关 键 词：

温室 收获 机器人 自动 导向 车车 设计 CAD 开题

资源描述：

温室收获机器人自动导向车车身设计含3张CAD图带开题,温室,收获,机器人,自动,导向,车车,设计,CAD,开题

内容简介：

XXX设计(XX)外文资料翻译院 系专业学生姓名班级学号外文出处Sensors and Actuators A 118(2005)183-189指导教师评语：指导教师签名： 年 月 日用于自动导向车的高精度超声波进出站引导系统F.Tonga,r,S.K.Tsoa,T.Z.Xub中国香港城市大学智能协会自动控制和操作系统小组设计中国厦门，361005，厦门大学海洋系2003年10月9日接受：2004年6月29日接受修订本；2004年6月29日采用；2004年9月29日联机登录摘要随着自动导向车在现代制造中物料处理方面越来越广泛的应用，其导向技术已经引起广大的关注。这篇论文的目的就是研究基于自动导向车二维的超声波随着自动导向车在现代制造中物料处理方面越来越广泛的应用，其导向技术已经引起广大的关注。这篇论文的目的就是研究基于自动导向车二维的超声波传感器。作为一种可靠的和经济的方法，超声波传感器的应用还面临着某些缺点：如精度低。现代作业中采用一种新颖的高精度、低成本的超声波进出站引导系统用于基于进出站引导工作站的AGV定位。传感器。作为一种可靠的和经济的方法，超声波传感器的应用还面临着某些缺点：如精度低。现代作业中采用一种新颖的高精度、低成本的超声波进出站引导系统用于基于进出站引导工作站的AGV定位。在准备的方案中，电磁波被作为同步触发飞行时间计数器。进出站引导系统分别由超频率音频接受装置和分别处于同一高度的烟火散发装置的工作站和自动导向车组成。自动导向车的二维位置由两个超声波发射信号和两个接收器获得。为了保证基于超声波传感器中低消耗狭窄的带宽中进出站引导的准确性，采用一种适当运算法则的超声波传感器用来实现高精度距离修正（1mm）。建立在自动导向车上样本上的实验结果证实了预测的想法，自动导向车的位随着自动导向车在现代制造系统中越来越广泛的应用，其进出站引导技术已经置精度在X和Y方向上可以达到3mm，在定位上可以达到12004 Elsevier B.V.所有权利声明保留关键词：自动导向车；进出站引导；平衡；超声波探测；1. 绪论物料处理无论在花费还是时间上都是现代加工制造工业中的一个重要组成部分。确切的说，统计表明处理只占据一项典型工作5%的时间，其他时间都被花费在存储和运输过程1,2中。由于在机动性、适应性和效率方面的优胜，自动导向车在现代制造体系中扮演着一个越来越重要的角色。这样，对自动导向车技术的研究就对自动处理系统的产业化显出重要的意义。自动导向车的主要任务就是将物料从一个工作台传送另一个工作场所。在行走的过程中，自动导向车要定位自身的二维位置。传统的方法是依靠埋在地下的电线或者是固定位置上的反射激光。 当自动导向车行走到一个工作台附近的时候，它就进入到进出站引导模式。自动导向车进出站引导中一个更大困难的问题是其对工作台的位置和相对关系两方面都要得到控制。同时，为了满足机械接触地面的需要，进出站引导程序必须采用比正常行走更高的精确性，这个问题已经被多个学者所提出。Vazetal3开发出了基于进出站引导系统的可移动机器人上的红外线传感器反射体，这种装置的精度在X和Y方向可以达到5mm,在方位上可以达到1。利用超声波传感器和CCD-线-相机，Roth的进出站引导系统的精确度在方位和距离上可以低于5mm。 超声波传感器被广泛应用于可移动机器人中。因为基于进出站引导解决方案上的超频率响的缘故，进出站引导的准确性基本上依靠点与点间的超声波距离修正的精确性。一般地，依照描述的精确性，已经有很多高精度的超声波装置。人造偏光板能在10.7m5的范围内产生1的平行排列单位。由Figueroa开发出的一对发射器-接收器传感器能在0.9m6的范围内达到0.152mm的探测精度。 虽然如此，目前几乎没有可用于评估窄带传感器精确性效果的系统。考虑到窄带传感器的特点，传统方法将会导致因测量时间的延迟而造成的精确性较低。为了解决这个问题，去卷积策略能通过排除传感器有限波段的影响而提供一个可行的解决方案7,8。但去卷积方法的不足之处在于操作设计到分割过程，这对数据误差是十分敏感的，特别是当被一个很小的数值分开时。尽管内含参数补偿能减小数值的影响，但是这种最佳参数的选择是十分困难的。在目前的研究中，由于采用了相应的均等化处理来减少上述方法的负面因素，一种高精度性低成本的超声波进出站引导方案浮出水面。考虑到和超频率音箱相比具有极高的传播速度，电磁波被用作为触发信号来测量超声波的传播时间。 剩下的内容组织如下：第一，介绍位置的描述和关系的确定方案的理论基础。接下来，形成高精度处理方法和系统的硬件及软件设计的轮廓。2. 理论基础和系统描述 作为一种已确定的技术，超声波传感器已经被广泛地用于移动机器人10-12的配置系统。这些系统可以分为两类，一类是脉冲-反射模式，另一种是发射器-接收器模式。发射器-接收器模式系统可以避免目标反射面问题，因而可靠性更高。Mahajan和Figueroa13,Martin et al 15已经研发出了用于移动机器人的发射器-接受器模式的超声波导航系统。为了能达到高度的可靠性和精确性，现行方案中采用发射器-接收器模式。整个进出站引导系统由两部分组成：安装在自动导向车上的发射装置和位于工作台上的接受装置。发射装置和接受装置安装在同一高度，这样可以将进出站引导问题化成二维模型来处理。同时，在触发测量时间的同步系统中采用电磁波。2.1位置和方位测量原理 在自动导向车的二维位置问题中采用三角测量方法方案。通过测量Dca和位于自动导向车上的发射器c和接受台A之间的Dcb以及已知位置B，自动导向车的二维位置就可以如图1所示那样被确定下来。 二维位置计算如下： (1) (2)准确来说，进出站引导高精度的位置定位和关系测量，关系测量时基于输出定位完成的，两点确定一条直线，借助安装在自动导向车上的两个发射器F,H，自动导向车是靠工作台上的接受装置D，E定位，这样我们就可以测得自动导向车离工作台的相对位置关系，相对位置关系计算如下： (3) 图1 2D位置和方位测量法则为了确保方位测量的准确性，发射台M在AGV上被首先定位在E，D两点的中信线上，其优先级在AGV方位测量之上。图2 同步时间系统2.2同步系统设计电磁波作为同步系统来控制时间计数器。同步原理介绍：在发射台发射出超声波脉冲的同时，无线发射器模块发射出一个同步电磁脉冲，接受装置在接到触发同步脉冲时计数一次，然后等待接受超声波脉冲去停止计数器工作并记录下时间，不同接受位置的时间值被用作计算AGV位置和方位关系时的参数。3 高精度超声波研究 3.1精确度考虑 上节所述的位置和关系测量原理意味着AGV的导航精度已经达到了超声波探测的精度。尽管如此，超声波探测在传统上应用中还是存在着精度较低的问题。普遍认为存在着许多因素如空气的扰动、湿度、温度，传感器未校准和传感器的带宽限制着超声波探测的准确性，其中，传感器的带宽是主要原因。从一个小带宽传感器获得的信号渐渐地在时间领域上从开始值到达峰值，这就导致探测精度低。根据Cramer-Rao理论，限制了尺寸的时间变化的确定下限，那么系统就有16 (4)这里T是测量时间的宽度，SNR是信号与接收信号的噪音的比率，fl和fh分别是系统带宽的较低限和较高限。经验证，和发射信号的带宽及空气媒介带宽相比较，传感器的带宽是整个系统带宽的主要限制。结果在相同的信噪比下，低带宽传感器导致了较低的探测精度。这项工作中，分析得出超声波传感器的带宽是探测精度较低的主要原因，通过廉价的，频带受限制的超声波传感器，采用同等化处理方法，可以达到探测精确性。3.2传感器同等化 传感器平衡装置是一个倒转滤波器，过去常用来平衡传感器中的小带宽从而增大整个系统的带宽。借助去卷积滤波器的帮助，传感器的影响可以通过使用倒转系统得到减小。理想地，传感器的误差和自身的滤波器将会导致延时脉冲： （5）这里c是延时常数。在去卷积处理后，整个系统的时间区分精度在传感器的波段影响后有所提高。3.3建立在适应运算法则基础上的传感器同等化 为避免第一节所提到的直接去卷积方法的缺点，人们采用适当的运算法则去调整有限脉冲响应（FIR）传感器均衡的分量。同等化在两个阶段完成：滤波器系数在反复使用一个参考信号中得到调整；第二，去卷积法通过从调整获得的固定系数中实现。 在调整阶段，为了获得较高的信噪比下得接受信号x（k），发射器和接受器传感器被固定在确定的地方，其间只有一个很小的距离（设置为0.5m）。将发射激励d（x）输入到发射器，接收器传感器接收到的信号就作为调整过程的输入，参考信号为延时激励信号。然后根据最小平均数平方法则就可以得出设备安装检查平衡装置的重量Wk.系数Wk的变换规则如下： （7） （8）这里u是一个常量，用来控制收敛率；e（k）是与系统相适应的误差：N是输入参考信号的长度；L是滤波器系数的长度；d（k）和x（k）分别是参考信号和输入信号。引用延时常量c用来确保倒转模型的因果关系。系统采用的传感器是频率为3khz，带框为3db廉价的扑通模型T/R40-16.工作在中心频率是40khz的弯曲震动模式下，它能够适应空气媒介。尽管如此，对于超声波探测系统来说，她的带宽是限制其精确性的主要因素。因此，同等化过程用来减少带宽的影响从而提高了探测的精确性。4. 实验工作4.1高精度超声波探测4.1.1传递时间的确定方法 在目前的工作中，获得的信号可以达到1Msps 8-bitADC的水平。