ASTRO-TSP

Speaker: Mitchell Humphries, George Matta and John Korah (Faculty Mentor)
Department of Computer Science
California State Polytechnic University, Pomona

Meeting ID: 833 7638 0746
Passcode: 469907
https://us06web.zoom.us/j/83376380746?pwd=QVY4MWNMKy9HTno3dDQxQit2ZEFGZz09&from=addon

Time: 9-10am

Place: Cahill 273

Title: ASTRO-TSP: Traveling Salesman Problem based Solutions for Scheduling Astronomical Observations
Abstract: In this presentation, we introduce three algorithms designed to automate the scheduling of astronomical observations in an observatory. Scheduling astronomical observations is computationally challenging due to the need to consider observation windows, time dependencies, and varying priorities. Representing our problem as a variant of the well studied traveling salesman problem (TSP), we investigated scheduling solutions that utilized greedy algorithm-, integer linear programming (ILP)- and genetic algorithm- based methodologies. The genetic algorithm proved to produce schedules which best maximized the utilization of an observatory for a given observation period. While the greedy algorithm performed well for a limited number of observations, the genetic algorithm based solutions proved to be more scalable and produced better results for larger problems.  Furthermore, we leveraged parallel processing designs on multi-core and graphical processing unit (GPU) architectures to scale up with larger problem sizes while being able to generate the schedules within time constraints. Our techniques have the potential to greatly increase an observatory's efficiency while also saving astronomers time by reducing tedious trial and error.  We will present example results generated from lists of targets and expected observing conditions for the Next Generation Palomar Spectrograph.