Inconsistent behavior for update rate between stvl and navigation2/nav2_costmap_2d

[sjusner]

When comparing the behavior of the stvl and the nav2 obstacle layer some differences in behavior emerged during troubleshooting an issue. In the obstacle_layer (more precisely the used observation buffer) stale observations get purged when they are older than the specified observation_keep_time_ (set via the observation_persistence parameter) compared to `clock_->now() here. For the stvl, this is compared to the last update time here.

This leads to inconsistent behavior when sensor data is suddenly dropping out.

• For the obstacle_layer, the observations get removed from the buffer after the set time as would be expected.
• The stvl keeps observations past the set observation_persistence as stale buffer entries get only removed when new observations are coming in (due to the update of last_updated_).

I suggest to replace the stale check in the stvl to compare with the current time instead of the last_updated_. Further, as commented in this issue, the update of last_updated in the reset function seems misplaced, and interferes with the expected_update_rate calculation.

Any thoughts on this?

[SteveMacenski]

compared to `clock_->now() here. For the stvl, this is compared to the last update time here.

I checked some of the older git commits and it looks like the behavior in Nav2 was updated but STVL wasn't updated with it. This used to be how Nav2's / ROS Navigation's observation buffer handled it.

Here's the PR that made the change in Nav2 ros-navigation/navigation2#1872

Id support that update. Can you open the PR?