2022年垂直多关节型工业机器人设计方案99.docx

《2022年垂直多关节型工业机器人设计方案99.docx》由会员分享，可在线阅读，更多相关《2022年垂直多关节型工业机器人设计方案99.docx（15页珍藏版）》请在淘文阁 - 分享文档赚钱的网站上搜索。

1、精品学习资源附录Modular Visual Servo Tracking for Industrial RobotsRobert Fuelep Biro robert.birogtri.gatech.edu Georgia Institute of Technology Atlanta, GA, 30332Gary McMurrayGeorgia Tech Research Institute Atlanta, GA, 30332欢迎下载精品学习资源AbstractHarvey Lipkin Mechanical EngineeringGeorgia Institute of Technol

2、ogy Atlanta, GA, 30332欢迎下载精品学习资源Visuallyservoing a robot to track a moving work-piecehas been demonstrated in literature using specially customized equipment. In this work, a simple modular architecture is presented using off-the- shelf components with serial line interfaces. Themethod has widesprea

3、d application for existing industrial robots whose capabilities can be upgraded with-outaltering proprietary originalequipment proven tobe reliable. As proof of concept, an ADEPT robotis visually servoed to track workpieces on a conveyorbelt moving up to 400 in/min. Simulation results areshown to co

4、mpare reasonably well with experimentaldata.Advantages and limitations of the implementation are discussed including the crucial effect of delays.1 Introduction1.1 Visual servoingThe purpose of visual servoing is to track an objectofunknown prior location using only relative visualfeedback informati

5、on provided by a camera.Traditionally,the locationand orientationofan objectinan absolute framemustbe determined beforeoperation begins. Lookand move schemes analyze asingle picture o.-lineto provide the world coordinatesof the object. Visual servoing is different in that therobot trajectory depends

6、 on a continuously updatederror signal in a relative frame.1.2 Dynamic visual servoingThe goal of dynamic visual servoing is to achieveaccurate high speed robotic tracking, performing trajectory control in the image plane. As robot speedincreases, the update rate of the feedback device mustincrease.

7、 While customized high speed vision systemsexist, most commercially available cameras欢迎下载精品学习资源provide 30or 60 frames per second, resulting in an overall processing rate of 2 to 15 images per second.The dominant effects ofusing slowlyupdatingsensors are overshoot and oscillationswhichlead to limitcy

8、cles and instability. Consider a discrete closed loopsystem consisting of a feedforward integrator and azero-order hold in the feedback, simulating sensor delay. The effects of changing the delay are shown inFigure 1. The system represented by a solid line hasthe same sampling rate of 16 ms througho

9、ut whilethe dottedline system has a feedback delay of 200ms. Given a step input, the first system exhibits afirst order exponential decay response while the othershows dramatic overshoot and oscillation.1.3 Industrial implementationMost implementation issues are not documented,relying on computer si

10、mulations7 or high performance specialized equipment410 in order to validate theories. In this paper, a paradigm is presentedfor retrofitting current industrial robots using separate, PCbased vision and controls modules. Onlyrelative coordinate frames are used, eliminating theneed for absolute coord

11、inate frames. For demonstration purposes, an ADEPT ONE robot has been usedfor visual servoing.2 Previous workThe problem with applying current research to anindustrial environment can be divided into four areas,absolute coordinate frames, performance characteristics, stability, and specialized hardw

12、are.2.1 Absolute coordinate frameSome researchers56 use the look and move approach which involves transforming the features to anabsolute coordinate frame and then commanding theend-effector to move to the location. Bishop2 concentrates on improved camera models. For these systems, the error signal

13、is formed combining a world objectposition,calculated fromthe camera output,andthe end-effectorlocation,determined using a kinematicmodel applied to jointencoder data. The advantage in a relativeframe is that the camera directlyproduces an error signal.2.2 Performance characteristicsAnotherapproach

14、is toonlyperformstatic visualservoingwhere the goalis toreach a target positionwithout regard to time3. By using very low controller gains, the serious effects of latency due to timedelays and different sampling rates may be ignored.2.3 StabilityWhilemany researchers provideproofsthat thetrackingerr

15、orconverges tozero,they do not mentionthe boundedinputboundedoutput stability of thestate variables and control effort7101213.Thereis no discussion of the possibilitythat the estimatedparameters may tend towards nonrealizable valuesarising from amplifier saturation or motor torque limits.Most algori

