14-02-2013, 04:28 PM
A Kinematic Calibration Method for Industrial Robots Using Autonomous Visual Measurement
A Kinematic Calibration.pdf (Size: 437.23 KB / Downloads: 100)
Abstract
Several new methods have been developed to achieve practical accuracy for offline programming of
robots and its applicability to the real world. In this paper, a new kinematic calibration method is proposed
to automatically improve absolute positioning accuracy of robots. Key points of the method include
autonomous measurement and the automatic generation of measuring poses. A new visual feedback
motion control method of the robot is proposed to achieve accurate measurement. An algorithm is also
proposed to improve the condition of measuring poses automatically. The effectiveness of the proposed
methods and algorithm was investigated through experiments with actual robots.
INTRODUCTION
Offline teaching is now being recognized as a necessity to
shorten start-up time of industrial robot systems, and to
thus, grow industrial robot applications. It is also well known
that, improving absolute positioning accuracy of a robot,
and improving detection accuracy of the location of
workpieces in a matter of minutes are two very challenging
hurdles to putting offline teaching into practical use. The
authors have previously developed workpiece position
detection methods using autonomous visual measurement
in robotic cells [1]. Currently, the improvement of absolute
positioning accuracy of an industrial robot with serial
mechanisms by kinematic calibration is being studied
intensively.
Various methods of kinematic calibration for an industrial
robot have been reported [2][3]. However, very few
calibration methods have been experimentally confirmed as
practical for use on the shop floor. Industrial robot
positioning errors observed on the shop floor include those
due to varying environmental conditions, such as changes
in temperature and load, that are difficult to predict prior to
robot shipment. In addition, it is impossible to predict errors
due to plastic deformation in robot links resulting from
mechanical damage. To decrease these types of errors, a
practical kinematic calibration method which cannot only
achieve indicated positioning accuracy but also be used
easily and fast without changing the set location of the robot,
is required.
Automatic Calibration Process
As a minimum requirement for autonomous measurement
using a CCD camera, the operator must move the camera
mounted on the tip of robot arm to a position where the
target can be viewed by the camera, and teach the position
to the robot as the initial measuring pose. To improve ease
of operation and reduce total operating time, a 2-step
automatic measuring process is proposed which does not
require the specification of initial values for camera
mounting and target positions.
1. Inch the robot keeping the target in the view and
measure robot poses around the single measuring point
using visual touch-up. Then roughly identify the
positions of the camera and the target but with enough
accuracy to achieve the next measurement for
kinematic calibration.
Error model
The error model for the experimental robot is shown in
Figure 8. Since robot links and drive systems have been
processed and assembled under strict dimensional control,
elastic deformations are assumed as the main cause of
absolute positioning inaccuracies at the tip of the robot arm.
By considering that elastic deformation occurs mainly in the
reducer mounted on each rotation axis, elastic
deformations caused by load are defined as a shift in the
zero point on the reducer of each rotation axis. Because the
relative position among the tip of robot arm, PVT (Tip of
virtual pin) and PM (measuring point on the target) are used
in identification, zero point error of J1 and J6 are dependent
on error of pVT,0 and pM,0. Therefore, independent zero point
errors of J2, J3, J4, and J5 axes
defined as kinematic parameters of robot pk, and errors
[ΔXVT,ΔYVT,ΔZVT]T included in pVT are defined as fixture
parameters of virtual pin PVT, errors [ΔXM,ΔYM,ΔZM]T
included in pM are defined as fixture parameters of target
PM. In this paper a total of 10 calibration parameters were
identified.
Calibration Result
Relationship among variations in identification results,
condition number and measuring method
At first, the initial vale of (DW,DP,DR) was set as
(5deg,5deg,5deg). With preset thresholds and searching
step (each 5deg in WPR), several sets of measuring poses
were determined by using the proposed algorithm. In this
case, time required for generating, adjusting and evaluating
a set of measuring poses is approximately 15sec by using
the robot controller. To evaluate the effectiveness of
condition number the threshold was changed several times
and then several sets of measuring poses were obtained.
Four sets were selected randomly for the evaluation. In
order to evaluate the relationship between variations in the
identification results and the condition number, 10 sets of
measuring data obtained from each set of measuring poses
were used for the identification.
CONCLUSIONS
A new automatic kinematic calibration method using an
autonomous measurement and an algorithm for
determining measuring poses based on the condition
number was proposed and its effectiveness was
experimentally investigated. The conclusions are
summarized as follows:
1. In order to improve accuracy of the identification and
shorten measuring time, a visual touch-up (rather than
stereo measurement) was proposed.
2. In order to increase efficiency of the measurement in the
calibration, an algorithm was proposed to determine a
set of measuring poses based on the condition number
and working space of an industrial robot on the shop
floor.
3. Using the methods proposed in 1 and 2, an automatic
calibration method was proposed.
4. The effectiveness of the proposed methods was
evaluated through experiments with an actual industrial
robot.