Integrated tokamak control using NSFsim simulator

NSFsim is an integrated modeling framework designed for tokamak scenario development, combining a free-boundary Grad-Shafranov equilibrium solver with a 1D transport model. NSFsim allows for self-consistent prediction of plasma evolution, coil currents, and induced currents in passive structures. The NSFsim-toolkit, a Python-based interface, manages data exchange between modules and external codes, supporting flexible integration and control-oriented workflows. Current capabilities include interfacing with heating and current drive models such as TRAVIS, and the TGLF code for turbulence-informed transport. Reinforcement learning tools for magnetic and kinetic control are embedded, supporting controller training via access to plasma states or synthetic diagnostics. Open access via FusionTwin.io provides both GUI and API support (Python, C++, MATLAB/Simulink), facilitating custom controller development and replay of experimental scenarios. NSFsim has been successfully used in experiments on the DIII-D tokamak, where machine learning-based controllers trained in simulation demonstrated robust magnetic shape control.