Scheduling ESPRESSO follow-up of TESS Targets. I. Myopic versus non-myopic sampling

Radial-velocity follow-up of stars harbouring transiting planets detected by TESS is expected to require very large amounts of expensive telescope time in the next few years. Therefore, scheduling strategies should be implemented to maximize the amount of information gathered about the target planetary systems. We present one random scheduler and two types of uniform-in-phase schedulers: one myopic, which selects targets one-at-a-time, and one non-myopic that efficiently explores all the possible combinations between stars to be observed and available time slots. We compare these strategies with respect to the bias, accuracy and precision achieved in recovering the mass and orbital parameters of transiting and non-transiting planets from the mock radial-velocity follow-up of a sample of 50 TESS target stars, with simulated planetary systems containing at least one transiting planet with a radius below $4R_{\oplus}$. For each system and strategy, 10 radial-velocity datasets were generated and analysed using a fully Bayesian methodology. We find the myopic strategies lead to a biased estimation of the order of 5\% of the mass of the transiting exoplanets while the non-myopic scheduler is able to provide an unbiased (<1\%) measurement of the masses while keeping the relative accuracy and precision around 16\% and 23\% respectively. The number of non-transiting planets detected is similar for all the strategies considered, although the random scheduler leads to less biased and more accurate estimates for their mass and orbital parameters, possibly due to a higher mean number of scheduled radial-velocities for the datasets associated with non-transiting planets detections.