Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “latency”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 217 records · Page 12
All sourcesNASA NTRSDOE OSTI
NASA NTRS2002

A Bayesian Approach to Estimating Coupling Between Neural Components: Evaluation of the Multiple Component, Event-Related Potential (mcERP) Algorithm

Accurate measurement of single-trial responses is key to a definitive use of complex electromagnetic and hemodynamic measurements in the investigation of brain dynamics. We developed the multiple component, Event-Related Potential (mcERP) approach to single-trial response estimation. To improve our resolution of dynamic interactions between neuronal ensembles located in different layers within a cortical region and/or in different cortical regions. The mcERP model assets that multiple components defined as stereotypic waveforms comprise the stimulus-evoked response and that these components may vary in amplitude and latency from trial to trial. Maximum a posteriori (MAP) solutions for the model are obtained by iterating a set of equations derived from the posterior probability. Our first goal was to use the ANTWERP algorithm to analyze interactions (specifically latency and amplitude correlation) between responses in different layers within a cortical region. Thus, we evaluated the model by applying the algorithm to synthetic data containing two correlated local components and one independent far-field component. Three cases were considered: the local components were correlated by an interaction in their single-trial amplitudes, by an interaction in their single-trial latencies, or by an interaction in both amplitude and latency. We then analyzed the accuracy with which the algorithm estimated the component waveshapes and the single-trial parameters as a function of the linearity of each of these relationships. Extensions of these analyses to real data are discussed as well as ongoing work to incorporate more detailed prior information.

Shah, Ankoor S.↗
NASA NTRS2002

Attentional sensitivity and asymmetries of vertical saccade generation in monkey

The first goal of this study was to systematically document asymmetries in vertical saccade generation. We found that visually guided upward saccades have not only shorter latencies, but higher peak velocities, shorter durations and smaller errors. The second goal was to identify possible mechanisms underlying the asymmetry in vertical saccade latencies. Based on a recent model of saccade generation, three stages of saccade generation were investigated using specific behavioral paradigms: attention shift to a visual target (CUED paradigm), initiation of saccade generation (GAP paradigm) and release of the motor command to execute the saccade (DELAY paradigm). Our results suggest that initiation of a saccade (or "ocular disengagement") and its motor release contribute little to the asymmetry in vertical saccade latency. However, analysis of saccades made in the CUED paradigm indicated that it took less time to shift attention to a target in the upper visual field than to a target in the lower visual field. These data suggest that higher attentional sensitivity to targets in the upper visual field may contribute to shorter latencies of upward saccades.

NASA Discipline Neuroscience↗
NASA NTRS2001

Effect of stance width on multidirectional postural responses

The effect of stance width on postural responses to 12 different directions of surface translations was examined. Postural responses were characterized by recording 11 lower limb and trunk muscles, body kinematics, and forces exerted under each foot of 7 healthy subjects while they were subjected to horizontal surface translations in 12 different, randomly presented directions. A quasi-static approach of force analysis was done, examining force integrals in three different epochs (background, passive, and active periods). The latency and amplitude of muscle responses were quantified for each direction, and muscle tuning curves were used to determine the spatial activation patterns for each muscle. The results demonstrate that the horizontal force constraint exerted at the ground was lessened in the wide, compared with narrow, stance for humans, a similar finding to that reported by Macpherson for cats. Despite more trunk displacement in narrow stance, there were no significant changes in body center of mass (CoM) displacement due to large changes in center of pressure (CoP), especially in response to lateral translations. Electromyographic (EMG) magnitude decreased for all directions in wide stance, particularly for the more proximal muscles, whereas latencies remained the same from narrow to wide stance. Equilibrium control in narrow stance was more of an active postural strategy that included regulating the loading/unloading of the limbs and the direction of horizontal force vectors. In wide stance, equilibrium control relied more on an increase in passive stiffness resulting from changes in limb geometry. The selective latency modulation of the proximal muscles with translation direction suggests that the trunk was being actively controlled in all directions. The similar EMG latencies for both narrow and wide stance, with modulation of only the muscle activation magnitude as stance width changed, suggest that the same postural synergy was only slightly modified for a change in stance width. Nevertheless, the magnitude of the trunk displacement, as well as of CoP displacement, was modified based on the degree of passive stiffness in the musculoskeletal system, which increased with stance width. The change from a more passive to an active horizontal force constraint, to larger EMG magnitudes especially in the trunk muscles and larger trunk and CoP excursions in narrow stance are consistent with a more effortful response for equilibrium control in narrow stance to perturbations in all directions.

