Qwen helps National Astronomical Observatories research team build scientific-grade telescope simulation system

QwenWork News

Recently, QwenWork assisted a research team from the National Astronomical Observatories in building a large-aperture, research-grade telescope digital simulation system—completed in just three days at a cost of less than 1,000 yuan.

In contrast to previous practices, similar systems typically required outsourcing to software vendors, taking about three months to develop and costing tens of thousands of yuan. QwenWork has reduced the barrier to building such systems by two orders of magnitude.

01

Why Build an Observation Simulation System?

The universe is vast and boundless; the starlight swept by telescopes originates from distant deep space.

Transient events such as supernova explosions and gamma-ray bursts are fleeting flashes in the long history of the cosmos, yet they serve as critical windows for studying cosmic evolution and extreme physical processes.

These events often occur suddenly and vanish quickly—capturing them successfully requires more than just high-performance telescopes. What's also needed is a "brain" that is perceptive, intelligent, and capable of autonomous decision-making.

Given the scarcity of research-grade telescope resources, real equipment cannot afford the cost of repeated trial and error. Simulating observation workflows in advance within a simulation system has thus become an effective way to minimize actual losses.

Therefore, before conducting actual observations, research teams must first construct a "simulated telescope" to fully emulate the entire observation process. Only after verifying the feasibility of their plans can they deploy real instruments to gaze into the deep sky.

QwenWork was honored to serve as the underlying support platform for this simulation system.

02

How Does the Simulation System Work?

QwenWork helped the research team at the National Astronomical Observatories build a digital simulation system for a large-aperture, research-grade telescope, integrating information on telescope components, sensor statuses, and observational environmental conditions into standardized MCP interfaces for real-time monitoring and invocation by intelligent agents (Agents).

This is akin to converting each part of the telescope into a row of "universal sockets," enabling plug-and-play functionality for Agents—allowing them not only to monitor status in real time but also to remotely control devices.

Two specialized models operate collaboratively under Agent orchestration:

Cross-survey alignment timing model: Identifies early-stage supernova candidates from multiple public sky survey datasets, answering the question of "what to observe";

Short-term local weather prediction model: Assesses real-time observing conditions, guiding "how to observe."

Ultimately, the Agent can autonomously complete a closed-loop process of "status perception → task planning → plan generation → workflow validation," based on scientific priority, real-time telescope status, and target visibility windows.

03

Related Agents Successfully Flagged 8 Early Supernova Candidates

This Agent framework is not limited to simulation—it has already been integrated with the "Sitian" Pathfinder and "Sitian" prototype telescopes.

After researchers submit observation requests, the Agent generates observation plans by considering target priorities, weather conditions, and equipment status. It invokes control modules through various MCP interfaces and feeds results back to researchers. After mission completion, it further optimizes future plans based on performance data and expert feedback, distilling effective experiences into reusable skills.

To date, these Agents have successfully flagged eight early-stage supernova candidates—eight times capturing the universe’s fleeting "blinks" before the light faded, with two instances triggering actual observations when weather permitted.

This practice represents not only the construction of a simulation system for a single telescope but also an exploration into new organizational and operational models for astronomical research.

04

AI Is Truly Participating in Scientific Execution—Research Paradigms Are Shifting

Astronomical observation has evolved from human eyesight to AI-assisted execution: starting with manual operation, progressing to automated workflows with predefined procedures—where changes in weather or equipment still required human intervention. Today, with the growing integration of large models, Agents, scientific models, and telescope control systems, AI now possesses the ability to dynamically plan, invoke tools, validate strategies, and adjust approaches based on real-time environments and scientific objectives.

Meanwhile, the simulation system can repeatedly execute tasks in a virtual environment, continuously recording equipment states, environmental changes, decision-making processes, and outcomes—generating high-quality training data to develop more advanced telescope control models (such as VLA), laying the foundation for higher-level intelligent observations on real instruments in the future.

"AI is becoming increasingly embedded in the scientific research process. Researchers focus on asking questions and setting goals, while Agents take on parts of the observational execution, plan validation, and iterative tasks. This shows that AI’s role in science is expanding beyond auxiliary analysis into actual research execution." — Dr. Li Yuyang, expert at the National Astronomical Observatories' Artificial Intelligence Promotion Committee

Over four hundred years ago, Galileo used his homemade telescope to observe Jupiter’s four moons for the first time, providing key evidence for the heliocentric theory.

Four centuries later, AI helps researchers capture fleeting flashes from across the universe.

While AI has not replaced those who gaze at the stars, it has freed scientists from repetitive tasks, allowing them to refocus on their core mission: asking scientific questions.

Let Agents handle observation, verification, and iteration—reserving time for what matters most: discovery.

When the next beam of starlight from deep space reaches Earth, an artificial intelligence will already be waiting.

We dedicated to serving clients with professional DingTalk solutions. If you'd like to learn more about DingTalk platform applications, feel free to contact our online customer service or email at This email address is being protected from spambots. You need JavaScript enabled to view it.. With a skilled development and operations team and extensive market experience, we’re ready to deliver expert DingTalk services and solutions tailored to your needs!