Reverse behavior in TrajectoryPlannerROS

[Matt Anderson]

Hello, I have a diff drive robot that I am attempting to integrate with nav stack, specifically move_base. I am having a difficult time when my long rectangular robot reaches a dead end (such as the end of a hallway). https://drive.google.com/file/d/0ByAqfl4quhJdVUhvOVNFZHloWm8/view?usp=sharing is a link to a video from rviz that demonstrates the problem. It could be solved by simply reversing until it can turn around, but it appears the planner isn't willing to generate reverse trajectories for long enough to make it happen. Is anyone facing this problem, or are there typical work arounds for a robot that doesn't mind reverse? Or is there some method/planner you could point me toward to best implement this behavior?  Since I saw(lidar) the objects on the way in reverse is really just as good as forward in my case (objects aren't moving).

[Martin Günther]

Hi Matt,

You should set the `min_vel_x` parameter of your local planner (e.g.,
DWAPlannnerROS or TrajectoryPlannerROS) to something negative (e.g. -0.3
for 0.3 m/s backwards).

BTW, this kind of question is better asked at http://answers.ros.org .

Cheers,
Martin

> --
> You received this message because you are subscribed to the Google
> Groups "ros-sig-navigation" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to ros-sig-navigat...@googlegroups.com
> <mailto:ros-sig-navigat...@googlegroups.com>.
> For more options, visit https://groups.google.com/d/optout.

[Matt Anderson]

Hello Martin,

Thanks for your reply. It turns out this is only true for DWAPlannerROS.  It appears TrajectoryPlannerROS will only generate negative x trajectories when escaping. Additionally, at least in my case, DWAPlannerROS doesn't preform as well when the robot is moving backwards.  However, it could be that the turning point of my robot isn't in the center.  Anyway, I will post any further questions on the answers sight...I didn't know which to use until finding this question after your post.

> an email to ros-sig-navigation+unsub...@googlegroups.com
> <mailto:ros-sig-navigation+unsub...@googlegroups.com>.