Difference between "trav_map" and "obstacle_map" in core/physics/scene.py

[stanfordvl]

Difference between "trav_map" and "obstacle_map" in core/physics/scene.py

[ChengshuLi]

Sorry for my belated reply. I think "v3" map is generated using a slightly different heuristic. Neither of them are completely perfect. Feel free just to use the normal map and ignore the "v3" map.

[ChengshuLi]

Thanks for the question.

We use two different heuristics to generate trav_map and obstacle_map (e.g. floor_trav_0.png and floor_0.png). They are used together in the code to best represent the traversability of a scene.

iGibson/gibson2/core/physics/scene.py

Line 218 in 822ff1a

trav_map[obstacle_map == 0] = 0

Essentially, we take a conservative approach and treat an area as traversable or empty only if both trav_map and obstacle_map represent that area as empty (with pixel value 255)

For example, this is the "trav_map" of a scene called "Rs".

And this is the "obstacle_map" of the same scene.

Please let me know if you have any further question.

[WSPeng]

Thank you, I got it now.

[WSPeng]

Hi, just go another related question, some image has name ended with "_V3". What is the difference between the "v3" map and normal map? Thank you!

[yimengli46]

Is there an obstacle_map for the interactive scenes, like Rs_int?

I want to avoid the agent collide with the furniture.