Local path planner (dwa) has strange behavior

asked 2018-10-10 13:48:06 -0500

Hi there!

I am trying to let a robot with omni-directional drive (ridgeback) follow a global path generated with global_planner::GlobalPlanner from navigation package. I am using the ridgeback_gazebo simulation for this. The local planner is dwa_local_planner::DWAPlannerROS. I am using standard parameters for both global and local planners as well as the cost maps from ridgeback package (see here), but also tried to modify multiple parameters. However, I get the strange behavior that sometimes the robot stops even if no object is near and turns on point. This can be seen here at 0:20 : Video.

I would love to know why this occurs. I also tried TrajectoryPlannerROS and set the weights such that following the global path has extremely high priority, but unfortunately with the same behavior.

Platform: Ubuntu16.04 with ROS Kinetic Versions: Everything installed from apt in the last 7 days

Video description: This is a recorded map from the ridgeback_gazebo standard simulation environment, yellow displays the global path, blue the local path and the local costmap can be seen. I am using amcl / map_server for localization. Thank you for your time!

answered 2018-10-10 16:02:28 -0500

shiv_rar
46 ●3

Hey, just did a quick look at your params and just had some minor advice on what you could adjust, to kind of improve.

First thing is the inflation layer:

inflation_radius: 0.25

This should be comparable to the full size the robot footprint, in your case (0.96 length and 0.8 width).

From how the costmap keep out and decay regions work having a small inflation_radius in comparison to the footprint is going to cause some problems when it tries to make that U - turn like movement in the global plan.

Further reading here: http://wiki.ros.org/costmap_2d

My recommendation is to try adjusting the parameters of the inflation layer like so:

inflation_layer:

 enabled:              true
 cost_scaling_factor:  0.3  # exponential rate at which the obstacle cost drops off (default: 10)
 inflation_radius:     1.0

If this doesn't solve your planning issue is it possible to post a screenshot or error messages that come up on the command line during the execution?

edit flag offensive delete link

Comments

First of all, thank you so much for your help!

I modified the inflation layer params to your suggested params. This definitely helps. However, at certain angles, the behavior still occurs, even with inflation_radius: 2.0

Screenshot