在集中的适应性调整过程中，获得的重量被作为传感器均衡器的固定信号保存下来。图4（A）和（B）分别是同等化前后超声波信号获得的时间记录。为了容易估计时间，通过同等化过程由固定在确定位置（当前方案为0.5m）的一对传感器获得的时间记录被相交相关操作中用作标准信号。同样地，通过比较现行方法和古典互相关系的目的后，在无需同等化过程的情况下通过一对传感器就可以获得标准信号。 在取得了标准信号后，根据相应的接受信号和标准信号之间的相互关系，时间就可以通过AGV上的发射装置和工作站上的接受装置确定下来。图4（C）和（D）中配有插图部分展示了在没有相应同等化标准信号和有相应同等化标准信号时的接受信号的规格化交相互关系曲线。相互关系曲线的顶点被选定为超声波的到达时间（图4（C）和（D）中标记有黑色箭头处）。 图4 信号波形（A）（B）和自身同等化前后的规格化相反关系曲线（C）（D）相应标准信号 4.1.2校准过程和传播速度的调整传播时间的确定方法包括和实际到达时间之间的偏移量。如果信号的形成没有实质性的变化，这个偏移量可作为常量来处理，并且可以在开始时校准一次。校准过程可由安装在相距为2m的固定位置上的两个传感器测出的400个测量值得平均值实现。除传播时间之外，超声波的传播速度C的精确性也会影响输出的修正值。在现代研究中，为准确起见，温度因素也被计算在内。为了减少受温度变化影响的C造成的探测误差，现采用补偿方法。温度T（）和C（m/s）的关系可由下式得出18: (9)传播速度C和在每次测试开始时测量的温度相对应。4.1.3 同等化结果考虑到在没有同等化时的实际情况，原始信号来源于两个T/R40-16传感器，这两个传感器覆盖了时间记录（见图4（A）上的700us的范围并展示了交相关曲线上真实点附近的许多逐渐减弱的无意义点（见图（C）。另一方面，如图4（B）所示，使用传感器平衡装置后，均衡器输出信号的时间幅值聚减为约100us，和交相关曲线上容易确定的相关顶点联系在一起（见图（D）。因此，描述准确性的到达时间的增加量就可以通过均衡器采集的更少的无意义交相关顶点的影响预测出来。4.1.4 同等化探测精度的提高 在校准和调整了传播速度后，精确的曲线就可以通过分别由传统相关方法得出的从2-9m的7个不同距离的测量误差绘出。均衡器同等化过程前后探测精度的对比结果绘在图5上。两种误差都决定于距离，如Ep。（4）所协定的那样，信噪比会随着距离的增加而减少。类似地，因为传感器均衡器的使用扩大了整个系统的带宽，探测精度比在1-3mm不同距离里使用传统方法得到的精度有了明显的提高。可以得出结论，这种系统的探测精度在3m的范围内为0.5mm。可以解决AGV工作站进出站引导系统的需求精度。4.2AGV进出站引导的实验11，18 在实验室里借助自动导向车样本完成进出站引导实验，实验环境是有围墙和角落的2D平台。实验结果如下：4.2.1进出站引导实验的试验台 为了达到导航实验的目的，采用如图6所示一辆被命名为“XD-AGV1”的自动导向车样本作为导航实验平台，自动导向车装有两个驱动轮（左右轮）和一个辅助轮（前轮），每个驱动轮都由一个步进电机驱动这种设计能够使自动导向车很容易准确地调整其与工作站的位置和相对方位关系。图5 同等化前后的探测精度图6 AGV“XD-AGV1”的样本照片4.2.2 AGV二维定位 准确地说，实现方位需要在自动导向车进出站引导区域才能完成，进出站引导区域离两个工作塔台的中心大约1m的距离，在靠近这两个塔台中心线5cm内，自动导向车的进出站引导模式之前的实验定位结果如图7所示。工作站（见图1）的中信在图7上被设为远点（0,0）。这样我们就可以看到，随着自动导向车离工作台距离的增加，由于本章提到的信噪比减弱的原因，定位误差也是出现增加的趋势。设L=750mm,在没有噪音干扰的空气扰动下（基本上是由风扇引起的），器信号振幅波动程度为3.28dB，在实验台上就可以按计划方案实现准确的二维定位精度3mm，这个精度比其他商业上利用的类似自动导向车的精度要高，例如德国Blechert的H750-24型运输机器人自动导向车和中国SIA-AGV-T500/T100方位测量在自动导向车准确地定位在两个工作台的中心线上后，它就开始为最终的进出站引导操作进行方位测量。自动导向车上的超声波发射台发出超声波信号，时间久可以由接收器得出，根据式3就可以由这个时间计算出自动导向车和工作站之间的角度，如图8所示是在50个不同的角度的50个方位测量结果。图8 在五个不同方向上的50次测量结果）由图可以看出，当自动导向车被定位在与工作站相对成一个大角度时（20左右，见图8），由于两个超声波传感器未对准而引起的强烈波形变形，大角度的区分误差就比小角度（小于20，见图8）的大。 考虑到最终的方位测量精度是由进出站引导操作中小角度的估计结果决定的，这样定位精度可以达到1，这在图8中得到了证实。5绪论 在目前的工作中，研究人员已研发出一种新型的超声波二维定位系统用于自动导向车的进出站引导。传统的三角测量方法被用来计算二维位置，方位关系是由自动导向车两个超声波传感器的布置决定的。考虑到超声波传播的估计时间，结合传统超声波测量方案的缺点，需要特别注意的是二维定位的高精度问题和应用传感器同等化过程时的方位关系测量。 采用发射器-接收器模式的时间探测方法、同步电磁波和传感器同同等化过程，在自动导向车试验台上进行的进出站引导实验就可以获得满意的可靠性和精度。在自动导向车试验台上进行的进出站引导实验就可以获得满意的可靠性和精度。在自动导向车试验台上获得的最终进出站引导精度指数在二维方位上为3mm，在方位测量上为1。 所描述的进出站引导方案具有成本低，动力充足，精度高的特点，在向着和其他类似的机动交通工具的实际进出站引导系统方向发展有着很大的潜力。下一步的工作就是要研究两个传感器未对准时测量误差的影响，以便采用新方法来获得较高的探测精度。参考文献1F.liu,CIMS and Manufactory Automation,Publishing House of engineering industry,Beijing,1998.2X.Zhang.J.Feng,Y,Wang,Modern manufactory material handling and simulation,Publishing house of Tsinghua University,Beijing,1998.3P.Mira Vaz,R.Ferreira,V.Grossmann,M.I.Ribeiro,in:Proceedings of ISIE97,Guimaraes,Portugal,Docking of a mobile platform based on infrared sensors 2(1997)735-740.4H.rOTH,K.Schilling,Navigation and Docking Manoeuvres of Mobile Robots in Industrial Environments.4(1998)735-7405C.Biber,S.Ellin,E.Shenk,J.Stempeck,The polariod ultrasonic ranging system,Audio Engineering Society Preprint A-8(1991)6J.F.Figueroa,E.Doussis,A hardware-level method to improve to improve the range and accuracy of an ultrasonic ranging system,ACUSTICA 78(1993)226-232.7J.Salazar,A.turo,J.A.Orega,Deconvolution problem to produce ultrasonic short pulses,IEEE Proceeding of Science and Measurement Technology 145(1998)317-320.8S.K.Sin,C.H.Chen,A comparison of decinvolution technique for the ultrasonic nondestructive evaluation of materials,ISEE Transaction on Image Processing 11(1992)3-10.9J.Salazar,A.Turo,J.A.Ortega,Deconvolution problem to produce ultrasonic short pulses, IEEE Proceeding of Science and Measurement Technology 145(1998)317-310.10D.Zhong,A trianggulation mode ultrasonic angle and range measurement device,Measurement and Control Technogogy 4(1999)39-4111F.Tong,T.Z.Xu,in:Proceedings of the CCR2000,ChANGSHA An ultrasonic navigation sensor for mobile robot (2000) 292-295.12J.F.Figueroa,An Ultrasonic Ranging System for Robot