News for hackers. Stuff that matters.

A comment on Avoiding fusion plasma tearing instability with deep reinforcement learning.

Re:Avoiding fusion plasma tearing instability with deep rein... by chrisgp on Saturday March 02, 2024 @06:06PM (#39576630)
Why doesn't simpler control theory work here, like PID controllers?
[ Parent | on HN ]
Re:Avoiding fusion plasma tearing instability with deep rein... by theptip on Saturday March 02, 2024 @07:12PM (#39577070)
PID controllers are for linear or mildly-nonlinear systems. Plasma fluid dynamics are about as nonlinear as it gets.
[ Parent | on HN ]
Re:Avoiding fusion plasma tearing instability with deep rein... by itishappy on Saturday March 02, 2024 @06:18PM (#39576715)
> ... Calculating tearing stability requires massive computational simulations based on resistive magnetohydrodynamics or gyrokinetics, which are not suitable for real-time stability prediction and control during experiments. This suggests the need for AI-accelerated real-time instability-avoidance techniques.
[ Parent | on HN ]