcda etc. “ Mobile robot system to aid the daily life of bedridden persons in the private house”, Proceedings of 2nd European Conference on Advancement Rehabilitation Technology, 24.4( 1993) 2. T.Komeda etc. I Mobile robot system to aid the daily life of bedridden persons in the private house(2nd report)”, Proceedings of 3rd European Conference on Advancement Rehabilitation Technology, pp.179-181(1995)摘 要照顧臥床不起的病人的人在思想和身體上都有很大的負(fù)擔(dān)。日本老年人口數(shù)量不斷增加，臥床不起病人的護(hù)理問(wèn)題將成為一個(gè)日益嚴(yán)重的社會(huì)問(wèn)題。為了解決這個(gè)問(wèn)題，我們正在為臥床不起的病人開(kāi)發(fā)一個(gè)小型的移動(dòng)機(jī)器人系統(tǒng)。這個(gè)系統(tǒng)的目的是把一個(gè)小物體半自動(dòng)地放到房間的某個(gè)地方。該移動(dòng)機(jī)器人由機(jī)械手、視覺(jué)傳感器單元和移動(dòng)單元組成。該系統(tǒng)通過(guò)攝像頭和監(jiān)視器向患者提供有關(guān)周?chē)h(huán)境的信息。當(dāng)患者在監(jiān)視器上引導(dǎo)目標(biāo)對(duì)象時(shí)，系統(tǒng)測(cè)量其三維位置，移動(dòng)單元靠近該對(duì)象，操作器拿起并將其帶回患者。該系統(tǒng)由圖像信息和人通過(guò)基于圖像的界面進(jìn)行控制。幫助臥床不起者日常生活的移動(dòng)機(jī)器人系統(tǒng)1 引言我們通常在精神和身體上有很大的負(fù)擔(dān)來(lái)幫助臥床不起的病人。隨著老年人口的不斷增加，這項(xiàng)幫扶工作將成為一個(gè)社會(huì)問(wèn)題。我們正試圖通過(guò)開(kāi)發(fā)一個(gè)小型移動(dòng)機(jī)器人系統(tǒng)來(lái)解決這個(gè)問(wèn)題。這個(gè)系統(tǒng)的目的是讓目標(biāo)物體半自動(dòng)地放在房間的某個(gè)地方。當(dāng)然，幫助臥床不起的人有很多種工作，例如換衣服、給藥、打掃床鋪等。有了這個(gè)系統(tǒng)。然而，我們認(rèn)為，即使機(jī)器只完成一種類(lèi)型的工作，而不是一個(gè)人，對(duì)幫助者的負(fù)擔(dān)應(yīng)該在思想和身體上減少。2 系統(tǒng)建設(shè)圖 1 顯示了移動(dòng)機(jī)器人的概述。該移動(dòng)機(jī)器人由機(jī)械手、視覺(jué)傳感器單元和移動(dòng)單元組成。該系統(tǒng)通過(guò)攝像頭和監(jiān)視器向患者提供有關(guān)周?chē)h(huán)境的信息。當(dāng)患者在監(jiān)視器上引導(dǎo)目標(biāo)對(duì)象時(shí)。該系統(tǒng)測(cè)量其三維位置。然后，移動(dòng)單元接近目標(biāo)對(duì)象，操作器拿起并將其帶回患者。移動(dòng)單元的 Si7e 寬 600 mm，長(zhǎng)800 mm。高度 550 毫米，它有空間裝載控制器和電池。移動(dòng)單元由兩個(gè)直流伺服電機(jī)和兩個(gè)輪子驅(qū)動(dòng)。每個(gè)電機(jī)都有一個(gè)用于位置控制的脈沖編碼器。操縱器的長(zhǎng)度約為 1255mm，有 4 個(gè)自由度和一只手。它的重量約為 3 公斤力。每個(gè)關(guān)節(jié)的執(zhí)行機(jī)構(gòu)是帶脈沖編碼器的直流伺服電機(jī)來(lái)控制其位置。在機(jī)械手的手腕部位安裝一對(duì)應(yīng)變計(jì)。因此，可以在控制器的反饋回路中測(cè)量和使用目標(biāo)對(duì)象的重量。3 視覺(jué)傳感器視覺(jué)傳感器系統(tǒng)由 CCD 攝像機(jī)和由個(gè)人計(jì)算機(jī)控制的圖像處理器組成。攝像機(jī)安裝在操縱器的手腕部分。這種設(shè)計(jì)使得攝像機(jī)和操縱器可以同時(shí)移動(dòng)。然后，系統(tǒng)通過(guò)攝像機(jī)和監(jiān)視器向操作員提供 0-7803-3280-6/96/%5.001996 IEEE-2789-周?chē)畔?。如果一個(gè)目標(biāo)物體是定向的，它測(cè)量它的三維位置。該移動(dòng)機(jī)器人系統(tǒng)的目標(biāo)對(duì)象為日常使用的弧形物體，具有不同的大小和形狀。很難對(duì)圖像處理進(jìn)行實(shí)時(shí)識(shí)別和定位。為了解決這個(gè)問(wèn)題，我們把日常使用的東西放在一個(gè)有圓柱形固定尺寸把手的托盤(pán)上。此夾點(diǎn)是視覺(jué)傳感器的目標(biāo)對(duì)象。因此，我們可以很容易地測(cè)量圖像平面上目標(biāo)物體的大小，作為 CCD 相機(jī)的像素?cái)?shù)，并使用一個(gè)簡(jiǎn)單的公式計(jì)算到物體的距離（圖 2） 。利用圖像進(jìn)行三維距離測(cè)量，并對(duì)目標(biāo)物體進(jìn)行跟蹤視覺(jué)跟蹤，可實(shí)現(xiàn)實(shí)時(shí)處理。2 號(hào)監(jiān)視器屏幕是 1 號(hào)監(jiān)視器屏幕的后 O.S秒圖像。1 號(hào)監(jiān)視器屏幕顯示目標(biāo)對(duì)象的中心和特征點(diǎn)。2 號(hào)監(jiān)視器屏幕顯示視覺(jué)傳感器通過(guò)掃描圖像找到新的中心和特征點(diǎn)。在 O.S秒之后，視覺(jué)傳感器將找到另一個(gè)新的中心和特征點(diǎn)。4 視覺(jué)跟蹤移動(dòng)單元由個(gè)人電腦、控制板和電機(jī)驅(qū)動(dòng)器控制。計(jì)算機(jī)根據(jù)目標(biāo)對(duì)象的位置計(jì)算每個(gè)驅(qū)動(dòng)電機(jī)的驅(qū)動(dòng)脈沖數(shù)，并將信息發(fā)送到控制板?？刂瓢瀹a(chǎn)生DDA 脈沖分配，在兩個(gè)電機(jī)之間進(jìn)行同步操作2。電機(jī)驅(qū)動(dòng)器通過(guò)觀察反饋脈沖進(jìn)行速度控制和位置控制。然而，即使我們使用反饋脈沖完美地控制移動(dòng)單元的兩個(gè)電機(jī)，由于每個(gè)驅(qū)動(dòng)輪的直徑和路面狀況的差異，移動(dòng)單元也不總是準(zhǔn)確地移動(dòng)。因此，我們使用視覺(jué)跟蹤方法來(lái)解決這個(gè)問(wèn)題。這個(gè)移動(dòng)機(jī)器人有一個(gè)來(lái)自 CCD 攝像機(jī)的圖像信息，每 0.5 秒更新一次。