End-Effector Positon Measurement,Ph.D.Thesis,Mechanical Engineering Department,The Pennsylvania State University,1988.13A.Mahajan,F.Figueroa,An automatic self-installation and calibration method for a 3D position sensing system using ultrasonics,Rob.Autonomous Sys.28(1999)281-294.14G.Tesch,in：Proceedings of theAUV 2000,State College,Penneylvania,Acoutic-based position discrimination of a moving robot(2000).June 28-29.15M.Martin Abreu,R,Ceres,L.Calderon,Measuring the 3D-position of a walking vehicle using ultrasonic and electomagnetic wave,Sens.Actuators 75(1999)131-138.16G.F.Brian,Time-delay estimation techniques applies to the acoustic detection of jet aircraft transit,J.Acousit.Soc.Am.7(106)(1996)255-264.17B.Farhang-Boroujeny,Adaptive Filters:Theory and Applications,John Wiley and Sons,Chichester,1999.18F.Tong,Research of ultrasonic navigation system for AGV,Ph.D.Thesis,Oceanograhy Department of Xiamen university,2000.A high precision ultrasonic docking system used for automatic guided vehicleF.Tonga,r,S.K.Tsoa,T.Z.XubConsortium for Intelligent Design,Automation and Machatronics(CIDAM),City University of Hong Kong, Hong Kong SAR,ChinaOceanography Department of Xiamen,361005,ChinaReceived 9 October 2003;received in revise form 29 June 2004;accepted 29 June 2004 Available online 29 September 2004Abstract With the more and more wide applications of automatic guided vehicle (AGV) in the material handling of modern manufactory system, its docking techniques have drawn extensive attention. The objective of this paper is to investigate the ultrasonic sensor based AGV 2D docking. As an established and cost-effective method, the application of classic ultrasonic sensor still faces certain drawbacks, e.g., low accuracy. In the present work, a novel high-precision, low-cost ultrasonic docking system is developed to locate the AGV relative to the docking workstation. In the proposed scheme, with electromagnetic wave used as system synchronization to trigger the time of flight (TOF) counter, the docking system consists of the ultrasound reception and emission beacons positioned at the same height of docking workstation and AGV, respectively. The 2D position of AGV is obtained from the TOF of ultrasonic signal from the emitter to two receivers. To ensure docking accuracy on low-cost narrow bandwidth ultrasonic transducer, a transducer equalizer based on the adaptive training algorithm is employed to realize high ranging precision (1 mm). Experimental results obtained on a demo AGV are presented and discussed, confirming the validity of the proposed solution, with an AGV docking precision of around 3 mm in x and y, 1 in bearing.Keywords AGV;Docking;Equalization; Ultrasonic rangingAbstract With the more and more wide applications of automatic guided vehicle (AGV) in the material handling of modern manufactory system, its docking techniques have drawn extensive attention. The objective of this paper is to investigate the ultrasonic sensor based AGV 2D docking. As an established and cost-effective method, the application of classic ultrasonic sensor still faces certain drawbacks, e.g., low accuracy. In the present work, a novel high-precision, low-cost ultrasonic docking system is developed to locate the AGV relative to the docking workstation. In the proposed scheme, with electromagnetic wave used as system synchronization to trigger the time of flight (TOF) counter, the docking system consists of the ultrasound reception and emission beacons positioned at the same height of docking workstation and AGV, respectively. The 2D position of AGV is obtained from the TOF of ultrasonic signal from the emitter to two receivers. To ensure docking accuracy on low-cost narrow bandwidth ultrasonic transducer, a transducer equalizer based on the adaptive training algorithm is employed to realize high ranging precision (1 mm). Experimental results obtained on a demo AGV are presented and discussed, confirming the validity of the proposed solution, with an AGV docking precision of around 3mm in x and y, 1 in bearing. 