Fast Machine Learning for Science Conference 2026

Real-Time Plasma State Estimation via FPGA-Accelerated Quantized Recurrent Probabilistic Neural Networks

Not scheduled

12m

Presentation Contributed Talks

Speakers

Aiken Xie (Columbia University) Jose Daniel Gaytan Villarreal (Carnegie-Mellon University (US))

Description

We evaluate the FPGA deployment of a quantized recurrent probabilistic neural network (RPNN) online state estimator for real-time tokamak control. To navigate the strict resource and latency limitations of fusion environments, we detail a workflow utilizing quantization-aware training (QAT) via the Brevitas framework, followed by translation into an efficient FPGA implementation using a directive-based C++ High-Level Synthesis (HLS) pipeline. We report hardware resource utilization alongside low, deterministic inference latencies while preserving the model's baseline state estimation accuracy. We demonstrate that the FPGA implementation's deterministic timing meets the latency requirements of operating as part of the plasma control system (PCS), enabling model predictive control (MPC)-style lookahead frameworks for the control of fusion plasmas.