PACMAN Artificial Intelligence Framework Developed to Control Nuclear Fusion Systems in Milliseconds

Researchers from the Princeton Plasma Physics Laboratory and Princeton University have developed the PACMAN artificial intelligence framework, capable of making critical millisecond-level decisions in nuclear fusion reactors while keeping human control intact.

AI-Powered Fusion Management

In nuclear fusion systems, particles hotter than the center of the sun can spin out of control in a timeframe too short for human operators to react. The newly developed software framework delegates these split-second decisions to artificial intelligence.

Working Structure of the PACMAN Framework

Known as PACMAN, the artificial intelligence framework consists of stages such as gathering real-time measurements, checking for errors, predicting plasma behavior through models, and enforcing hardware safety limits.

Successful Experiments at the DIII-D Facility

The research team successfully tested this system in five different experiments on the tokamak device at the DIII-D National Fusion Facility in San Diego, proving the system's flexibility and reliability.

Human Supervision and Safety Measures

PACMAN does not push humans out of the loop; the framework mandates hardware safety limits independently of artificial intelligence suggestions, and physicists regulate the oversight.