Teaching a controller to tame a tokamak
A fusion plasma can lose control in milliseconds. Can a learned controller keep it stable better than hand-tuned rules?
In one paragraph
A simulation study compares a reinforcement learning controller for tokamak magnetic coils against a conventional hand-tuned feedback system. In simulated shots, the learned controller cut disruptions from 12 percent to 4 percent under a range of disturbances. The study is simulation only and has not yet been tested on a real machine.
Fusion Physics & Space
Briefing reviewed by an editor before publication. The paper belongs to its authors; this page summarises it and links to the original. Reviewed by Dr. G. Aaltonen-Diallo. Paper licence: CC BY 4.0. Corrections: support@papersays.com.