2004 Elsevier B.V. All rights reserved. Keywords: AGV; Docking; Equalization; Ultrasonic ranging1.Introduction Material handling is a significant part of the manufacturing process, both in terms of cost and time. Indeed, statistics show that the processing time only occupies 5% of the manufacturing time of a typical job, the remainder is spent in storage and transportation processes 1,2. With advantages such as mobility, flexibility, efficiency and so on, the automatic guided vehicles (AGV) are playing a more and more important role in modern manufactory system. Thus the research of AGV technology makes important sense for the industrialization of automatic handling system. The main task of AGV is to transport materials from one workbench to another in industrial environment. During thestage of running, AGV need to locate its 2D position, traditionally depending on electrical wires buried in the floor, or reflecting laser from fixed position. When AGV runs near to a workbench, it enters docking mode. A more difficult problem raised upon the docking of AGV is that both the position and the bearing relative to the workbench must be controlled. Meanwhile, to meet the need of mechanical interface, the precision needed in docking procedure is much higher than that in simply running. This topic has been extensivelyaddressed by several researchers. Vaz et al. 3 developed an infrared sensor and reflector based docking system for mobile robot with an accuracy of 5mm in x and y,1 in bearing. Using ultrasonic sensors and CCDline-camera, Roths docking system 4 claims accuracy in orientation and distance of less than 5 mm. Ultrasonic sensors are widely used in mobile robot applications. For ultrasound based docking solution, the ac-curacy of docking is in essence dependent on the precision of pointpoint ultrasonic ranging. Generally characterized as limited precision, there have been several ultrasonic devices that possess high accuracy. Polaroid claims a ranging unit with a resolution of 1% for a range of 10.7m 5. A transmitterreceiver pair developed by Figueroa yields an accuracy of 0.152mm for a range of 0.9m 6.Nonetheless, few previous systems addressed the effect of low-bandwidth transducer on ranging precision. Consider the low-bandwidth characteristics of the transducer, the classical correlation method will lead to poor accuracy of time delay measurement. To cope with this problem, the deconvolution strategy can offer a feasible countermeasure by removing the influence of band limited transducer 7,8. But the main drawback of the direct deconvolution method is that the operation involves a division process, which is very sensitive to data errors, especially when divided by a small value. Though the inclusion of a compensator parameter can alleviate the effect of small value, unfortunately, optimal selection of this parameter proved to be difficult 9. In the present study, with the adoption of an adaptive equalization processing to mitigate the negative factors above-mentioned, the design of a highprecision, low-cost ultrasonic docking scheme is presented.Consider its great propagation speed compared to ultrasound, the electromagnetic wave is adopted as the trigger signal for measuring TOF of ultrasonic wave. The remainder of the present paper is organized as follows.First, the theoretical basis of the described position and bearing estimation scheme is introduced. Next, the high precision processing method, system hardware and software design are outlined. Finally, experimental studies are carried out with the use of a demo AGV platform in the laboratory. The results obtained demonstrate the satisfactory performance of proposed scheme. Corresponding author. Tel.: +852 2784 4608; fax: +852 2788 8423.E-mail address: .hk (F. Tong).0924-4247/$ see front matter ?2004 Elsevier B.V. All rights reserved.doi:10.1016/j.sna.2004.06.0262. Theoretical basis and system description As an established technique, ultrasonic sensor systems