Odometry is relative to when the calculation was initiated/reset (or at least: that's how it's typically implemented in many ROS nodes/drivers). It's not world-referenced like you seem to be expecting.

So you where telling this is not possible?? but this problem was not there with turtlebot3. In turtlebot3 when i change the intial position of robot and checked the odom value it shows as same as of the intial pose value that i gave. This thing is not happening with turtlebot2

So you where telling this is not possible?

No, that's not what I'm writing at all.

What I wrote is that in most cases, odometry calculation is relative. And it's typically relative to when the algorithm was started. So that makes the point you start the algorithm 0, 0, 0 automatically.

but this problem was not there with turtlebot3. In turtlebot3 when i change the intial position of robot and checked the odom value it shows as same as of the intial pose value that i gave. This thing is not happening with turtlebot2

well, it sounds like it then that the authors of the TB2 node(s) implemented things differently than those of the TB3 nodes.

I would not know about that..

The only thing I wanted to point out is that you seemed to expect odometry to always be absolutely referenced against some external point, which I don't believe is necessarily true. And this may explain what you are seeing.

If the TB3 authors decided to initialise their odom calculations with ..

.. some externally referenced "start state", that would seem to be a design decision.

You'd have to ask them about that.