I would point out that currently the system requires "a map, a destination, and a position estimate". While AMCL does maintain localization of the robot based on odometry+laser, it does not localize a robot without some apriori knowledge (the "initial pose").

Given that solving the kidnapped robot problem is still a fairly open research question, I don't think we want to presuppose that we will have such a solution in the new system.

@mikeferguson - That makes sense, I agree. I updated the README to include "a current pose" as an input.

As for the "Kidnapped Robot" problem, I don't expect this to be solved as part of this project. However I do expect it will be solved in the near future with a combination of positioning technologies, and when it is, this project should work the same as if a user designated an initial pose manually. :)

Any other comments or can I merge this?