I am trying to build an occupancy grid map using odometry and 2D lidar data through graph optimization. I am not aware of any reliable loop closing methods using only the 2D laser scans. I have been reading about ICP for scan registration and am thinking that it may be what I need for loop closure detection.
What are some common methods of loop closure detection given only the 2D laser scans?
The most traditional method is to keep looking at the trajectory and see if your current location is close enough to the previously visited place. Once this happens run the ICP. If ICP converged normally, that is your loop closure.
A bit more advanced method is doing a place recognition. Generate a keyframe every few meters and extract a feature descriptor from the keyframe. Whenever you create a new keyframe compare this descriptor to the previously created descriptor. If you find any similar descriptor, you are likely re visited the place. Run ICP and see if it is really the case.
