Happy to report that our group's first two papers on the automated scheduling of Doppler observations are out. Over a typical six-month observing semester, a dozen or so research groups (ours included) pool time to execute ~3000 Doppler observations at Keck Observatory. Today, we're mostly using the newly commissioned Keck Planet Finder, although the old workhorse HIRES is still used for some legacy projects. The different programs have a range of timing requirements dictated by the observability of targets, distance to the moon, orbital periods of suspected planets, and so on. This is a challenging problem since there are N! ~ 10^3500 possible orderings of these exposures. This problem is closely related to several well-known problems in computer science such as the ‘set packing problem’ and ‘traveling salesman problem.’
Former UCLA undergrad Luke Handley (now an astronomy PhD student at Caltech) led two papers developing the mathematics for automated scheduling. The first solves the 'traveling telescope problem' and determines the optimal ordering of a set of exposures throughout the night that minimizes slew times. The second deals with the challenge of sprinkling the right mix of observations from a set of programs into a set of allocated nights. I, Velibor Mišić (UCLA), Jack Lubin (UCLA), and Howard Isaacson (UC Berkeley) assisted Luke with these publications. We've had a lot of fun collaborating with Velibor Mišić at the UCLA Anderson School of Business who studies the optimal allocation of resources for all sorts of business applications. It was fun to see our work featured in the Anderson Review.
We'll continue to develop these techniques in subsequent works, and Jack Lubin is leading the charge at UCLA. Special thanks to the Heising-Simons Foundations for supporting our work.