Non-NASA Center↗
NASA NTRS1998

Relationship between selected orientation rest frame, circular vection and space motion sickness

Space motion sickness (SMS) and spatial orientation and motion perception disturbances occur in 70-80% of astronauts. People select "rest frames" to create the subjective sense of spatial orientation. In microgravity, the astronaut's rest frame may be based on visual scene polarity cues and on the internal head and body z axis (vertical body axis). The data reported here address the following question: Can an astronaut's orientation rest frame be related and described by other variables including circular vection response latencies and space motion sickness? The astronaut's microgravity spatial orientation rest frames were determined from inflight and postflight verbal reports. Circular vection responses were elicited by rotating a virtual room continuously at 35 degrees/s in pitch, roll and yaw with respect to the astronaut. Latency to the onset of vection was recorded from the time the crew member opened their eyes to the onset of vection. The astronauts who used visual cues exhibited significantly shorter vection latencies than those who used internal z axis cues. A negative binomial regression model was used to represent the observed total SMS symptom scores for each subject for each flight day. Orientation reference type had a significant effect, resulting in an estimated three-fold increase in the expected motion sickness score on flight day 1 for astronauts who used visual cues. The results demonstrate meaningful classification of astronauts' rest frames and their relationships to sensitivity to circular vection and SMS. Thus, it may be possible to use vection latencies to predict SMS severity and duration.

short duration↗
NASA NTRS2009

Combined MODIS/AMSR-E SST Composites for Regional Weather Applications

Recent applications of a high resolution MODlS composite SST product have clearly shown the importance of developing high-resolution SST data sets for coastal applications and modeling. In general, coupling between the oceans and atmospheres has been closely linked to SST gradients and fronts, indicating a need for high resolution SSTs, specifically in the areas of large gradients associated with coastal regions. Thus an accurate determination of SST gradients has become critical for determining the appropriate air-sea coupling and the influence on ocean modeling. Recent research is focused on improving the accuracy and spatial coverage of the current operational MODIS SST composite product provided by the Short-term Prediction Research and Transition (SPORT) project and distributed to the community. GHRSST-PP MODlS data and microwave AMSR-E data are being combined to produce composite data sets for both the West Coast and East Coast of the United States, including the Gulf of Mexico. The use of 1 km MODIS data has explicit advantages over other SST products including its global coverage and high resolution. The AMSR-E data will reduce the latency of the composites. A strategy for utilizing the error characteristics contained in the GHRSST data has been developed. This strategy will include using the error characteristics directly to calculate weights in the SST composites, uncertainty maps based on the composite biases and RMS errors, and latency products calculated in the compositing process. Recent accomplishments include the development of an enhanced compositing approach based on the error-weighted combination of recent clear MODIS SST values, where the error contributions come from measurement error, potential cloud contamination, and data latency sources. Future plans call for the inclusion of AMSR-E SST values with appropriate weights based upon measurement accuracy, MODIS-AMSR-E SST bias, and latency.

Jedlovec, Gary↗
NASA NTRS2012

Desert RATS 2011: Near-Earth Asteroid Human Exploration Operations

The Desert Research and Technology Studies (D-RATS) 2011 field test involved the planning and execution of a series of exploration scenarios under operational conditions similar to those that would be expected during a human exploration mission to a near-Earth asteroid (NEA). The focus was on understanding the operations tempo during simulated NEA exploration and the implications of communications latency and limited data bandwidth. Anchoring technologies and sampling techniques were not evaluated due to the immaturity of those technologies and the inability to meaningfully test them at D-RATS. Reduced gravity analogs and simulations are being used to fully evaluate Multi-Mission Space Exploration Vehicle (MMSEV) and extravehicular (EVA) operations and interactions in near-weightlessness at a NEA as part of NASA s integrated analogs program. Hypotheses were tested by planning and performing a series of 1-day simulated exploration excursions comparing test conditions all of which involved a single Deep Space Habitat (DSH) and either zero, one, or two MMSEVs; three or four crewmembers; one of two different communications bandwidths; and a 100-second roundtrip communications latency between the field site and Houston. Excursions were executed at the Black Point Lava Flow test site with a Mission Control Center and Science Support Room at Johnson Space Center (JSC) being operated with 100-second roundtrip communication latency to the field. Crews were composed of astronauts and professional field geologists and teams of Mission Operations, Science, and Education & Public Outreach (EPO) experts also supported the mission simulations each day. Data were collected separately from the Crew, Mission Operations, Science, and EPO teams to assess the test conditions from multiple perspectives. For the operations tested, data indicates practically significant benefits may be realized by including at least one MMSEV and by including 4 versus 3 crewmembers in the NEA exploration architecture as measured by increased Scientific Data Quality, EVA Exploration Time, Capability Assessment Ratings, and Overall Acceptability ratings by Crew, Mission Operations, Science, and Education & Public Outreach teams. A combination of text and voice was used to effectively communicate over the 100-second roundtrip communications latency and increased communication bandwidth yielded a small but practically significant improvement in Overall Acceptability as rated by the Science team, although the impact of bandwidth on scientific strategic planning was not assessed. No effect of increased bandwidth was observed with respect to Crew, Mission Operations, or EPO team ratings of Overall Acceptability.

