Accelerated Charged Particle Tracking with Graph Neural Networks on FPGAs
University of California San Diego（加州大学圣地亚哥分校）；Princeton University（普林斯顿大学）；University of Illinois at Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）；Fermi National Accelerator Laboratory（费米国家加速器实验室）；Massachusetts Institute of Technology（麻省理工学院）；Institute of Physics Belgrade（贝尔格莱德物理研究所）；European Organization for Nuclear Research (CERN)（欧洲核子研究组织（CERN））；California Institute of Technology（加州理工学院）；Purdue University（普渡大学）；HawkEye360；University of Illinois at Chicago（伊利诺伊大学芝加哥分校）
Aneesh Heintz, Vesal Razavimaleki, Javier Duarte, Gage DeZoort, Isobel Ojalvo, Savannah Thais, Markus Atkinson, Mark Neubauer, Lindsey Gray, Sergo Jindariani, N… 展开作者
Aneesh Heintz, Vesal Razavimaleki, Javier Duarte, Gage DeZoort, Isobel Ojalvo, Savannah Thais, Markus Atkinson, Mark Neubauer, Lindsey Gray, Sergo Jindariani, Nhan Tran, Philip Harris, Dylan Rankin, Thea Aarrestad, Vladimir Loncar, Maurizio Pierini, Sioni Summers, Jennifer Ngadiuba, Mia Liu, Edward Kreinar, Zhenbin Wu
备注：
8 pages, 4 figures, To appear in Third Workshop on Machine Learning and the Physical Sciences (NeurIPS 2020)
We develop and study FPGA implementations of algorithms for charged particle tracking based on graph neural networks. The two complementary FPGA designs are based on OpenCL, a framework for writing programs that execute across heterogeneous platforms, and hls4ml, a high-level-synthesis-based compiler for neural network to firmware conversion. We evaluate and compare the resource usage, latency, and tracking performance of our implementations based on a benchmark dataset. We find a considerable speedup over CPU-based execution is possible, potentially enabling such algorithms to be used effectively in future computing workflows and the FPGA-based Level-1 trigger at the CERN Large Hadron Collider.
Low Latency Edge Classification GNN for Particle Trajectory Tracking on FPGAs
National Yang Ming Chiao Tung University（国立阳明交通大学）；University of California San Diego（加利福尼亚大学圣迭戈分校）；University of Washington（华盛顿大学）；University of Illinois at Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）
University of Pennsylvania（宾夕法尼亚大学）；University of California San Diego（加州大学圣地亚哥分校）；National Yang Ming Chiao Tung University（国立阳明交通大学）；University of Illinois at Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）；Princeton University（普林斯顿大学）；University of Washington（华盛顿大学）
The determination of charged particle trajectories in collisions at the CERN Large Hadron Collider (LHC) is an important but challenging problem, especially in the high interaction density conditions expected during the future high-luminosity phase of the LHC (HL-LHC). Graph neural networks (GNNs) are a type of geometric deep learning algorithm that has successfully been applied to this task by embedding tracker data as a graph -- nodes represent hits, while edges represent possible track segments -- and classifying the edges as true or fake track segments. However, their study in hardware- or…
Hardware-Accelerated Event-Graph Neural Networks for Low-Latency Time-Series Classification on SoC FPGA
Graduate School of Science and Technology（科学与技术研究生院）；Embedded Vision Systems Group, Computer Vision Laboratory（嵌入式视觉系统组，计算机视觉实验室）；AGH University of Krakow（克拉科夫AGH大学）；CEA-List, Université Grenoble Alpes（CEA-List，格勒诺布尔阿尔卑斯大学）
University of Tokyo（东京大学）；Institute of Physics Belgrade（贝尔格莱德物理研究所）；European Organization for Nuclear Research (CERN)（欧洲核子研究中心（CERN））；California Institute of Technology（加州理工学院）；Columbia University（哥伦比亚大学）；University of California San Diego（加州大学圣迭戈分校）；Massachusetts Institute of Technology（麻省理工学院）；Northwestern University（西北大学）；Fermi National Accelerator Laboratory（费米国家加速器实验室）；HawkEye360；University of Illinois at Chicago（伊利诺伊大学芝加哥分校）
Graph neural networks have been shown to achieve excellent performance for several crucial tasks in particle physics, such as charged particle tracking, jet tagging, and clustering. An important domain for the application of these networks is the FGPA-based first layer of real-time data filtering at the CERN Large Hadron Collider, which has strict latency and resource constraints. We discuss how to design distance-weighted graph networks that can be executed with a latency of less than 1$μ\mathrm{s}$ on an FPGA. To do so, we consider a representative task associated to particle reconstruction …
The determination of charged particle trajectories (tracking) in collisions at the CERN Large Hadron Collider (LHC) is one of the most important aspects for event reconstruction at hadron colliders. This is especially true in the high conditions expected during the future high-luminosity phase of the LHC (HL-LHC) where the number of interactions per beam crossing will increase by a factor of five. Deep learning algorithms have been successfully applied to this task for offline applications. However, their study in hardware-based trigger applications has been limited . In this paper, we study d…
