How to make an accurate map?
It seems like the slam algorithm is really tricky. I tried in my living room and it works fine(demo: https://youtu.be/ml_pgbPEIz8 and https://youtu.be/yCTHY2m6ByE , there are happening roughly the same time). However I tried in some rooms in my school, since I want to do a demo of navigation for students, it doesn't work well. I tried in a room with hard floor and a room with rough carpet. Both them doesn't give good map. I wonder if the slam algorithm itself doesn't work well for a simple small confined square shaped room, because it can't have a good prediction about its' location if all corners looks similar. The two rooms I tried are all standard college small classroom. The shapes are square and with lots of chairs in the middle.
What would be the best environment to demo slam? (I guess the best is hard floor room that's a little more complicated like this https://youtu.be/lkW4-dG2BCY ) Are there good resources regarding to this problem?
Thanks!
Slam tends to do well in environments where the sensor can see the walls, and the walls and obstacles have some unique features. Putting a few large boxes in the middle of your classroom may help provide unique features.
I would suggest you to use Hector SLAM which gives better maps for indoors.