Abercromby, Andrew↗
NASA NTRS2013

Coarse-Grain Bandwidth Estimation Scheme for Large-Scale Network

A large-scale network that supports a large number of users can have an aggregate data rate of hundreds of Mbps at any time. High-fidelity simulation of a large-scale network might be too complicated and memory-intensive for typical commercial-off-the-shelf (COTS) tools. Unlike a large commercial wide-area-network (WAN) that shares diverse network resources among diverse users and has a complex topology that requires routing mechanism and flow control, the ground communication links of a space network operate under the assumption of a guaranteed dedicated bandwidth allocation between specific sparse endpoints in a star-like topology. This work solved the network design problem of estimating the bandwidths of a ground network architecture option that offer different service classes to meet the latency requirements of different user data types. In this work, a top-down analysis and simulation approach was created to size the bandwidths of a store-and-forward network for a given network topology, a mission traffic scenario, and a set of data types with different latency requirements. These techniques were used to estimate the WAN bandwidths of the ground links for different architecture options of the proposed Integrated Space Communication and Navigation (SCaN) Network. A new analytical approach, called the "leveling scheme," was developed to model the store-and-forward mechanism of the network data flow. The term "leveling" refers to the spreading of data across a longer time horizon without violating the corresponding latency requirement of the data type. Two versions of the leveling scheme were developed: 1. A straightforward version that simply spreads the data of each data type across the time horizon and doesn't take into account the interactions among data types within a pass, or between data types across overlapping passes at a network node, and is inherently sub-optimal. 2. Two-state Markov leveling scheme that takes into account the second order behavior of the store-and-forward mechanism, and the interactions among data types within a pass. The novelty of this approach lies in the modeling of the store-and-forward mechanism of each network node. The term store-and-forward refers to the data traffic regulation technique in which data is sent to an intermediate network node where they are temporarily stored and sent at a later time to the destination node or to another intermediate node. Store-and-forward can be applied to both space-based networks that have intermittent connectivity, and ground-based networks with deterministic connectivity. For groundbased networks, the store-and-forward mechanism is used to regulate the network data flow and link resource utilization such that the user data types can be delivered to their destination nodes without violating their respective latency requirements.

Cheung, Kar-Ming↗
NASA NTRS2013

Understanding the Effects of Spaceflight on Head-trunk Coordination during Walking and Obstacle Avoidance

Prolonged exposure to spaceflight conditions results in a battery of physiological changes, some of which contribute to sensorimotor and neurovestibular deficits. Upon return to Earth, functional performance changes are tested using the Functional Task Test (FTT), which includes an obstacle course to observe post‐flight balance and postural stability, specifically during turning. Aims: To quantify changes in movement strategies during turning events by observing the latency between head‐andtrunk coordinated movement. Hypothesis: It is hypothesized that subjects experiencing neurovestibular adaptations will exhibit head‐to‐trunk locking ('en bloc' movement) during turning, exhibited by a decrease in latency between head and trunk movement. Sample: FTT data samples were collected from Shuttle and ISS missions. Samples were analyzed three times pre exposure, immediately post‐exposure (0 or 1 day post) and 2‐to‐3 times during recovery from the microgravity environment. Methods: Two 3D inertial measurements units (XSens MTx) were attached to subjects, one on the head and one on the upper back. This study focused primarily on the yaw movements about the subject's center of rotation. Time differences (latency) between head and trunk movement were calculated at two points: the first turn (Fturn) to enter the obstacle course (approximately 90⁰ turn) and averaged across a slalom obstacle portion, consisting of three turns (approximately three 90⁰ turns). Results: Preliminary analysis of the data shows a trend toward decreasing head‐to‐trunk movement latency during post‐flight ambulation, after reintroduction to Earth gravity in Shuttle and ISS astronauts. Conclusion: It is clear that changes in movement strategies are adopted during exposure to the microgravity environment and upon reintroduction to a gravity environment. Some subjects exhibit symptoms of neurovestibular neuropathy ('en bloc movement') that may impact their ability to perform post‐flight functional tasks.

