September 15, 2022 Master Thesis Open Access

Greta Heine

Citation Style Language JSON Export

{
  "abstract": "<p>Efficient reconstruction of charged particle trajectories is a crucial yet very difficult step&nbsp;in the analysis pipeline of high energy physics (HEP) experiments. Recent work has&nbsp;shown that graph neural networks (GNNs) are well suited for the pattern recognition&nbsp;task of track finding, where tracking detector hits can be naturally represented as nodes&nbsp;and particle track segments as edges. The interaction network (IN) GNN architecture&nbsp;provides computationally efficient edge classification of the high-dimensional and sparse&nbsp;tracker data, which is especially crucial for implementation in constrained computing&nbsp;environments such as field programmable gate arrays (FPGAs). This work describes the&nbsp;overall workflow for implementing and systematically analyzing an IN-based classification&nbsp;of track segments on FPGAs for the anti-Proton&nbsp;Annihilation at&nbsp;DArmstadt (PANDA)&nbsp;forward tracking system (FTS). This workflow includes data preprocessing, graph building, GNN-based edge classification and a series of FPGA implementation design studies&nbsp;concerning latency, resource utilization, and classification quality using the high level&nbsp;synthesis for machine learning (hls4ml) compiler. The presented final implementation of&nbsp;the GNN-based track segment classifier on a Xilinx Zynq&nbsp;provides an overall inference latency of about&nbsp;0.99&nbsp;&mu;s&nbsp;using about&nbsp;34 %&nbsp;of available digital signal processors (DSPs) and&nbsp;85 %&nbsp;of available lookup tables (LUTs). This work enables the acceleration of charged particle tracking on heterogeneous computational resources toward real-time track reconstruction for the PANDA experiment. The discussed methods and studies could be easily adapted and used in other HEP experiments for accelerated charged particle tracking.</p>", 
  "author": [
    {
      "family": "Greta Heine"
    }
  ], 
  "id": "22136", 
  "issued": {
    "date-parts": [
      [
        2022, 
        9, 
        15
      ]
    ]
  }, 
  "title": "A Fast AI-Based Track Reconstruction on FPGA for the PANDA Experiment", 
  "type": "thesis"
}