Constraints Checking

The constraint checking process is the first filter applied by SchedulerCore when generating candidate windows. A window that does not pass a constraint is discarded without computing its score.

Filtering pipeline

For each active profile and each time interval of the night:

flowchart TD
    A["Time interval [t_start, t_end]"] --> B["is_active = True\nFacility assigned"]
    B -->|fail| X["✗ Discard"]
    B --> C["Altitude > min_altitude\nAirmass < max_airmass"]
    C -->|fail| X
    C --> D["Lunar separation > min_moon_distance\nLunar illumination < max_moon_illumination"]
    D -->|fail| X
    D --> E["Night type compatible\nTransparency sufficient"]
    E -->|fail| X
    E --> F["Seeing < min_seeing\n(if forecast available)"]
    F -->|fail| X
    F --> G["Category checks\nPERIODIC: within transit window\nTRANSIENT: before fade_date"]
    G -->|fail| X
    G --> H["✓ Candidate Window → ScoringStrategy"]

Implementation

The checks are performed inside SchedulerCore.generate_candidate_windows(), which iterates over a time grid sampled every N minutes (configurable). For each time sample:

1. AstronomyEngine.get_target_altaz_coords() — Computes altitude.
2. AstronomyEngine.get_airmass() — Computes airmass.
3. AstronomyEngine.get_moon_separation() — Distance to the Moon.
4. AstronomyEngine.get_moon_phase() — Lunar illumination.
5. Comparison against the thresholds in SchedulingProfile.

Category-specific constraints

PERIODIC

Periodic targets are only candidates within their transit window (transit_start → transit_end). Outside this window, PeriodicStrategy returns 0.0 directly:

if time < self.transit_start or time > self.transit_end:
    return 0.0

TRANSIENT

Transient targets have a fade_date after which they are no longer candidates:

if time > profile.fade_date:
    return 0.0

Adding custom constraints

To add an extra constraint, extend the generate_candidate_windows() method in SchedulerCore or create a subclass and override the filtering method. The strategy system also allows integrating custom logic directly inside a scoring function.