performance and the known responses of the joints. The parameters were adjusted to achieve

acceptable responses from the joints and respectable performance in the robot's picking process.

With the tuned velocity and position controllers, all of the performance requirements were met.

 After tuning the velocity and position controllers, vision controllers were added to the

software package. These controllers used information from the vision system to align the robot

with a targeted fruit. These controllers were first tuned to achieve acceptable performance

with nonmoving fruit. After showing exceptional steady-state errors of 6 pixels or less, their

response to moving fruit was investigated. From the worst case motions of fruit as determined in

the fruit motion test, amplitude ratio and phase lag requirements were specified. These

requirements could not be met by the vision controllers. By utilizing frequency response curves

for the design of the controllers, it was found that meeting these amplitude ratio requirements

would cause the system to be unstable. Also, the vision controllers could not be tuned to achieve

the small phase lags which were required. These obstacles were overcome by altering the

intelligence base of the robot to either wait for the velocity of moving fruit to slow to a more

achievable rate or to search for another fruit.


 Conclusions

 While incorporated into the robot software environment, these controllers for the

orange-picking robot provided acceptable performance, enabling the robot to successfully locate

and pick fruit. Ultimately, the controllers would have provided a very fast response to input

signals with a minimum steady-state error and overshoot. Since the robot had been designed

and constructed before the controllers were designed, major changes to the construction of the

manipulator were not possible that would make the control process easier and better. However,

these controllers furnished sufficient responses of the joints to changing input signals. When

incorporated with the intelligence software of the robot, picking rates of 1 fruit every 4 seconds

were achieved. Also, in 75 percent of the attempts to pick a fruit, a fruit was removed from the

tree. These controllers also aided in proving the concept of mechanically picking oranges