runBehaviour() in recovery and costmap update

[GMahmoud]

Hi guys,

I have worked recently on a rotate recovery behavior. So I tested your implemented plugin and I noticed that the the local costmap (obstacle layer only) is not updated when the behavior is running. This issue is caused by the while(n.ok()){ ... } loop running in the main thread. In my opinion, it is better to create a new std::thread to start the recovery behaivor and to avoid blocking the costmap updates for the forward simulation.

What do you think ?

Cheers MG

[mikeferguson]

That does seem reasonable -- my only concern would be if it introduces any race conditions (but from a quick look it appears that the recovery behavior does not alter the cost map, so I think we're OK).

[casiarobot]

A case is that nobody can cancel the navigation task when in recovery behavior. Especially, a situation u have to stop the robot. U can publish /cmd_vel to zero, but when stop it the robot will continue running rotate recovery behavior.