Madansingh, S.↗
NASA NTRS2014

Understanding the Effects of Spaceflight on Head-trunk Coordination during Walking and Obstacle Avoidance

Prolonged exposure to spaceflight conditions results in a battery of physiological changes, some of which contribute to sensorimotor and neurovestibular deficits. Upon return to Earth, functional performance changes are tested using the Functional Task Test (FTT), which includes an obstacle course to observe post‐flight balance and postural stability, specifically during turning. The goal of this study was to quantify changes in movement strategies during turning events by observing the latency between head‐and‐trunk coordinated movements. It was hypothesized that subjects experiencing neurovestibular adaptations would exhibit head‐to‐trunk locking ('en bloc' movement) during turning, exhibited by a decrease in latency between head and trunk movement. FTT data samples were collected from ISS missions. Samples were analyzed three times pre‐exposure, immediately postexposure (1 day post) and 2‐to‐3 times during recovery from the microgravity environment. Two 3D inertial measurements units (XSens MTx) were attached to subjects, one on the head and one on the upper back. This study focused primarily on the yaw movements about the subject's center of rotation. Time differences (latency) between head and trunk movement were calculated at two points on the obstacle course: the first turn to enter the obstacle course (approximately 90⁰ turn) and averaged across a slalom obstacle portion, consisting of three turns (approximately three 90⁰ turns). Preliminary analysis of the data shows a trend toward decreasing head‐to‐trunk movement latency during postflight ambulation in slalom turning after reintroduction to Earth gravity in ISS astronauts. It is clear that changes in movement strategies are adopted during exposure to the microgravity environment and upon reintroduction to a gravity environment. Most ISS subjects exhibit symptoms of neurovestibular changes ('en bloc head and trunk movement) which may impact their ability to perform post‐flight functional tasks.

Madansingh, S.↗
NASA NTRS2014

Understanding the Effects of Spaceflight on Head-trunk Coordination During Walking and Obstacle Avoidance

Prolonged exposure to spaceflight conditions results in a battery of physiological changes, some of which contribute to sensorimotor and neurovestibular deficits. Upon return to Earth, functional performance changes are tested using the Functional Task Test (FTT), which includes an obstacle course to observe post-flight balance and postural stability, specifically during turning. The goal of this study was to quantify changes in movement strategies during turning events by observing the latency between head-and-trunk coordinated movements. It was hypothesized that subjects experiencing neurovestibular adaptations would exhibit head-to-trunk locking ('en bloc' movement) during turning, exhibited by a decrease in latency between head and trunk movement. FTT data samples were collected from 13 ISS astronauts and 26 male 70-day head down tilt bed rest subjects, including bed rest controls (10 BRC) and bed rest exercisers (16 BRE). Samples were analyzed three times pre-exposure, immediately post-exposure (0 or 1 day post) and 2-to-3 times during recovery from the unloading environment. Two 3D inertial measurements units (XSens MTx) were attached to subjects, one on the head and one on the upper back. This study focused primarily on the yaw movements about the subject's center of rotation. Time differences (latency) between head and trunk movement were averaged across a slalom obstacle portion, consisting of three turns (approximately three 60⁰ turns). All participants were grouped as 'decreaser' or 'increaser,' relating to their change in head-to-trunk movement latency between pre- and post- environmental adaptation measures. Space flight unloading (ISS) showed a bimodal response between the 'increaser' and 'decreaser' group, while both bed rest control (BRC) and bed rest exercise (BRE) populations showed increased preference towards a 'decreaser' categorization, displaying greater head-trunk locking. It is clear that changes in movement strategies are adopted during exposure to an unloading environment. These results further the understanding of vestibular-somatosensory convergence and support the use of bed rest as an exclusionary model to better understand sensorimotor changes in space flight.

