[Question] experience with ProbabilisticOccupancyMap

[facontidavide]

Hi,

I noticed that the class FrequencyOccupancyMap seems to be preferred over the ProbabilisticOccupancyMap.

I have taken a look at the underlying algorithm and seems like the latter mimics the behavior of Octomap, somehow.

I wonder what your experience and recommendations are, when using one over the other.

[eupedrosa]

Hi @facontidavide.

In the literature, you will always find the ProbabilisticOccupancyMap as the "map" to use. But in practice, the FrequencyOccupancyMap provides better results for mapping. At least from my experience.

I usually use the ProbabilistiOccupancyMap for perception applications, where the occupancy map has to be more dynamic and quick to change. The FrequencyOccupancyMap is more conservative, it takes more effort (or observations) to go from occupiend to free.

And yes, the ProbabilisticOccupancyMap does mimic Octomap. But for integration, you can chose to use DDA like Octomap or the more simpler integer Bresenham algorithm, which is faster. The outcome is not very different due to the amount of data that we usually use in perception.

[facontidavide]

Thanks!