...
- trim list of candidate fields to the ones that are above some cutoff airmass
- estimate the co-added depth for each one, compute their depth merit functions
- determine the (partial-credit) time since last observed for each onefield in each band, and compute the temporal merit function for each field/passband combination
- compute merit function for each field and passband, and calculate nominal sky merit function vs. t by looking forward until temporal merit hits 0.9. Store best merit within that interval.. Compute penalty for observing it now if better chance later, for each field and passband.
- Pick the top 150 fields for each passband and add up merit function for each band. Select the highest total as the starting filter for the night. That establishes the first hour's worth of observations
Adopt a block-wise approach. While taking data during a block of 150 pointings (which will take about an hour), remove them from the list of candidate fields by setting weights to zero, or whatever.
Decide whether we should change filters: Compute sum-of-merit for next-best 150 observations in current filter, as opposed to most important ~100 observations in other bands. This imposes the filter change penalty properly.
Decide on best order for observing most important fields. This is now just optimizing the sequence of 150 pointings.
...
Some references
LSST science book
...