SLAM but without actually driving the Robot?

Question

I am just starting with SLAM and have to implement SLAM to get a dimensions (map) of an indoor space.

I have looked at the turtlebot3 slam and slam_toolboox packages with algorithms like gmapping and cartographer. As far as I have understood, these packages requires the user to drive the robot around the room for the 1st time and subsequently we get a map of the room, is my understanding correct ?

If yes, are there any different approaches where the robot drives autonomously while also creating a map in the process ?

Edit: I plan on using ROS2 for implementation initially in a simulation and eventually on a real robot.

Thanks in advance!

Answer 1

This is a recent post on the ROS Discourse which could be of relevance:

Autonomous Mapping with SMACC & Isaac Sim.

Answer 2

As far as I have understood, these packages requires the user to drive the robot around the room for the 1st time and subsequently we get a map of the room, is my understanding correct?

Yes, SLAM systems often split operation into a 'creating a map' stage and a 'using a map' stage. Both of the stacks you've mentioned fall into that category.

are there any different approaches where the robot drives autonomously while also creating a map in the process

This is sometimes called "frontier exploration". There's a ROS 1 package (and this one) that's available on the ROS 1 repos, and it looks like there are some packages available on GitHub for ROS 2 that might be worth looking at (for example, this or this), but your best bet might be on the ROS 2 repos.

But you could also make your own, the general operation is pretty simple:

1. Start mapping with whatever existing stack
2. Start autonomously driving around by picking a goal within the known map that is near an unknown part of the map.
3. This should extend the map, so return to 2 until you're happy with the size of the map or you can't find any more goals.