Madansingh, S.↗
NASA NTRS2015

Measured Response for UAS Integration into the National Airspace System

The measured response (MR) is the response time of aircraft to Air Traffic Controller (ATCo) commands and clearances. The overall MR can be broken up into several components, including the pilot verbal latencies (MR1), the time between the end of an ATCo clearance and the beginning of the pilots read back, and the execution initiation latency (MR2), the time between the end of the ATCos clearance and when the pilot begins to execute a maneuver. The MR is a crucial concern for the integration of Unmanned Aircraft Systems (UAS) into the National Airspace System (NAS) due to potentially greater latencies stemming from remote pilot communication and command execution. As a result, it is important to quantify what latencies in verbal responding and command execution are acceptable for safe and efficient operations in the NAS. The present studies begin to address these issues in a series of four simulations supported by the UAS Integration into the NAS program.

UAS in the NAS↗
NASA NTRSDate not supplied

Zero-Trust Architecture for Autonomous Edge Computing

We are at the apex of an aviation revolution where autonomy will play a central role in enabling complex, multi-agent systems to communicate, interact, and collaborate on a myriad of applications spanning autonomous swarms to wild-fire management. Autonomy is not an absolute but rather a spectrum ranging from a system requiring significant human intervention to one requiring little to none [1]. For example, the extreme, in the case of an autonomous aircraft, is one that operates independently in the airspace interacting with all other elements (air traffic controllers, other pilots) as if it were a human pilot. Critical to this vision is an architecture that enables autonomous agents to interact with minimal latency. Edge computing is an emerging architecture where compute and storage is pushed to the ‘edge’ of the network in order to minimize the round-trip time from agent to resource thereby mitigating the latency associated with cloud-only based approaches. Additionally, services can generate massive amounts of data (e.g., video feeds), which may require analysis in near real-time. Moving this data to the cloud for further processing may not be feasible due to latency, bandwidth, and cost. Privacy, security, and reliability can also be improved by edge computing architectures. However, this geo-distributed and dynamic* architecture complicates the establishment of unambiguous network security boundaries and can lead to vulnerabilities including man in the middle attacks, replay attacks, physical security breaches of edge nodes, signal interception, etc. This motivates the need for zero-trust architectures [2–4] which de-emphasize the notion of static network perimeters and, as the name implies, do not instill any innate trust in any particular agent. It is required that all agents must be authorized and approved in every transaction. In this paper, we present a zero-trust architecture suitable for edge-computing applications that demand significant low-latency, security, privacy, and reliability.

zero trust↗
DOE OSTI2021

Design Space Exploration of Emerging Memory Technologies for Machine Learning Applications

Memory design space exploration methods study memory systems’ performances and limitations before implementation. The computer memory design space has grown exponentially because of the enormous growth of memory types, memory controllers, and application software. Computer simulators are commonly used for memory design space exploration. However, complex memory simulations take an enormous amount of time. Hence, in this paper, we proposed a machine learning-based design space exploration method for dynamic random-access memory and non-volatile memory systems. We applied our method to the CosmoGAN and LeNet applications to predict the following six memory response parameters: (i) bandwidth, (ii) power, (iii) average latency, (iv) average total latency, (v) memory reads, and (vi) memory writes. Our experimental results show that machine learning models can predict memory response parameter values faster than simulations. We used support vector machine, random forest, and gradient boosting machine learning models. We observed that the support vector machine provides better performance for bandwidth, average latency, and average total latency. The random forest model works better for memory reads and writes. The gradient boosting model provides superior prediction performance for power. We provide a detailed discussion on learning curve characteristics, error analysis, and memory type recommendation.

Hasan, S M Shamimul↗
DOE OSTI2020

On the Marriage of Asynchronous Many Task Runtimes and Big Data: A Glance

The rise of the accelerator-based architectures and reconfigurable computing have showcased the weakness of software stack toolchains that still maintain a static view of the hardware instead of relying on a symbiotic relationship between static (e.g., compilers) and dynamic tools (e.g., runtimes). In the past decades, this need has given rise to adaptive runtimes with increasingly finer computational tasks. These finer tasks help to take advantage of the hardware by switching out when a long latency operation is encountered (because of the deeper memory hierarchies and new memory technologies that might target streaming instead of random access), thus trading off idle time for unrelated work. Examples of these finer task runtimes are Asynchronous Many Task (AMT) runtimes, in which highly efficient computational graphs run on a variety of hardware. Due to its inherent latency tolerant characteristics, Latency-sensitive applications, such as Graph Analytics and Big Data can effectively use these runtimes. This paper aims to present an example of how the careful design of an AMT can exploit the hardware substrate when faced with high latency applications such as the ones given in the Big Data domain. Moreover, with its introspection and adaptive capabilities, we aim to show the power of these runtimes when facing the changing requirements of the application workloads. We use the Performance Open Community Runtime (P-OCR) as our vehicle to demonstrate the concepts presented here.

