Robotics StackExchange | Archived questions

clearing obstacles from local costmap faster

hi everyone

when we are using navigation stack, how we could clear faster the obstacles from local costmap?? I mean some dynamic obstacles which pass across the robot and it recognizes them intime and show them in costmap, but the robot is slow to clear them after they pass and it causes some problems in navigating.

This is my costmapcommonparams.yaml and localcostmapparams.yaml. Should i change updatefrequency or publishfrequency??

costmapcommonparams.yaml:

obstacle_range: 2.5 raytrace_range:
3.0 footprint: [[ -0.146,  -0.372],
        [ 0.154, -0.372],
        [ 0.354, 0],
        [0.154, 0.372],
        [-0.146,  0.372]]

inflation_radius: 0.55

observation_sources: laser

laser: {sensor_frame: laser,
data_type: LaserScan, topic: /scan,
marking: true, clearing: true}

localcostmapparams.yaml:

local_costmap:   global_frame: odom  
robot_base_frame: base_link  
update_frequency: 5.0  
publish_frequency: 2.0  
   static_map: false   
   rolling_window: true   
   width:6.0   
   height: 6.0   
   resolution: 0.05

Thanks:))

Asked by mohammad on 2014-12-31 15:57:07 UTC

Comments

Also having the same problem. Anyone have a fix?

Asked by musiqsoulchild on 2015-03-17 00:03:22 UTC

Do you have clearing observations that show that the obstacle is no longer there?

Asked by David Lu on 2015-03-17 11:22:11 UTC

Not sure what you mean by clearing observations but when i restart move_base, the costmap works fine and doesn't show obstacles that aren't there. Not sure what parameters to change.

Asked by musiqsoulchild on 2015-03-21 21:55:07 UTC

Obstacles are meant to stay in the costmap until something gives evidence that they're not there. That evidence could be a clearing observation, i.e. a laser reading that is behind where the obstacle is.

Asked by David Lu on 2015-03-22 14:10:03 UTC

So I'm using a hokuyo urg 04 lx and for some reason when that sees a long hallway, it displays NaN in the /scan topic. This causes the local costmap to show an obstacle, and if i keep moving straight (and there is a hallway perperdincular), the obstacle keeps moving in parallel because of NaN scan.

Asked by musiqsoulchild on 2015-03-22 15:24:06 UTC

I tested it without long hallways and didn't have this problem (no NaN in the scan topic), however when it sees a long hallway, the NaN ruins the local costmap. I tried to use the laser scan filters but I don't know how to filter NaN scans.

Asked by musiqsoulchild on 2015-03-22 15:26:39 UTC

have you tried the

  • ~//inf_is_valid (bool, default: false)
  • ~//observation_persistence (double, default: 0.0)

as described "http://wiki.ros.org/costmap_2d/hydro/obstacles"

Asked by daniel_dsouza on 2015-07-17 12:32:15 UTC

Using ROS indigo with the layered costmap and costmap plugins -- is there a way to set a decay rate or something similar so that human scans fade away even if the robot doesn't scan the area again? The output of rosparam get / | grep -in *observation_persistence* is empty for me

Asked by j_salfity on 2017-03-08 14:29:36 UTC

@j_salfity a costmap layer with decay rate can be found in the spatio_temporal_voxel_layer from @stevemacenski. Answering to this comment also made me realize that Steve also implemented nonpersistent_voxel_layer which would be as simpler version (no decay though).

Asked by awesomebytes on 2019-06-17 02:41:01 UTC

Answers