A novel FPGA based online tracking algorithm for helix track reconstruction in a solenoidal field, developed for the PANDA spectrometer, is described. Employing the Straw Tube Tracker detector with 4636 straw tubes, the algorithm includes a complex track finder, and a track fitter. Implemented in VHDL, the algorithm is tested on a Xilinx Virtex-4 FX60 FPGA chip with different types of events, at different event rates. A processing time of 7 $μ$s per event for an average of 6 charged tracks is obtained. The momentum resolution is about 3\% (4\%) for $p_t$ ($p_z$) at 1 GeV/c. Comparing to the al…
New heterogeneous computing paradigms on dedicated hardware with increased parallelization, such as Field Programmable Gate Arrays (FPGAs), offer exciting solutions with large potential gains. The growing applications of machine learning algorithms in particle physics for simulation, reconstruction, and analysis are naturally deployed on such platforms. We demonstrate that the acceleration of machine learning inference as a web service represents a heterogeneous computing solution for particle physics experiments that potentially requires minimal modification to the current computing model. As…
During the High Luminosity LHC, the CMS detector will need charged particle tracking at the hardware trigger level to maintain a manageable trigger rate and achieve its physics goals. The tracklet approach is a track-finding algorithm based on a road-search algorithm that has been implemented on commercially available FPGA technology. The tracklet algorithm has achieved high performance in track-finding and completes tracking within 3.4 $μ$s on a Xilinx Virtex-7 FPGA. An overview of the algorithm and its implementation on an FPGA is given, results are shown from a demonstrator test stand and s…
In Run-3, beginning in 2022, the LHCb software trigger will start reconstructing events at the LHC average crossing rate of 30 MHz. Within the upgraded DAQ system, LHCb established a testbed for new heterogeneous computing solutions for real-time event reconstruction, in view of future runs at even higher luminosities. One such solution is a highly-parallelized custom tracking processor ("Artificial Retina"), implemented in state of the art FPGA devices connected by fast serial links. We describe the status of a realistic prototype for the reconstruction of pixel tracking detectors that will r…
The high instantaneous luminosities expected following the upgrade of the Large Hadron Collider (LHC) to the High Luminosity LHC (HL-LHC) pose major experimental challenges for the CMS experiment. A central component to allow efficient operation under these conditions is the reconstruction of charged particle trajectories and their inclusion in the hardware-based trigger system. There are many challenges involved in achieving this: a large input data rate of about 20--40 Tb/s; processing a new batch of input data every 25 ns, each consisting of about 15,000 precise position measurements and ro…
Finding tracks downstream of the magnet at the earliest LHCb trigger level is not part of the baseline plan of the upgrade trigger, on account of the significant CPU time required to execute the search. Many long-lived particles, such as $K^0_S$ and strange baryons, decay after the vertex track detector, so that their reconstruction efficiency is limited. We present a study of the performance of a future innovative real-time tracking system based on FPGAs, developed within a R\&D effort in the context of the LHCb Upgrade Ib (LHC Run~4), dedicated to the reconstruction of the particles downstre…
In this work, we present a new approach for fast tracking on multiwire proportional chambers with neural networks. The tracking networks are developed and adapted for the first-level trigger at hadron collider experiments. We use Monte Carlo samples generated by Geant4 with a custom muon chamber, which resembles part of the thin gap chambers from the ATLAS experiment, for training and performance evaluations. The chamber has a total of seven gas gaps, where the first and last gas gaps are displaced by ~1.5 m. Each gas gap has 50 channels with a size of 18-20 mm. Two neural network models are d…
Low latency optical-based mode tracking with machine learning deployed on FPGAs on a tokamak
Columbia University（哥伦比亚大学）；Fermi National Accelerator Laboratory（费米国家加速器实验室）；Lehigh University（利哈伊大学）；Northwestern University（西北大学）；Drexel University（德雷塞尔大学）
