Measured acceleration · Physics

Deep reinforcement learning helps avoid tearing instabilities in fusion plasma

Deep reinforcement learning helps avoid tearing instabilities in fusion plasma: measured or workflow-level change in a research process.

Summary

The Nature paper reports a reinforcement-learning controller trained with a dynamic model of future tearing-instability likelihood. In DIII-D tokamak experiments, the controller kept estimated tearing likelihood below a threshold while tracking a stable operating path under challenging conditions.

AI role

AI helps choose, automate, optimize, speed up, or scale part of the research workflow in physics.

Narrative role

This belongs in the timeline because it records a measurable or operational change in a research workflow, search process, simulation, or experiment loop.

Caveat

Process gains do not automatically imply final scientific, clinical, industrial, or field-level impact.