Automation, operations & scheduling

How observing works

Users choose the number of unit telescopes and targets, the spectral resolution, and the cadence or time-criticality of their observations. Like the VLT, the TDT is a PI-driven facility: proposals are ranked by a central time allocation committee, and users ask for unit-hours, which can be used in any configuration. Large and multi-year programmes may be offered by the observatory, but the TDT is not just a survey instrument: flexibility and agility are central to the concept.

Agentic scheduling

Every TDT observation is effectively a hard target of opportunity, so scheduling conflicts are inevitable. The large number of units eases this, but full-array observations, to maximise depth or cadence, will be in high demand. To make the most of near-real-time transient detections, which may be the norm in the 2040s, the whole array must be reconfigurable within a minute or less. No facility in the world comes close to this level of scheduling complexity.

The TDT will therefore use an agentic approach, in which AI agents each manage one aspect of the facility, within strictly defined optimisation parameters:

• Telescope manager agents, one for each unit telescope.
• A scheduling agent, which allocates units to observations across the array.
• A data reduction agent, which processes the data as they arrive.
• A maintenance agent, which handles preventive maintenance.

For example, the maintenance agent may tell the manager agent of unit 52 that a component shows signs of wear and must be taken out of service. The scheduling agent then reschedules the full array, taking into account the time-criticality and depth constraints of upcoming observations.

From alert to spectrum

The TDT will integrate tightly with alert streams from time-domain surveys and multi-messenger triggers. Software will evaluate alerts in real time, assign priorities and allocate telescope resources, often within minutes or seconds of discovery.

• Autonomous scheduling, continuously re-optimised for weather, target visibility and evolving priorities.
• Real-time data reduction, producing calibrated spectra and light curves fast enough to inform same-night decisions.
• Feedback loops, in which early results can automatically trigger deeper or longer observations.
• Scalable data architectures, for continuous data streams from a hundred units, including photon-event streams from future photon-counting detectors.

Automation is not just about efficiency; it enables new observing strategies. The TDT can monitor sources adaptively, follow variability as it evolves, and coordinate a hundred telescopes without human bottlenecks. The result is not simply faster follow-up, but a new way of observing the dynamic Universe.