feat: Clearing expired cells after timeout

[renan028]

Description

AB#22914
In obstacle layer, adding logic to clear cells if they are not scanned after some time.

Costmap has a persistent logic, thus each cells' costs remain the same until new scan pass through it or "hit" an obstacle on it.
I am adding an option to clear the cell after some "expiration" time, i.e., if the cell cost does not change after a while (cell_clearing_timeout), we change it back to NO_INFORMATION.

cell_clearing-2024-10-21_11.59.34.mp4

[corot]

Are we creating a Timer for each cell? Sounds rather overkilling
What's the impact in performance?
And why not using STVL instead?

[renan028]

Are we creating a Timer for each cell? Sounds rather overkilling What's the impact in performance? And why not using STVL instead?

It is a ros::Time for each cell, not ros::Timer
ros::Time is not overkill, it has 4 bytes (sec) + 4 bytes (nsec), thus same as double.

What's the impact in performance?

If the logic is not used (value -1), the impact is zero because it exits instantly.
I will investigate the impact when the feature is on.

And why not using STVL instead?

STVL has the same logic with OpenVDB, each voxel has a timestamp, but it is double.
I think it is overkill to use STVL, since it introduces the whole OpenVDB (tree) complexity and conversions, which we don't need. Besides, it does not have the Raytrace logic outside the map, which we added in the obstacle layer as an option, so we cannot use it.
I think we could either keep it as a parameter here, or create a new costmap layer with the new logic of cell decay.