Then, I should define the objects as dynamic, to let Bullet to calculate the position of the object.
No, that's not what I intended.
I elaborated on the needs of DYN
objects in a specific thread about a month ago: Getting the rid of kinematic objects
. Here, I consider the need to "hybridize" a kinematic
object with some dynamic
object behaviour and come to the conclusion that KIN
objects are not intended to be the solution for that specific usage.
Whatever you have the same need is unclear to me.
I cannot really tell if Bullet is going to compute the position according to velocity but I expect it won't - kinematic
objects are considered static
AFAIK (with the exception to have their transforms updated; only if not sleeping) - what I expect is it will pull out the velocity vector for energy transfer computations. The object will still not move but hopefully it will push stuff around much more.
But that's just speculation.
Of course, if you can get the rid of the KIN
object that would be just fine. There are a few cases in which this is not advisable nor viable.