A Large Ion Collider Experiment (ALICE) is one of the four major experiments conducted at the CERN Large Hadron Collider (LHC). The ALICE detector is currently undergoing an upgrade for the upcoming Run 3 at the LHC. The new Inner Tracking System (ITS) sub-detector is part of this upgrade. The front-end electronics of the ITS is composed by 192 Readout Units, installed in a radiation environment. Single Event Upsets (SEUs) in the SRAM-based Xilinx Kintex Ultrascale FPGAs used in the ITS readout represent a real concern. To clear SEUs affecting the Kintex configuration memory, a secondary Flash…
This work reports a cost-effective, simple front-end readout and DAQ of a RPC-based muon scattering tomography system under construction. The Time-over-threshold property of NINO ASICs has been exploited to achieve precise tracking by extraction of position information. The use of a low-cost FPGA for event selection and TOT measurement has been demonstrated for a prototype RPC which has been triggered by a cosmic ray hodoscope of three plastic scintillators. The relation between input pulse width and TOT has been established by analyzing the waveforms of the RPC signals. The TOT profile at dif…
The Compact Muon Solenoid (CMS) experiment at the Large Hadron Collider (LHC) is designed to study a wide range of high energy physics phenomena. It employs a large all-silicon tracker within a 3.8 T magnetic solenoid, which allows precise measurements of transverse momentum ($p_\mathrm{T}$) and vertex position. This tracking detector will be upgraded to coincide with the operation of the High-Luminosity LHC, which will provide luminosities of up to $7.5\,\times\,10^{35}\,\mathrm{cm^{-2}s^{-1}}$ to CMS, or 200 collisions per 25 ns bunch crossing. This new tracker must maintain the nominal phys…
With the High Luminosity LHC upgrades, incorporating tracking information into the CMS Level-1 trigger becomes necessary in order to maintain a manageable trigger rate and good trigger performance e.g. to retain thresholds for electroweak physics. The main challenges Level-1 track finding faces are the large data throughput from the detector at a collision rate of 40 MHz and a 4 microsecond latency budget to reconstruct charged particle tracks with sufficiently low transverse momentum to be used in the Level-1 trigger decision. Dedicated all-FPGA hardware systems with time-multiplexed architec…
ALICE (A Large Heavy Ion Experiment) is one of the four major experiments at the Large Hadron Collider (LHC) at CERN. The High Level Trigger (HLT) is an online compute farm which reconstructs events measured by the ALICE detector in real-time. The most compute-intensive part is the reconstruction of particle trajectories called tracking and the most important detector for tracking is the Time Projection Chamber (TPC). The HLT uses a GPU-accelerated algorithm for TPC tracking that is based on the Cellular Automaton principle and on the Kalman filter. The GPU tracking has been running in 24/7 op…
Physics collisions at 13 TeV are expected at the LHC with an average of 40-50 proton-proton collisions per bunch crossing under nominal conditions. Tracking at trigger level is an essential tool to control the rate in high-pileup conditions while maintaining a good efficiency for relevant physics processes. The Fast TracKer is an integral part of the trigger upgrade for the ATLAS detector. For every event passing the Level-1 trigger (at a maximum rate of 100 kHz) the FTK receives data from all the channels of the silicon detectors, providing tracking information to the High Level Trigger in or…
The MuPix Telescope is a particle tracking telescope, optimized for tracking low momentum particles and high rates. It is based on the novel High-Voltage Monolithic Active Pixel Sensors (HV-MAPS), designed for the Mu3e tracking detector. The telescope represents a first application of the HV-MAPS technology and also serves as test bed of the Mu3e readout chain. The telescope consists of up to eight layers of the newest prototypes, the MuPix7 sensors, which send data self-triggered via fast serial links to FPGAs, where the data is time-ordered and sent to the PC. A particle hit rate of 1 MHz pe…
We present a fast general-purpose algorithm for high-throughput clustering of data "with a two dimensional organization". The algorithm is designed to be implemented with FPGAs or custom electronics. The key feature is a processing time that scales linearly with the amount of data to be processed. This means that clustering can be performed in pipeline with the readout, without suffering from combinatorial delays due to looping multiple times through all the data. This feature makes this algorithm especially well suited for problems where the data has high density, e.g. in the case of tracking…