當(dāng)移動(dòng)單元接近目標(biāo)對(duì)象時(shí)，計(jì)算機(jī)每 0.5 秒計(jì)算一次目標(biāo)對(duì)象在監(jiān)視器上的位置。如果位置從監(jiān)視器的中心移動(dòng)，移動(dòng)單元將被控制以改變方向以在中心捕獲它。這種方法能夠使物體連續(xù)地保持在監(jiān)視器的中心（圖 3） 。通過(guò)測(cè)量監(jiān)視器上目標(biāo)物中心與移動(dòng)機(jī)器人角度之間的關(guān)系，確定了該方法。圖 4 顯示了移動(dòng)機(jī)器人接近 1800 mm 時(shí)的一個(gè)結(jié)果。結(jié)果表明，當(dāng)移動(dòng)機(jī)器人開(kāi)始接近目標(biāo)時(shí)，目標(biāo)對(duì)象的中心位置約為 360 像素點(diǎn)（全像素為 512） ，當(dāng)移動(dòng)機(jī)器人運(yùn)行約750 mm 時(shí)，目標(biāo)對(duì)象的中心位置被捕捉到監(jiān)視器的中心位置（256 像素點(diǎn)） 。該方法能夠在 750-1800 毫米的運(yùn)行過(guò)程中使中心保持在 5 個(gè)像素范圍內(nèi)，并且對(duì)該移動(dòng)機(jī)器人具有足夠的精度。5 控制器和接口我們的移動(dòng)機(jī)器人可以半自動(dòng)地將目標(biāo)從房間內(nèi)尚未確定的位置帶出。但是，如果操作者想要一個(gè)特定的目標(biāo)，他必須通過(guò)計(jì)算機(jī)與機(jī)器人通信。例如，如果目標(biāo)位于機(jī)器人的右側(cè)，他必須用鼠標(biāo)光標(biāo)指向監(jiān)視器上的箭頭，以轉(zhuǎn)動(dòng)機(jī)器人在監(jiān)視器上找到目標(biāo)。此外，他還必須教機(jī)器人什么是目標(biāo)。此外，該機(jī)器人還被不擅長(zhǎng)機(jī)器人的操作人員使用，并且是殘疾人。因此，我們開(kāi)發(fā)了一個(gè)基于以下概念的機(jī)器人與操作員之間的接口系統(tǒng)。(1)簡(jiǎn)單易行。操作人員可以像無(wú)線遙控車(chē)一樣方便地控制機(jī)器人，而且這種輸入方法必須簡(jiǎn)單。(2)不需要安全高空間，對(duì)于軟性 TIC 機(jī)器人，每次 TIC 機(jī)器人或操作員識(shí)別到危險(xiǎn)動(dòng)作時(shí)，都會(huì)中斷動(dòng)作。該機(jī)器人在機(jī)械手手腕部位安裝有小型攝像機(jī)，操作者通過(guò)監(jiān)視器可以識(shí)別出末端輔助因子和開(kāi)關(guān)的方向。但是，通過(guò)攝像機(jī)來(lái)控制帶有周?chē)畔⒌腡IC 機(jī)器人是不夠的，操作者也會(huì)對(duì)機(jī)器人的狀態(tài)信息進(jìn)行縮頸。例如，如果機(jī)器人的位置達(dá)到其工作空間的上限，即使操作員指示向上移動(dòng)，也不可能移動(dòng)機(jī)器人。在多 CPU 的配置下，每個(gè)任務(wù)補(bǔ)償機(jī)器人的運(yùn)動(dòng)，操作者可以根據(jù)需要選擇現(xiàn)金任務(wù)作為對(duì)象。有了這個(gè)系統(tǒng)。操作員可以簡(jiǎn)單地從現(xiàn)有的幾個(gè)任務(wù)中選擇機(jī)器人運(yùn)動(dòng)任務(wù)作為指令（圖 7） 。這些任務(wù)并不總是處于活動(dòng)狀態(tài)，而是始終處于待機(jī)狀態(tài)，以便從操作員那里獲得命令。這意味著，如果操作員發(fā)現(xiàn)有必要，他可以中斷運(yùn)動(dòng)以執(zhí)行另一個(gè)可以從危險(xiǎn)路徑逃脫的任務(wù)。當(dāng)機(jī)器人在運(yùn)動(dòng)過(guò)程中發(fā)現(xiàn)危險(xiǎn)時(shí)，也需要對(duì)安全性進(jìn)行補(bǔ)償，這意味著機(jī)器人不能識(shí)別危險(xiǎn)，而只能識(shí)別危險(xiǎn)。在我們的系統(tǒng)中，在機(jī)械手的手腕部分放置一對(duì)應(yīng)變片，機(jī)器人可以使用這些應(yīng)變片測(cè)量手腕部分的載荷，并且在這部分產(chǎn)生過(guò)大力的危險(xiǎn)。利用時(shí)間切片方法，當(dāng)機(jī)器人識(shí)別出危險(xiǎn)時(shí)，可以不受影響地中斷機(jī)器人的運(yùn)動(dòng)。這意味著這個(gè)系統(tǒng)同時(shí)監(jiān)視操作員和機(jī)器人的危險(xiǎn)。6 結(jié)論(1)開(kāi)發(fā)了由機(jī)械手、移動(dòng)單元和視覺(jué)傳感器組成的由圖像信息控制的移動(dòng)機(jī)器人系統(tǒng)。(2)我們開(kāi)發(fā)了一種視覺(jué)跟蹤方法來(lái)校正運(yùn)動(dòng)方向，該系統(tǒng)操作簡(jiǎn)單，響應(yīng)速度快，精度高。(3)我們開(kāi)發(fā)了移動(dòng)機(jī)器人與操作員之間的接口系統(tǒng)。該界面可以觀察機(jī)器人運(yùn)動(dòng)的各個(gè)任務(wù)，并利用真實(shí)的狀態(tài)信息顯示圖形運(yùn)動(dòng)。我們?cè)囍鴻z查該系統(tǒng)的功能，操作簡(jiǎn)單，操作方便，同時(shí)也證實(shí)了該系統(tǒng)的安全性，今后我們將改進(jìn)該系統(tǒng)，使其更具互動(dòng)性，并嘗試應(yīng)用于避障。感謝日本福利設(shè)備發(fā)展中心對(duì)本項(xiàng)目的支持。參考文獻(xiàn)1T.Komcda 等， “幫助臥床不起者在私人住宅中日常生活的移動(dòng)機(jī)器人系統(tǒng)” ，第二屆歐洲先進(jìn)康復(fù)技術(shù)會(huì)議論文集，24.4（1993）2T.Komeda 等， 私人住宅中臥床不起的人的日常生活輔助移動(dòng)機(jī)器人系統(tǒng) （第二次報(bào)告） ， 第三屆歐洲先進(jìn)康復(fù)技術(shù)會(huì)議論文集 ，第 179-181頁(yè)（1995 年）MOBILE ROBOT SYSTEM TO AID THE DAILY LIFE FOR BEDRIDDEN PERSONSTakashi KOMEDA, Hiroaki MATSUOKA, Yasuhiro OGAWA, Mitsuyoshi FUJII, Tateki UCHIDA, Masao MIYAGI, Hiroyuki KOYAMA and Hiroyasu FUNAKUBOSHIBAURA INSTITUTE OF TECHNOLOGY FUKASAKU307, OMIYA, 330, JAPANABSTRACT People who care for bedridden patients have great burdens in mind and body. The number of old-aged people in the population is increasing in Japan, the problem of care for bedridden patients will become an increasingly social problem. We