have been widely used in positioning system of mobile robot 1012. These systems can be divided into two categories, one is pulseecho mode, and the other is transmitterreceiver mode. Systems of transmitterreceiver mode can avoid the problem of identifying the target reflecting surface and get better reliability. Mahajan and Figueroa 13, Tesch 14, Martin et al. 15 have developed transmitterreceiver mode ultrasonic navigation systems for mobile robot. To reach high reliability and precision, transmitterreceiver mode is employed in the proposed scheme. The whole docking system consists of two parts: the transmitter beacons mounted on AGV and the receiver beacons at workbench. Both the emission and reception beacons are positioned at the same height to simplify the docking into a problem of 2D position discrimination. Meanwhile, the electromagnetic wave is used as system synchronization to trigger the TOF measuring.Fig. 1. Principle of 2D localization and bearing measurement2.1.Principle of position and bearing measurement The triangulation scheme is adopted to estimate the 2D position of AGV conveniently. By the measurement of ranges DCA and DCB between emitter C on AGV and receiver beacons A and B on known position, AGVs 2D location can be determined as shown in Fig. 1a. The 2D location calculation is as follows: (1) (2) Accurately docking needs both high precision localization and bearing measurement. Bearing measurement is performed based on the localization output. Because two points can define a line, with two transmitters F, H mounted on the AGV (shown in Fig. 1b) located by receiver beacons D, E at workstation respectively, we can measure the AGVs bearingrelative to the workstation. The bearing calculation is as follows: (3) To assure the accuracy of bearing measurement, transmitter beacon M is firstly located to the centerline of E, D points prior to the AGV begins the bearing measurement. Fig. 2. Synchronization timing of system.2.2. System synchronization design The electromagnetic wave is used as system synchronization to trig the TOF counter. Synchronization principle (shown in Fig. 2) is that: at the same time of transmitter beacon emitting an ultrasonic pulse, wireless transmitter module emits an electromagnetic synchronization pulse, reception side triggers a counter once receiving the synchronization pulse, then waits the receiving of ultrasonic pulse to stop the counter and record the TOF. The TOF values from different reception positions (See TOF1 and TOF2 in Fig. 2) are used as the parameters for the calculation of AGV position or bearing in Eqs. (13).3. Research of high precision ultrasonic ranging3.1. Accuracy consideration The principle of localization and bearing measurement in the above section means that the precision of AGV navigation is up to the precision of ultrasonic ranging. However, ultrasonic ranging method is traditionally characterized as low accuracy. It is generally understood that there are several factors such as air turbulence, humidity, temperature dependence, transducer misalignment and bandwidth of transducer that limit the accuracy of ultrasonic ranging. Among them, the bandwidth of transducer is a main reason. The signal obtained from transducer with a small bandwidth climb slowly from its beginning to its peak in time-domain, which causes a low ranging precision. According to the CramerRao theory, the lower bound of time estimation variance of band limited measurement system is 16: (4) where T is width of measurement time, SNR is the signal to noise ratio of received signal, fL and fH is lower and higher bound of bandwidth of system, respectively. It is recognized that,compared to the bandwidth of emission signal and air medium, bandwidth of the transducer is the main limitation of the whole system bandwidth. As a result, under the same SNR, low bandwidth transducer leads to low ranging accuracy. In this work, base on the analysis that the small bandwidth of ultrasonic transducer is the main reason of low ranging