Paradigma del desarrollo sostenible

As was stated in section 6.2, the PUMA manipulator was required to follow a moving workpiece in two dimensional space. The stereo vision system tracked the workpiece as it moved relative to the cameras and thus relative to the manipulator shoulder. The spatial data gathered by the vision system was then used to modify the path of the manipulator such that the manipulator end flange remained fixed relative to the target.

An inertial system comprising an extensible ’finger’ connecting the manipulator to the target was used to measure errots between the desired (target) trajectory and the actual (manipulator) trajectory.

6.5.1. Hardware

A vertically mounted x-y plotter was used to move the workpiece. The targets consisted of contrasting circles fixed at varying depths on the workpiece, thus creating a three dimensional moving target as shown in figure 6.2. The workpiece was attached to the pen holder of the plotter.

rigid workpiece -

holding the targets motion of workpiece

motion of workpiece

rod connecting workpiece centre to

PUMA end flange

2dof ball joint

flat white discs (targets) raised above the level of the dark workpiece by differing amounts

FIG. 6.2 - FIXED TARGETS MOUNTED ON A WORKPIECE

The relative position between the target and the PUMA was measured with the moveable 'finger' shown in figure 6.3. This was mounted between the PUMA end flange and a rotary joint connected to the plotter pen holder. It comprises two potentiometers able to move ±30° in a rigid housing and giving a voltage according to the x and y

displacement of the PUMA relative to the target. Any z-axis motion is taken up by a linear variable differential transducer (LVDT) with a stroke of 44mm.

These three voltages were read by a Data Translation 2801-A analogue-to-digital converter mounted in an Elonex 386 PC. The output voltages to the x-y plotter were generated in the same PC by an Amplicon PC24 digital-to-analogue converter. Prior to each set of tests a calibration program was run, moving the target whilst holding the PUMA motionless. This gave a relationship between the voltages recorded and the length of misalignment that this represented.

PUMA manipulator end flange X and y motion potentiometers manipulator motion potentiometer housing LVDT 2 dof ball joint raised 3D _ vision targets flat workpiece following the path of the x-y plotter workpiece motion

FIG. 6.3 - RELATIVE ERROR MEASUREMENT PROBE IN ELEVATION

The experimental equipment was logically divided into the four systems shown in figure 6.4. These systems are:

• TARGET SYSTEM: The target system PC drives the x-axis and y-axis of the x-y

plotter independently. Each axis is driven in a separate sinusoidal pattern, varying in frequency and amplitude between tests. There is, additionally, a random phase difference between the start points of each axis. As each new motion is commanded, the PC logs the error voltage signals from the measurement probe and the requested current target position. All data is stored as ASCII text matrices in *.m files for subsequent off-line manipulation using the MATLAB mathematical software [Matlab UG 92]. For test one (section 6.5.2 details the tests), it also sends the new motion to the PUMA system controller, via an RS 232 serial link running at 19200 baud. For tests two and three, it does not send any information to directly control the PUMA system.

• PUMA SYSTEM: This is the PUMA manipulator and the Unimate controller. This controls the manipulator position according to the external information received via an external cable link either from the TARGET system PC or from the ARM system PC.

ARM SYSTEM: This sends the position data to the PUMA system for tests two and three (section 6.4.2 details the tests). For test two it receives the true plotter position from the TARGET system. For test three, it receives the target position from the VISION system.

VISION SYSTEM: This sends three dimensional target position information to the ARM system for test three. Although the targets move in a two dimensional plane, the cameras are set up such that they are not directed along an axis perpendicular to this plane. Thus, relative to the cameras, the targets are moving in three dimensions.

VISION SYSTEM PUMA SYSTEM TARGET SYSTEM

x-y plotter PUMA 560 manipulator plotter control signal error data stereo camera

system PUMA 560 _controller

test 2 & 3 vision system ^ controller = 68030 processor vision ^stem PjCE ARM s ystem P ARM SYSTEM

KEY: arrows indicate lata flow w hicIÏÏSyl^ control

d& orsa^data ^

6.5.2. Experiments

The experiments were aimed at establishing the performance of the vision based motion compensation system. Using the system designations shown in figure 6.4, the following three tests were carried out:

1. Errors due to the PUMA system. This involved only the PUMA and TARGET system. The method was to send the identical position data fi'om the plotter control PC to both the plotter and the PUMA manipulator and measure the position errors. 2. Errors due to the ARM and PUMA system. This involved sending the identical

position data to the plotter and the ARM system PC and measuring the position errors.

3. Errors due to the VISION, ARM and PUMA system. This involved sending the position data to the plotter only. The VISION system then calculated the target positions, transmitted these to the ARM system PC which commanded the PUMA accordingly. Again the position errors were measured.