Speaker
Description
Gran Telescopio Canarias (GTC), the largest optical and infrared telescope in operation, is a complex cyber-physical infrastructure where a segmented 10.4 m mirror, cryogenic instruments, a moving enclosure and thousands of sensors must work together every night. A night lost to a fault is science that never comes back, and nineteen years of operation have left a vast record of incidents, logs and expertise that today lives largely in the memory of its engineers and operators.
Not every replica is a twin. A digital model describes the system, a digital shadow mirrors its state in real time, and a digital twin closes the loop and returns validated decisions to the physical asset. GTC currently operates a digital shadow: an Industry 4.0 platform built on open standards (IIoT sensing, FIWARE-based context management, a Delta Lake lakehouse) feeds a 3D replica of the telescope synchronised in real time. This talk presents the programme to climb the next step, a cognitive digital twin that remembers, diagnoses, anticipates and advises, with human supervision and deterministic safety layers at every stage.
The twin is the state and memory of a wider ecosystem of AI agents built on a common platform,several of them already in production or under test: retrieval over technical documentation and incident history, evidence based reconstruction of what happened around a fault, decision support for daily operations, and an orchestrator that turns procedures into executable and auditable workflows. AI recommends; the telescope control system executes.
We discuss the infrastructure choices behind it, in particular the balance between local and cloud AI: on premises inference for sovereignty, latency and cost, frontier models where they add value. We close with lessons learned and the roadmap ahead.