precision, an equalization processing method is employed to reach high ranging accuracy with low-cost,band limited ultrasonic transducer.3.2. Transducer equalization A transducer equalizer is an inverse filter used to counterpoise the small bandwidth of transducer to widen the system bandwidth. With the aid of a deconvolution filter, the effect of transducer can be mitigated by compensating its impulse response hTR, using an inverse system, hde. Ideally, the cascade of the transducer and its inverse filter will simply be a delayed impulse:Hde*hTR =(c) (5) where c is a constant delay. After the deconvolution process, the time discrimination precision of the whole system will be enhanced with the removing of transducers band limited effect 9.3.3. Transducer equalization based on adaptive algorithm To avoid the drawback of direct deconvolution method mentioned in Section 1, adaptive algorithm is used to train the weights of a finite impulse response (FIR) transducer equalizer. The equalization is performed in two stages: first, the filter coefficients are trained iteratively to convergence using a reference signal; second, deconvolution is carried out with the fixed coefficients obtained from training. In the training stage, transmitter and receiver transducers are positioned at fixed positions, with a very small distance (set at 0.5 m), in order to get receiving signal x(k) with high SNR. Feeding emitting impulse d(x) into transmitter transducer, received signal from receiver transducer acts as the input of training process. Reference signal is the delayed impulse signal (Fig. 3). Then the weightsWk of the FIR equalizer are trained according to the least mean square (LMS) algo-rithm. The iteration formula of coefficients Wk updating is17: k=1,2,3,.k (6) (7) (8)Fig. 3. Training process of the equalizer with LMS adaptive algorithm where is a gain constant used to control the convergence rate, e(k) is the error of adaptive system, N is the length of input and reference signal, L is the length of filter coefficients, d(k) and x(k) is reference signal and input signal respectively. The constant delay c is introduced to ensure the causality of inverse modeling. See Fig. 3. The transducer adopted in this system is the low-cost, general purpose Model T/R40-16, with a 3dB bandwidth of 3 kHz.Working on bending vibration mode with a centre frequency of 40 kHz, it is suitable for being used in air medium. However, for ultrasonic ranging system, its low bandwidth is the main factor limiting the precision. Therefore, equalization process is performed to mitigate its bandwidth to improve the ranging accuracy.4. Experimental work4.1. High precision ultrasonic ranging4.1.1. Determination method of TOF In the present work, the signals are acquired with a 1Msps 8-bit ADC. After the convergence of adaptive training process, the weights obtained are saved as the fixed taps of the transducer equalizer. Shown in Fig. 4(A) and (B) are the time histories of ultrasonic signal obtained before and after the equalization respectively. To facilitate the TOF estimation, time history obtained from transducer pairs positioned at a fixed distance (0.5m at the current scheme) through equal-ization process is firstly employed as the standard signal for cross-correlation operation. Similarly, for the aim of comparison between the proposed method and the classical correlation,standard signal is also acquired directly from the transducer pairs without equalization. After the acquirement of standard signal, the TOF between the emitter beacon on AGV and receptor beacon on workstation is determined according to the cross-correlation peak between corresponding received signal and standard signal. Illustrated in Fig. 4(C) and (D), normalized crosscorrelation curves between reception signals without and with equalization and corresponding standard signals are presented. The points tpeak associated with the peak of crosscorrelation curve are chosen as the arrival time of ultrasonic wave (marked