Creating Autonomous Scientific Stations for Occultations Using Machine-Learning and Smart Telescopes

Our team has developed and tested a trustworthy and efficient artificial intelligence to process the ever growing amount of occultation-related data provided by the 10,000 digital and smart telescopes part of the Unistellar network. ODNet is a Convolutional Neural Network for Asteroid Occultation Detection which detects reliably on raw CMOS data frames occultation events (precision of 91% & recall of 87% based on 66 observations) (cazeneuve et al. - submitted). Today the number of occultation events observed with the network is growing exponentially so this algorithm is currently used by our pipeline to alert on the possible presence of a positive occultation and initiate an accurate detection and characterization of this event.

We propose to expand this work to the concept of asteroid-tracking station, a small and autonomous observatory that would be able to scrutinize the sky in search of undocumented asteroid's occultation similarly to the Whipple mission concept.

With our preliminary estimation, we estimate that posting one station in a weather appropriate location would permit the characterization of known asteroids each year. Versatile, these stations could also be mobilized as part of larger campaigns, such as the characterizations of NASA targets (Buie, Marc W., et al. "Size and shape of (11351) Leucus from five occultations." The Planetary Science Journal 2.5 (2021): 202.) and help in planetary defense program (Marchis, F., et al. "The Contribution of the Unistellar Network for Planetary Defense: Empowering Crowd-sourcing Astronomy." 7th IAA Planetary Defense Conference. 2021.) similarly that what was done recently with recent Apophis occultation events observed in 2021 (Marchis, F. et al. "Citizen Science and Scientific Results from the World's Largest Network of Backyard Astronomers." 2021.).