The questions is quite straightforward. For the past few days I've been looking into point cloud registration. It looks like it all started with the basic ICP algorithm as described in P. J. Besl and N. D. McKay, "A method for registration of 3-D shapes".
Since then, there have been possibly thousands of papers describing faster or more accurate variants of this algorithm. Is there a go-to variation of the algorithm or some variations that stand out nowadays?
I know different papers focus on different parts of the algorithm. Others try to find a good correspondence between points, others try to increase the speed of specific methods of correspondence, others try to take advantage of RGB values or any other data collected from images accompanying the point cloud...
It's been chaotic trying to pick a specific algorithm for this. I'm mostly interested in real-time registration if that helps. I can use RGB data in my implementation but I'd also like to know efficient ways without image data.