with black arrows in Fig. 4(C) and (D).4.1.2. Calibration process and propagation speed adjustment This TOF estimation method may include an offset between the tpeak and the real time of arrival. Provided that the form of the signal does not suffer significant variations, this offset can be considered as constant and can be calibrated once at the beginning. The calibration process is carried out by the mean result of 400 measurements with transducer pairs at a fixed distance of 2.0 m. Besides the TOF, the accuracy of the propagation speed C of ultrasonic wave will also affect the ranging output. For simplification, only the factor of the temperature is accounted in the present study. To alleviate the ranging error caused by the C due to the variation of temperature, compensation measure is taken. According to the following relationship of the temperature T (C) and C (m/s) 18:C = 331.45 + 0.61 T (9)the propagation speed C is thus adjusted with the temperaturemeasured at the beginning of each test.Fig. 4. Signal waveform (A) (B) and its normalized cross-correlation curve (with corresponding standard signal) (C)(D) before and after the equalization.4.1.3. Results of equalization Consider the original situation without the equalization, the rawsignal from the T/R40-16 transducer pairs covers a big 700s time-width in time history (See Fig. 4(A) and exhibits many slowly attenuating false peaks around the actual peak in cross-correlation curve (See Fig. 4(C). On the other hand, as Fig. 4(B) illustrates, with the use of transducer equalizer, the time-width of the equalizers output signal decreases greatly to about 100s, associated with a easily identified correlation peak in cross-correlation curve (Fig. 4(D). Accordingly, the enhancement of arrival time discrimination accuracy can be straightforwardly predicted with the much less influence of false cross-correlation peaks resulting from the adoption of equalizer.4.1.4. Improvement of ranging precision with equalization After calibration and propagation speed adjustment, the precision curves are obtained by the smoothing of measurement errors at 7 different distances from 2 to 9m using proposed and classical correlation method respectively. The comparison results between the ranging accuracy before and after the transducer equalization process is plotted in Fig. 5. It is noted that both of the errors depend mainly on the distance, exhibiting general agreement with Eq. (4) as the SNR willdrop with the increasing of distance. Similarly, because the use of transducer equalizer broadens the bandwidth of whole system, ranging precision is obviously improved compared to classical method by about 13mm at different distance. Fig. 5. Ranging precision before and after the equalization. It can be concluded that the ranging precision of this system is 0.5mm within 3m, which can meet the precision requirement of AGV workstation docking.4.2. Experiment of AGV docking 11,18Docking experiments are carried out in the lab with the help of a demo AGV. The experimental environment is the laboratory room, which is assume to have a 2-D floor plane and composed of walls and corners. The results are as follow:4.2.1Test bed for docking experiments For the purpose of navigation experiments, a demo AGV named “XD-AGV1” shown in Fig. 6 is developed as nav-igation experiment platform. It is equipped with two driving wheels (left and right wheel) and a passive wheel (frontwheel). Each driving wheel is activated by a step motor. Such design facilitated the AGV to adjust its position and bearing relative to workstation accurately. Fig. 6. Photo of demo AGV “XD-AGV1”.4.2.2. AGV 2D localization Accurately localization experiment is carried out after the AGV enters a docking area, which is about 1m away from the midpoint of two workstation beacons, about 5 cm near the centerline of these two beacons. The experimental localization results before AGVs docking mode are shown in Fig. 7. The mid-point O of the workstation (See Fig. 1) is stationed