Artificial intelligence for the EChO mission planning tool

Abstract

The Exoplanet Characterisation Observatory (EChO) has as its main goal the measurement of atmospheres of transiting planets. This requires the observation of two types of events: primary and secondary eclipses. In order to yield measurements of sufficient Signal-to-Noise Ratio to fulfil the mission objectives, the events of each exoplanet have to be observed several times. In addition, several criteria have to be considered to carry out each observation, such as the exoplanet visibility, its event duration, and no overlapping with other tasks. It is expected that a suitable mission plan increases the efficiency of telescope operation, which will represent an important benefit in terms of scientific return and operational costs. Nevertheless, to obtain a long term mission plan becomes unaffordable for human planners due to the complexity of computing the huge number of possible combinations for finding an optimum solution. In this contribution we present a long term mission planning tool based on Genetic Algorithms, which are focused on solving optimization problems such as the planning of several tasks. Specifically, the proposed tool finds a solution that highly optimizes the defined objectives, which are based on the maximization of the time spent on scientific observations and the scientific return (e.g., the coverage of the mission survey). The results obtained on the large experimental set up support that the proposed scheduler technology is robust and can function in a variety of scenarios, offering a competitive performance which does not depend on the collection of exoplanets to be observed. Specifically, the results show that, with the proposed tool, EChO uses 94% of the available time of the mission, so the amount of downtime is small, and it completes 98% of the targets.