ASCL.net

Astrophysics Source Code Library

Making codes discoverable since 1999

ASCL Code Record

[submitted] A Neural Network for the Identification of Dangerous Planetesimals (Including scripts for data generation)

Two neural networks were designed to identify hazardous planetesimals that were trained on object trajectories calculated in a cloud computing environment. The first neural network was fully-connected and was trained on the orbital elements (OEs) of real/simulated planetesimals, while the second was a 1-dimensional convolutional neural network that was trained on the position Cartesian coordinates of real/simulated planetesimals. Ultimately, the network trained on OEs had a better performance by identifying one-third of known potentially hazardous objects including the 3 asteroids with the highest chance of impact with Earth (2009 FD, 1999 RQ36, 1950 DA) as established by NASA's Monte Carlo based Sentry system.

Code site:
https://github.com/mrteetoe/Hazardous-Asteroid-Identifcation-via-NNs

Views: 802