Real-time plasma state estimation runs on FPGA hardware
Real-Time Plasma State Prediction via FPGA-Accelerated Quantized Recurrent Probabilistic Neural Networks
Machine Learning
Summary
Estimating the state of plasma inside a Tokamak fusion device quickly is important but hard because control systems need very fast data. The authors present a complete process to run a special kind of neural network (called a recurrent probabilistic neural network) on an FPGA chip to do this estimation. They shrink the model and train it with techniques to make it run efficiently in hardware. Their design runs in less than 10 microseconds per step and fits within the hardware limits, making it suitable for use in controlling plasma in real time.
What this means in practice
- •For tokamak control engineers: Implement real-time plasma state estimation on FPGA devices to improve control system responsiveness within strict latency limits.
- •For embedded systems developers: Deploy quantized recurrent neural networks on FPGA platforms for low-latency, resource-efficient time series prediction tasks.
Authors
Daniel Gaytan-Villarreal, Aiken Xie, Tu Pham, Rohit Sonker, Chiara Amendola, Matteo Cremonesi, Cong Hao, Jeff Schneider
Abstract
Real time plasma state estimation for control of Tokamak devices are challenging due to the stringent latency requirements of the plasma control system (PCS). We present an end-to-end workflow for deploying a recurrent probabilistic neural network (RPNN) on FPGA hardware. We combine architecture size reduction with quantization-aware training via QKeras. The model is then synthesized using hls4ml, targeting a Xilinx Alveo U50 device. We report a design that fits comfortably within all four resource budgets (DSP, LUT, FF, BRAM) at deterministic sub-10~$μ$s single-timestep latency, meeting the requirements for real-time inference inside a model-predictive-control-style plasma control loop.