Basic Info

Issue:

Points remain in obstacle layer after we stopped publishing a point cloud on regular basis. Settings to reproduce:

      obstacle_layer:
        enabled: true
        observation_sources: pointcloud
        pointcloud:
          topic: "blabla"
          max_obstacle_height: 2.0
          data_type: "PointCloud2"
          expected_update_rate: 0.0
          observation_persistence: 1.0
          clearing: true
          marking: true

According to observation_persistence: 1.0 the point clouds used by the obstacle layers should time-out after 1 second.

Analysis:

• last_updated_ is updated on every new cloud that needs to be added to the buffer.
• purgeStaleObservations checks the cloud's time stamp with last_updated_.
• When we stop publishing a point cloud last_updated_ will never be updated.
• Hence the observations will not time-out and remain in the observation_list_.
• This may result in the following warning, when the robot drives away from the clouds in observation_list_:

Sensor origin at (x, y) is out of map bounds. The costmap cannot raytrace for it.

Description of contribution in a few bullet points

• Fixed it by replacing last_updated_ with nh_->now() in purgeStaleObservations

Description of documentation updates required from your changes

• None

Future work that may be required in bullet points

• None