controller for joint 1 was illustrated by the Horizontal Location of the Fruit Centroid. As

indicated in these diagrams, the vision controllers were not able to maintain perfect alignment

between the robot arm and the fruit. And, the resulting errors were greater than specified as

allowable for a successful pick. However, the controllers' actions still enabled the robot to be

able to track the fruit and remove it from the tree.

 The addition of the wait state in the intelligence base aided the controllers in the

tracking of moving fruit. This state provided a period of time for a swinging fruit to return to

the field of view of the vision system. This addition added a period of time to the pick cycle

time, but it made the difference in a successful pick attempt and an unsuccessful attempt. Thus,

in the overall picture, the time for removal of all of the fruit from a tree would be shortened.

 The use of the joint 2 velocity controller output for detection of a collision proved very

successful. The sensitivity of the controller to changes in the error provided an effective

method for detecting collisions before damage was done to the robot. This method of sensing the

force required to move the sliding joint also achieved good results when used for determining

when a fruit required more force for removal than was achievable by the arm.