are trying to solve this problem by developing a small mobile robot system for bedridden patients. The purpose of this system is to pick up and to bring a small object putting it somewhere inside the room semi-automatically. This mobile robot consists of a manipulator, visual sensor unit and mobile unit. This system gives information about the surrounding to the patient through the camera and monitor. When the patient directs an target object on the monitor, the system measures a 3-dimensional location of it, and the mobile unit approaches the object and the manipulator picks up and carries it back to the patient. This system is controlled by the image information and also by human through an interface based on images. 1. INTRODUCTION We generally have great burdens in mind and body to help bedridden patients. The population of old aged increasing today, this helping work will become a social problem. We are trying to solve this problem by developing a small mobile robot systeml. The purpose of this system is to bring the target object putting it somewhere inside the room semiautomatic ally. Of course, there are many kinds of works involved in helping bedridden persons, for example changing clothes, giving the medicine and cleaning the bed, etc With this system. however, we think that even if the machine achieves only one type of work instead of a human, the burdens on the helper should decrease in mind and body.2. SYSTEM CONSTRUCTIONFig. 1 shows the overview of the mobile robot. This mobile robot consists of a manipulator, a visual sensor unit and a mobile unit. This system gives information about the surrounding to the patient through the camera and the monitor. When the patient directs a target object on the monitor. the system measures a 3-dimensional location of it. The mobile unit then approaches the target object and the manipulator picks up and carrying it back to the patient. The si7e of the mobile unit is width 600 mm, length 800 mm. height 550 mm, and it has the space to load the controller and batteries. The mobile unit is driven by two DC servo motors and two wheels. Each motor has a pulse encoder for position control. The length of the manipulator is about 1255mm and it has 4 degrees of freedom and a hand. Its weight capacity is about 3 kgf. The actuator of each joint is DC servo motor with a pulse encoder to control its position. A pair of strain gages is put on to the wrist part of the manipulator. The weight of a target object can therefore be measured and used in the feedback loop of the controller.3.VISUAL SENSORThe visual sensor system consists of a CCD camera and an image processor which is controlled by a personal computer. The camera is attached on the wrist part of the manipulator. This design makes it possible to move the camera and manipulator simultaneously. Then, the system gives 0-7803-3280-6/96/%5.00 1996 IEEE - 2789 - surrounding information to the operator through the camera and the monitor. If a target object is directed, it measure a 3- dimensional location of it. The target objects for this mobile robot system arc things for daily using which have different size and shape. It is difficult to treat an image processing to recognize such things and “me its location in real time. To