Exploring multi-planet system wasp-148 using n-body simulation and deep learning

Abstract

The search of exoplanets this day is focusing on finding planet with small mass or exoplanet with similar mass with Earth. One of the detection methods that sensitive enough to detect this kind of planet is transit timing variation method (TTV). This method detects planet using its perturbation to the transiting planet that could be observed by its small variation on the transit curve. From the TTV signal, the planets parameter such as mass and eccentricity could be estimated. One of the methods to determine exoplanet parameters from TTV signal is using N-body simulation to find the best planet configuration that could replicate the TTV signal from observation. However, this method is computationally expensive because there are so many combinations of planet configuration that must be simulated to find the best configuration. By using machine learning technique, we can make this simulation faster and more efficient by predicting the initial condition before hands. We combine photometric data from space and ground based observation for exoplanet system WASP-148 and use that data as input on our machine learning model. Then finally we can estimate the exoplanet parameters using N-body simulation