16、thms are derived in the continuous timedomain, whereas digital implementation is strictly inthediscrete time domain. When designing in the discrete time domain, it is straightforward to accountfor time delays due to implementation,includingthesampling rate of the visionsystem, time for mathematical欢

17、迎下载精品学习资源computations,and operating system overhead. Hosoda and Asada8findthat theirgains needto be bounded, even though the continuous time proofshows stability for all gains.2.4 Specialized hardwareIt is necessary to consider the design constraints ofthe available hardware when formulating a strat

18、egyfor visual servoing. Both Corke4 and Nelson10 usea PUMA 560, but override the internal controller andimplement their own controllers. While this methodsimplifies development for a specific robot system, itdoes not provide for a general solution applicable tomost industrial robots.3 Adaptation of

19、current industrialrobotsThe most economical means of attaining visual servoing is to retain as much of the existing robot aspossible, augmenting it with commercially availableplatform independent components. The methodology should be applicable to a wide variety of industrial robots. Open architectu

20、res will allow for performance improvements in computers and image processing hardware.3.1 MethodologyOur strategy is to utilize the realtime trajectorymodification capabilities available on many robots,such as the ADEPT ONE robot through the Altercommand. The vision and trajectory control systems a

21、re located on separate personal computers, connected by serial interfaces as depicted in Figure 2.The vision computer captures an image and determines the Cartesian error vector from the center ofgravity of the object to the camera lens. This information is then sent to the controls computer which c

22、omputes avelocitycommand. Finally,the informationispassed to the ADEPTcontrollerwhichimplementsthe velocity commands. At no point is customizedhardware used.3.2 Experimental setup3.2.1 Industrial robotThe interface between the visual servoing componentsand the standard industrial robot is simply a s

23、erial communications connection at 19.2 kbps. TheADEPT MC robot controller calculates the inversekinematics, using a velocity input from the controlscomputer. Alter is a fairly standard command whoseinput is simply a differential change in end-effectorlocation. The only programming required on thero

24、bot is a host program that reads in the incrementalmovements over the serial communications line usingthe single precision IEEE floatingpoint format andplaces the movements into the Alter command whichruns as a background task.3.2.2 Controls computerThe controls computer, a 100 MHz Pentium basedpers

25、onal computer, acts as the system coordinator.Connected by serial cables to both the industrial robotand vision system, the controls computer reads inthe processed vision data, using the IEEE floatingpoint format, and calculates a velocity command. Itcan also display various diagnostic information t

26、o thescreen.Most industrial robots use older processors and interpreted languages which execute slowly. Instead欢迎下载精品学习资源ofbeing limited to the controller provided by the manufacturer, one can use any advanced control routine.The ability exists to easily test several different algorithms on the same

27、 experimental setup.At first a simple PID control system with integrator anti-wind-up was used. Later, a crude form ofpredictive control,using an exponential decay factorfor the vision input, was added to compensate for delays in the system. Recently, there has been some experimentation with adaptiv

28、e self-tuning controllers.Details of this control system can be found in 1.3.2.3 Vision computerThe current method utilizes a Pulnix TMC-74 colorcamera mounted on the robot end-effector in theeye-in-hand configuration.An Imaging Technologies ITIMLFrame Grabber and AM-CLRAcquisition Module reside in

29、a 33MHz 486 host computer. Pixels ofspecific colors are separated from the background andused to calculate a center of gravity. This allows object recognition without prior knowledge of shape, sizeor placement. The color contrast method is generallyinsensitiveto changing lightconditions,not requirin

30、gadditional lighting sources. When necessary to determine height above the object, an additional algorithmuses information from the size of the object11.Thevisionsystem presents the largestbandwidthbottleneckofthe system. Originallya Data Translation DT-2871 Frame Grabber with a DT-2858 Coprocessor

31、was used, but was only able to provide anupdate rate of 2 Hz. The current system can requireover 60 ms to acquire an image and then another 25ms for processing, a total rate of 10 Hz. As soonas one cycle is completed, the features are sent to thecontrols computer and image acquisition begins again.4

32、 Results4.1 Cycle timesThe time to process one image and transport theinformationthrough the module to the ADEPT variessignificantly,from less than 80 ms to almost 250 mswith a mode of roughly 100 ms. This has a tremendous impact on the controller gains. To compensatefor this, a system of fixed cycl

