AI can now control fusion plasma faster than humans can react
Princeton researchers have tested an AI system that can monitor and control fusion plasma in milliseconds, reacting far faster than a human operator. In one experiment, it predicted a damaging instability about 200 milliseconds before it appeared and adjusted the plasma to stop it from forming.
Researchers have developed a software framework called PACMAN that uses artificial intelligence to control fusion plasma faster than humans can react. Fusion systems generate particles hotter than the sun's core, which become unstable within mere milliseconds—far too quickly for human operators to respond. The PACMAN framework, created by researchers at the U.S. Department of Energy's Princeton Plasma Physics Laboratory and Princeton University, employs machine learning to make rapid decisions while maintaining strict safety controls.
The framework consists of four stages: collecting live measurements from the tokamak, checking for errors, estimating plasma behavior, determining necessary actions, and resolving conflicting instructions. PACMAN can predict plasma behavior in milliseconds, enabling the adjustment of systems like heating equipment, magnets, and gas injectors before instabilities disrupt the fusion reaction.
In experiments using the DIII-D National Fusion Facility tokamak, PACMAN successfully predicted a tearing mode instability 200 milliseconds in advance, allowing the plasma to be altered proactively. Additionally, PACMAN coordinated all six gyrotrons heating the plasma simultaneously to meet complex targets, a task previously unachievable.
Despite faster fusion experiments with PACMAN, human operators remain responsible for setting the system's objectives.
Written by urgent.news from ScienceDaily's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
