Speakers
Description
LEGEND operates an array of High-Purity Germanium (HPGe) detectors in combination with various integrated active veto systems immersed in a liquid argon environment, enabling an ultra-low radioactivity setting for optimal background discrimination. The scientific reach of the program and its ability to probe rare signals across a broad range of new physics can be further improved through implementation of a new AI-powered data acquisition (DAQ) pipeline, in comparison to the current pipeline which applies a traditional threshold-based hardware trigger coupled with offline data cleaning. The proposed work represents a unique opportunity to automate these efforts for the next-generation LEGEND-1000 experiment, which will produce petabyte-scale datasets over the course of its lifetime at the lowest energy thresholds.
A Genesis Mission Phase I effort, LEAP-FWD directly addresses these challenges by developing and demonstrating AI algorithms deployable on Field Programmable Gate Arrays (FPGAs) for real-time, low-latency event classification at the sensor edge. Building upon recent semi-supervised ML approaches that combine unsupervised affinity propagation with support vector machines, we will produce an advanced, hardware-optimized iteration for direct deployment on FPGAs towards operation within the LEGEND-200 experiment and the future LEGEND-1000 iteration. Success of this work will enable lower energy thresholds, reduced person-hours spent on data cleaning, and improved detection signal efficiency. A key part of this work is use of the PNNL-led high-level synthesis tool SODA, that enables a fully automatic path from AI/ML Python frameworks to efficient hardware implementations. These capabilities will lay the groundwork for an unsupervised spatiotemporal-aware anomaly detection algorithm that correlates signals across the entire detector array in real time.