Data processing: measuring – calibrating – or testing – Measurement system – Orientation or position
Reexamination Certificate
2001-05-14
2003-03-04
Hilten, John S. (Department: 2863)
Data processing: measuring, calibrating, or testing
Measurement system
Orientation or position
C702S127000
Reexamination Certificate
active
06529852
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The invention concerns a method and device for improving the pose accuracy of a mechanism having tolerances in a workspace, the mechanism being movable in at least one axis and having an effector, wherein at least one effector object is fixedly mounted on the effector and is eccentric relative to the at least one axis of the mechanism and at least one reference object is arranged in the workspace so that it is fixed relative to the mechanism. A computer system with measurement and control programs is operatively connected to the mechanism. The effector and reference object form a signal trigger/signal detector pair suitable for enabling the triggering and detection of at least binary signals, wherein the totality of the signal poses of the detector relative to the trigger device in which a signal is triggered on the detector may be described by at least one non-trivial characteristic equation.
2. Description of the Related Art
A number of methods for improving pose accuracy of a mechanism are known in the art. For example, FR-A-2 696 969 discloses a calibration method in which a laser beam is fastened at the last segment of a robot and a measurement plane close to the robot is used as reference. The robot approaches a series of hand poses (which are not more closely specified) in which the laser beam meets the measurement plane. After this, the coordinates of the contact points on the measurement plane are identified photographically. From these and from the associated joint configurations from the estimated values for the pose of the measurement plane and from the pose of the laser relative to the hand, the robot parameters are calculated by means of a variation of the Newton calculus of observations method (Levenberg-Marquart method).
In the primary embodiment of this reference, the measurement plane is a mirror or opaque projection screen and the laser beam which meets it is photographed by a camera positioned in front of the mirror. The use of markers on the projection screen is not explained any more closely and is apparently intended to serve the creation of a relation between the camera and projection screen. The measurement plane may also be designed as an optical matrix sensor.
In accordance with further embodiments, two measurement sequences are photographed. Between the two measurement sequences either the laser is mounted at another point, the position of which with respect to a tool support, e.g. point of the hand, is known or the position of the mirror is altered—the orientation is not mentioned in this reference.
The altered poses are not specified any more closely. The following explanation is given: if the measurement sequence is selected unintelligently, several values may come into question mathematically for identification of the parameters. To prevent such a problem, two different mirror poses are suggested. Problematical is the fact that, using a laser beam, it is not possible to identify the position of the laser or the tool relative to the root of the hand completely. In addition, two intersecting laser beams are used to identify the 5
th
and 6
th
parameters of the tool position. Moreover, the system is inexact due to the distortion of the image of the measurement plane by the camera.
Further attempts to identify robot parameters are apparently contained in an essay by Newman, Osborn, “A new method for kinematic parameter calibration via laser line tracking”, Proc. Int. Conf. Robotics and Automation, USA, Atlanta (1993), p. 160-165. In this reference, a laser beam is fixedly set up in space close to a robot. A special detector is fastened to the hand of the robot that consists of a planar, rectangular light sensor divided into four separate quadrants. The quadrants meet at a center-point. Each of the four quadrant sensors provides a brightness value. The robot moves the hand successively in various measurement poses defined by the fact that the brightness values provided by the four sensors are identical. The robot parameters are calculated using Newton's calculus of observations from the joint configuration of the measurement poses and the estimated values for the pose of the laser and the detector relative to the hand. The authors describe a principal structure of the experiment and then report on results of a two-dimensional simulation of their principle on the basis of a two-dimensional robot with two-joints. Not all kinematic parameters may be determined by this method. In particular, the pose of the robot relative to a prescribed coordinate system cannot be identified but only dimensionless parameters, thus making it impossible to derive an absolute size standard of the robot. It is not explained what has to happen in three-dimensional space if the sensor plane does not stand perpendicularly on the laser beam.
The re-calibration of small parameter modifications is described by an essay “Autonomous robot calibration using a trigger probe” by Zhong et al. in the US magazine Robotics+Autonomous Systems, 18 (1996) S. 395-410. Three plates are fastened in the vicinity of the robot which stand exactly perpendicularly with respect to each other. The robot takes up an omni-directional mechanical probe and approaches a series of hand poses which are not specified more closely and in which the probe touches the plate or triggers the internal contact of the probe. The associated joint configurations are evaluated by a neural network which supplies the robot parameters as result. True sizes cannot by determined by this method as only relative modifications are recognized. Some information is lacking, for example, the true distance between robot and plates, the authors ascertain.
The disclosures of WO96/30171 and WO 93/11915 describe methods and devices for the calibration of the axes of motion of industrial robots.
In WO 96/30171, a calibration device is used which consists of a calibration beam, e.g. a laser in the workspace of a robot, and an associated interrupter detector. A sphere with known radius is mounted on the hand of the robot. The robot heads for a series of hand poses, not specified more closely, in which the calibration beam of the sphere is interrupted. The calibration parameters are calculated by means of the Newtonian-Gauss method from the associated joint configurations, the estimated values for the robot pose, and the pose of the laser relative to the hand. In the preferred embodiment, the calibration beam has to stand perpendicularly on the x-y plane defined by the robot basis. With certain exceptions, six calibration parameters are calculated for each axis.
The accuracy of the calibration parameters may be increased by calculating them several times. In doing so the calibration beam is put into various positions in the workspace. The calibration parameters are then calculated as the mean value of the calibration parameters for the various beam poses. To obtain the greatest variations between the robot configurations used, the robot may be equipped with several calibration beams the pose of which is selected in such a manner that the greatest possible differences between the robot configurations during the various measurements are achieved.
According to the WO 93/11915, a calibration body is used which consists of a cuboid with exactly parallel lateral sides in the workspace of the robot. A sphere with known radius is mounted on the hand of the robot. The robot heads for a series of pairs of hand poses, not specified more closely, in which the sphere touches the cuboid once on any arbitrary side of the cuboid and then again on the opposite side. The presentation and the manner of the subsequent calculation indicate that the second contact point has to lie exactly on the perpendicular of the cuboid point opposite to the first point.
The calculation of the robot parameter takes place in iteration steps. In each step, the relevant coordinate differences of the associated pairs of hand poses are determined first of all on the basis of the current approximation values for all p
Knoll Alois
Kovacs Peter
Cohen & Pontani, Lieberman & Pavane
Hilten John S.
Sun Xiuqin
LandOfFree
Method and device for the improvement of the pose accuracy... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and device for the improvement of the pose accuracy..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and device for the improvement of the pose accuracy... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3010317