140 Pickable Fruit
 Motions
 120

 100 Determined Actual
 Fmaruit Motions
 S 80



 E 40

 20

 0
 0.0 0.2 0.4 0.6 0.8 1.0 1.2
 Frequency (HZ)



Figure 9.17. Pickable fruit motions as determined from the closed-loop response of the
 manipulator along with the fruit motions determined to exist in the grove.


 found to be 6 pixels or less which was much less than the 65 pixel maximum allowed error. For

 joint 0, a maximum overshoot of 40 pixels was evident in similar tests of the compensated vision

 system. Similar tests were conducted for the joint 1 vision controller as shown in Figure 9.19. In

 this case, the fruit was placed away from the setpoint in the horizontal direction. The initial

 centroid was determined to be at 37 pixels in the horizontal direction of the image plane. The

 setpoint was set to 192 pixels, and the vision controller was actuated at 36 ticks on the plot. The

 vision controller positioned the end-effector so that the steady-state position of the fruit's

 centroid was at 191 pixels in the image plane with no overshoot. The performance of the joint 1

 controller was better than that of the joint 0 controller providing steady-state errors of only 1

 pixel while an error of 47 pixels would still have been acceptable.

 The ability of the robot to track and pick moving fruit could not be totally solved by the

vision controllers. The over design of this first prototype citrus picker resulted in large link

inertias and the use of oversized actuators which contributed to its slow frequency response.

Because physical changes in the robot (e.g. different actuators and reduced link masses) which

would improve its frequency response were not practical, the solution for this obstacle was to