33、e times has been developed to provide consistent sampling rates and allowfor accurate modeling. Each module is forced to operate with a cycle time of 120 ms. In the event thata module has not finished processing data when thecycle is over, the last set of data is retransmitted.An unforeseen benefit

34、of this is to correct a featureinherent in implementing PID controllers which run ata faster rate than the feedback. Since integral controlisa sum of all previous error inputs, the control actionramps up as the same data is used for several cycles,causing overshoot. When the feedback data is notupda

35、ted, the rate of change of the error signal is zero,eliminating influence of derivative control.4.2 Simulation accuracyWhilecomputer simulationsprovide much insightintothe effect of delays and multiplesampling rates,it is imperative that the simulation provide accurateresults, leaving data collectio

36、n for only fine tuning thePID parameters.欢迎下载精品学习资源The system has been modelled using discrete-timeelements in Simulink, a graphic user interface forMATLAB. This includes all known delays in the system. Step responses of the simulation at various gainsare compared to actual results to fine tune the

37、simulation. As can be viewed in Figure 4, the simulationsprovide a good approximation of the actual systemstep response for a stationary object approximately 4in 100 mm from the camera starting point.4.3 Tracking performance utilizing Dynamic Visual ServoingThe system is able to track an object on a

38、 conveyor belt movingat 400 in/minwith0.4 in 10 mmsteady state error. The limitingfactorin conveyorbelttrackingis initialimage acquisition.The stationary end-effector initially moves towards the object athigh speed as soon as it comes into view. If the visionupdate rate is too slow, the end-effector

39、 will pass beyond and overshoot the object before the next image iscaptured. Tracking will only be possible if the objectremains within the field of view as the end- effectordecelerates and reverses direction.When the object moves slower than 200in/min, steady state error is less than one millimeter

40、. Owing to limited workspace, it is not possibleto determine the steady state error when the objectmoves faster than 300 in/min, but it is less than 1 in25 mm and decreasing. The fifth line shows the response when the object is at first moving 300 in/minthen accelerates to 500 in/min at 6 seconds. W

41、hileovershoot is too great to track an object moving initially faster than 400 in/min, once the error is small,object speed may be greatly increased.Importantly,the system is able to tracka randomlymovingobjectwithvelocitiesup to150 in/minwhilemaintainingthe object withinthe camera fieldofview.The m

42、ajorlimitationin tracking randommovement arises from a change in direction betweenvision updates. The object may fall out of the camerafield of view before corrective measures can be taken.While better feedback bandwidth does not necessarilylead to better conveyor tracking, faster vision has atremen

43、dous impact on random movements. Using theolder vision system, the robot was only able to trackrandom movements with a maximum velocity of 12in/min.4.4 Three dimensional operationWhenonlytrackingan objecton a conveyor, someovershoot may be acceptable. On the other hand,when servoing vertically to gr

44、asp the workpiece, overshoot can damage object, robot, and workcell.As thecamera approaches the object, field of view reducesperceptibly allowing for the possibility that the object may disappear from the visionsystem. This is aproblem inherent in high speed dynamic visual servoing using an offset e

45、yeinhand configuration.This is alleviated by using a two stage vertical axisprofile. The vertical axis trajectory, handled by aseparate algorithm,allows fastservoing toa specifieddistance above the object,then slower,more precise movement as it nears the object. Final closuredoes not begin untilthe

46、object is withina given radial distance from the center of the end-effector. Thevisual servoing system is able to grasp objects spaced12欢迎下载精品学习资源inches apart moving 75 in/min. A picture of theindustrial visual servoing system in operation can beseen in Figure 6.Figure 6: Three dimensional visual se

47、rvoing system4.5 Hardware limitations4.5.1 Robot controllerTheADEPTprovidesrealtimetrajectorycontrolonlyintheend-effectorCartesian frame.Direct jointcontrolis not possible. The simple host program onthe ADEPTstillrequires on average 80 ms to complete one full cycle of reading in 12 bytes, placing th

48、econverted numbers into the realtime trajectory algorithm, and performing four safety checks. While hardware upgrades exist to alleviate these problems, therelativecostforanADEPTupgradeisextraordinary,approximately50%oftheoriginal robot/controllerpurchase cost.4.5.2 System robustnessPerformanceislimitednotonlybyhardwareconstraints,butbysafetymeasures as well. Robotworkspace must be constrained to a rectangular areaover the conveyor belt, preventingthe end- effect