Jinio,
do you know anything on your plant, like transfer function (tf), magnitude and/or phase diagram, etc.?
If so, try to figure out with a tool (LabVIEW Control Design) or by hand something on the phase and gain margin.
Depending on this result and depending on what type of controller you are looking for you have to choose the correct gains/time constants.
The upper/lower limit is good for limiting the set value because depending on your controller it might happen that the set value increases to a multiple of the allowed stuff.
I highly recommend, to simulate the plant in combination with the controller before testing it on the real object.
If this isn't possible for you, start by using P=1, I and D = 0. Increase P as long as the critical point of the plant is reached. After that you can calculate the "safe" P, I and D stuff.
For more information on that method, look for the method called "Ziegler - Nichols".
BR,
ThSa