at the (0,0) point in Fig. 7. It thus can be seen that with the increasing distance of AGV with respect to the reception beacons, the error of localization also exhibits an increasing trend, due to the weakening ofSNRmentioned at the previous section.With the setting of L = 750 mm,under the condition of both silent and fluctuation air medium (generated by electric fan, with signal amplitude fluctuation degree 3.28dB), an accurate 2D localization precision not more than 3mm is achieved on the test bed AGV with the proposed scheme. This precision is superior to that of some similar commercially available AGV systems, e.g., Transrobot-H750-24 type AGV of Germany Blechert and SIA-AGV-T500/T1000 of China . Bearing measurementAfter AGV is accurately located to the centerline of two workstation beacons, it begins the bearing measurement for the final docking operation. Two ultrasonic beacons on AGV transmit ultrasonic signals, the TOFs of which are estimated in receptors and used to calculate the angle between AGV and the workstation according to Eq. (3). Shown in Fig. 8 are the 50 measurement results of bearing at five different angles.As seen from Fig. 8, when the AGV is located at a relatively big angles (around 20, See Fig. 8) with respect to the workstation, the angle discrimination errors are biggerthat that from small angles (less than 20, See Fig. 8), due to the strong waveform deformation caused by the misalignment between ultrasonic transducer pairs. Consider the final bearing measurement precision will be dependent on the estimation results at small angles in docking operation, aprecision of 1 is achieved, as confirmed in Fig. 8.Fig. 7. Result of AGV 2D localization.Fig. 8. Result of 50 measurements at five different bearings.4.2.4. Index of AGV docking performance 2D position location precision: 3mm Bearing measurement precision: 1.5. Conclusion In the present work, an innovative ultrasonic 2D localization system has been developed to be used for the docking of AGV. The classical triangulation scheme is employed to calculate the 2D position and the bearing is determined by the positioning results of two points on the AGV. Consider the TOF estimation of ultrasonic wave, in light of the drawbacks of traditional ultrasonic solution, special attention is given to the high precision in 2D localization and bearing measurement by adopting a transducer equalization process. Adopting the transmitterreceiver mode TOF ranging method, electromagnetic wave synchronization and transducer equalization process, satisfactory reliability and precision are achieved in the docking experiments on a test bed AGV. The final docking precision index obtained on the test bed AGV is 3mm in 2D localization, 1 in bearing measurement. The described docking scheme, which is low-cost, robust and has high-precision, has the potential of being developed into a practical docking system for the AGVs as well as other similar mobile vehicles. Further work will be carried out to investigate the effect of misalignment of transducer pairs on measurement errors for the development of new method with higher precision.6.Acknowledgments The authors are grateful for the funding of the China Nationa. High-tech Project (863 project) (No. 863-512-10-21) in support of the present work.References1 F. Liu, CIMS and Manufactory Automation, Publishing House of engineering industry, Beijing, 1998.2 X. Zhang, J. Feng, Y. Wang, Modern manufactory material handling and simulation, Publishing house of Tsinghua University, Beijing, 1998.3 P. Mira Vaz, R. Ferreira, V. Grossmann, M.I. Ribeiro, in: Proceedings of ISIE97, Guimaraes, Portugal, Docking of a mobile platform based on infrared sensors 2 (1997) 735740.4 H. Roth, K. Schilling, Navigation and Docking Manoeuvres of Mobile Robots in Industrial Environments. 4 (1998) 735740.5 C. Biber, S. Ellin, E. Shenk, J. Stempeck, The polariod ultrasonic ranging system, Audio Engineering Society Preprint A-8 (1991).6 J.F. Figueroa, E. Doussis, A hardware-level method