Pose array's variance converge(navigation stack, amcl)

edit

Hi, I'm using nav stack with my robot.
I'd like to set a final variance of a robot pose array.

I set amcl's parameters "update_min_a" and "update_min_d" to 0.0
because I'd like to have the robot estimate its pose
even if it stops.

In this case, a pose array showen on rviz converges as soon as amcl node runs.
It's what I like, but the array over-converges.
I mean, all arrows showing robot's pose lie on top of each other.

I guess that some threshold is offered to avoid that.
However, I changed all parameters listed in amcl's wiki and couldn't find
which one is it.

Could you tell me which parameter is used to avoid the above problem?

Thanks in advance.