solve this problem, we put things for daily using on a tray which has a cylindrical fixed size grip. This grip is the target object of the visual sensor. Therefore, we can easily measure the size of the target object on the image plane as the number of pixels of the CCD camera, and calculate the distance to the object using a simple formula(Fig.2). The treatment to measure a 3- dimensional distance using the image and a following visual tracking of the target object can be done in real time.Monitor screen No.2 is the image of after O.Ssec of monitor screen No.1. Monitor screen No.1 shows the center and feature points of the target object. Monitor screen No.2 shows that the visual sensor finds new center and feature points by scanning the image. After O.Ssec, the visual sensor will find another new center and feature points.4. VISUAL TRACKINGThe mobile unit is controlled by a personal computer, a control board and motor drivers. The computer calculates the number of drive pulses of each drive motor from the location of the target object, and sends the information to the control board. The control board generates DDA pulse distribution to do the synchronous operation between two motors2. Motor drivers observe the feedback pulse to do the rate control and the position control. However, even if we perfectly control two motors of the mobile unit using the feedback pulse, the mobile unit docs not always move exactly because of the difference of the diameter of each drive wheel and the road surface condition. Therefore, we use the visual tracking method to solve this problem. This mobile robot has an image information from CCD camera, and this information renews every 0.5 second. When the mobile unit approaches the target object, the computer calculate the position of the target object on the monitor every 0.5 second. If the position shifts from the center of the monitor, the mobile unit is controlled to change the direction to catch it in the center. This method is able to keep the object in the center of the monitor continuously(Fig.3). The method was confirmed by measure the relation between the center of the target object on the monitor and the angle of mobile robot. Fig.4 shows one of the results when the mobile robot has approached about 1800 mm. This result shows that the center of the target object positioned about 360 pixel point (full pixel is 512) when the mobile robot starts action to approach to the target, and is caught the center of the monitor (256 pixel point) when the mobile robot run about 750 mm. The method is able to keep the center within 5 pixels during 750 mm to 1800 mm running and enough precision for this mobile robot.5. CONTROLLER AND INTERFACEOur mobile robot can bring the target from the undecided position inside the room semi-automatically. However, if the operator wants a specific target, he must communicate with the robot through the computer. For example, if the target is at the right side of the robot, he must point out the arrow on the monitor by the mouse cursor to turn the robot to find the target on the monitor. In addition, he must teach the robot which is the target. Furthermore, this robot is used by the operator who is not professional about robots, and is handicapped.We have therefore developed an interface system between the robot and the operator based on following conceptions. (1) Easy and simple. The operator can control the robot easily like as radio-controlled car and also this input method must be simple.