I posted another video like this a while ago, but I always get amazed at them. I can't help but think how shape and boundaries are enough for vision. I'd like to write up what's the role of the contrast sign in producing these effects. For one, the shadows being black helps a lot. The black traces the robots leave on the street make it look like it is covered in snow. Here is a simple inversion of the video (no sound so you can play them together without interference):
Shadows become highlights, and the ground seems to be covered in water instead of snow. It would be cool to have another version of the vid with only the boundaries between the black & white regions.