Little tidbit in Astronomical Journal, Vol 168, #6 (Dec 2):
Ai, G. Yin, J. Cui, L. Learning Gravity Fields of Small Bodies: Self-adaptive Physics-informed Neural Networks a 242 ad7951
The AI hype continues. At this moment, there is no real pressure to land on a specific point on an asteroid; simply making a safe contact is already an achievement. In the future, however, we will strive to reach a given place: a resource-rich area, perhaps a pole (for fast rotators), or something we don’t even know about today. Another group had already covered this just before.
Ai et al.’s group now takes a turn. Unlike Sun et al.’s group, they start with a crude initial model of the asteroid’s gravity field, then use Artificial Intelligence to ‘fill in the gaps.’ The net goal is the same: to fly about a body with a varying, irregular shape and mass, without first testing out the variance and irregularities. And not in a chance way: both NEAR Shoemaker and Hayabusa 1 approached and ‘docked’ with their target asteroids, eventually. They just made many exploratory orbits first.