(2) Safety High space is not necessary, and for the softy tic robot can interrupt the movement every time tic robot or operator recognizes a dangerous motion. This robot has the small camera which is attached on the wrist part of manipulator, the operator can recognize the direction of the end cofactor and the sconce through the monitor. But, it is not sufficient to control tic robot with surrounding information through the camera, the operator also necks information about status of the robot. For example if the position of the robot reaches the upper limit of its working space, it is impossible to move the robot even if the operator directs to move UP.Under the configuration of Multi CPU, each task compensate robot motion, and the operator can choose cash task as an object according to his necessity. With this system. the operator can simply choose the robot motion task as an edict from existing several tasks (Fig. 7). These tasks are not always in active, hut always on standby to get the command from the operator. This means that if the operator finds it necessary, he can interrupt the motion to execute another tasks which can escape from a dangerous path.It is also necessary to compensate the safety when the robot finds a danger while the movement, this means not the operator but the robot can recognize the danger. In our system, a pair of strain gages is placed on the wrist part of the manipulator, the robot can measure the load at the wrist part using these strain gages, and rccogni7c the danger when excessiveness force arise at this part. Using the time slice method, we can interrupt the robot motion without influence when robot recognize a danger. This means that this system watches a danger both operator and robot.6. CONCLUSIONS(1)We have developed the mobile robot system which consists of a manipulator, a mobile unit and a visual sensor, and is controlled by image information. (2)We have developed a visual tracking method to correct the motion direction, it is simple way, high response and good accuracy to use this system.(3)We have developed the interface system between our mobile robot and the operator. This interface can observe each task of robot motions and show the graphical motion using the real status information. We tried to check functions of this system, it was simple and easy to operate the robot and also we confirmed that it is useful for safety, In future, we will improve this system more interactive and try to apply to the obstacle avoidance.ACKNOWLEDGMENT We wish to thank the Welfare Equipment Development Center of Japan for support this project.REFERENCES 1. T.Komcda etc. “ Mobile robot system to aid the daily life of bedridden persons in the private house”, Proceedings of 2nd European Conference on Advancement Rehabilitation Technology, 24.4( 1993) 2. T.Komeda etc. I Mobile robot system to aid the daily life of bedridden persons in the private house(2nd report)”, Proceedings of 3rd European Conference on Advancement Rehabilitation Technology, pp.179-181(1995)