Real-time optical boundary recognition for plasma configuration feedback control with a lightweight YOLO on EAST

The real-time control of plasma shape is a critical technical challenge for achieving sustained steady-state operation in future fusion reactors. This study proposes a lightweight YOLO-based method for real-time optical plasma boundary recognition to support the feedback control of plasma configuration on the EAST tokamak. The poloidal cross-sectional images of the plasma are captured by a high-speed camera in real time, enabling the system to perform pixel-level boundary recognition and reconstruction. The coordinates of the boundary control points extracted during this process are subsequently fed back to the plasma control system to regulate the plasma shape. Experiments conducted on the EAST tokamak in July 2025 successfully validated the proposed system. These experiments demonstrated the feasibility of utilizing a neural network model for the real-time control of the optical plasma boundary, providing a novel solution for plasma shape control in future fusion devices during steady-state discharges.