Added publishZeroVelocity() when state_ is changed to PLANNING.

[ryan-sandzimier]

If a robot is on it's way to the goal and a new MoveBaseActionGoal is published, the robot continues moving at the last commanded velocity for the duration of planning a path to the new goal. When the global planner is slow (such as for sbpl_lattice_planner), this is problematic. #750 and #751 fix this for the initial goal. However, if a new goal is published before the robot reaches the preceding goal, the same problem exists.

Solved this by adding publishZeroVelocity() when state_ is changed to PLANNING so that a zero velocity is commanded prior to making the plan.

[mikeferguson]

So, I know of at least a few people who leverage this "feature" - in that the robot won't stop real hard if they can actually get a new plan in time (basically, it's a lot smoother to let the robot coast for a second). I don't think we can merge this as-is into an LTS release that is 4 years into the cycle - but if you put it behind a parameter, that defaults to the old behavior, we could merge it.

[ryan-sandzimier]

So, I know of at least a few people who leverage this "feature" - in that the robot won't stop real hard if they can actually get a new plan in time (basically, it's a lot smoother to let the robot coast for a second). I don't think we can merge this as-is into an LTS release that is 4 years into the cycle - but if you put it behind a parameter, that defaults to the old behavior, we could merge it.

@mikeferguson Added a parameter that defaults to old behavior.