adaptive runtime, big data analysis↗
DOE OSTI2025

LUNA: LUT-Based Neural Architecture for Fast and Low-Cost Qubit Readout

Qubit readout is a critical operation in quantum computing systems, which maps the analog response of qubits into discrete classical states. Deep neural networks (DNNs) have recently emerged as a promising solution to improve readout accuracy . Prior hardware implementations of DNN-based readout are resource-intensive and suffer from high inference latency, limiting their practical use in low-latency decoding and quantum error correction (QEC) loops. This paper proposes LUNA, a fast and efficient superconducting qubit readout accelerator that combines low-cost integrator-based preprocessing with Look-Up Table (LUT) based neural networks for classification. The architecture uses simple integrators for dimensionality reduction with minimal hardware overhead, and employs LogicNets (DNNs synthesized into LUT logic) to drastically reduce resource usage while enabling ultra-low-latency inference. We integrate this with a differential evolution based exploration and optimization framework to identify high-quality design points. Our results show up to a 10.95x reduction in area and 30% lower latency with little to no loss in fidelity compared to the state-of-the-art. LUNA enables scalable, low-footprint, and high-speed qubit readout, supporting the development of larger and more reliable quantum computing systems.

Farooq, M. A. [Arizona State U., Tempe]↗
DOE OSTI2020

Cyber‐physical system testbed for power system monitoring and wide‐area control verification

The electric power system is intrinsically a cyber‐physical system (CPS) with power flowing inthe physical system and information flowing in the cyber‐network. Testbeds arecrucial for understanding the cyber‐physical interactions and provideenvironments for prototyping novel applications. This study proposes afour‐layer architecture for CPS testbeds with emphases on communication networkemulation and networked physical components. A configurable software‐definednetwork is employed to bridge physical components with wide‐area applicationsfor closed‐loop control. In order to distribute physically coupled devices intomultiple software simulations, this study proposes a data broker setup based ona distributed messaging environment to achieve low‐latency data streaming. Thedecoupled design with data streaming allows for building testbed components asmodules and running them in a distributed manner. Case studies verify the databroker setup for low‐latency sensing and actuation, as well as the communicationemulation setup for the desired network latency. Also illustrated is a replayattack scenario using synchrophasors in the Western Electricity CoordinatingCouncil (WECC) 181‐bus system for demonstrating the closed‐loop cyber‐physicalsimulation capability of the testbed.

Cui, Hantao↗
DOE OSTI2021

Fast convolutional neural networks on FPGAs with hls4ml

We introduce an automated tool for deploying ultra low-latency, low-power deep neural networks with convolutional layers on field-programmable gate arrays (FPGAs). By extending the hls4ml library, we demonstrate an inference latency of 5 µs using convolutional architectures, targeting microsecond latency applications like those at the CERN Large Hadron Collider. Considering benchmark models trained on the Street View House Numbers Dataset, we demonstrate various methods for model compression in order to fit the computational constraints of a typical FPGA device used in trigger and data acquisition systems of particle detectors. In particular, we discuss pruning and quantization-aware training, and demonstrate how resource utilization can be significantly reduced with little to no loss in model accuracy. We show that the FPGA critical resource consumption can be reduced by 97% with zero loss in model accuracy, and by 99% when tolerating a 6% accuracy degradation.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2022

Lightweight jet reconstruction and identification as an object detection task

We apply object detection techniques based on deep convolutional blocks to end-to-end jet identification and reconstruction tasks encountered at the CERN large hadron collider (LHC). Collision events produced at the LHC and represented as an image composed of calorimeter and tracker cells are given as an input to a Single Shot Detection network. The algorithm, named PFJet-SSD performs simultaneous localization, classification and regression tasks to cluster jets and reconstruct their features. This all-in-one single feed-forward pass gives advantages in terms of execution time and an improved accuracy w.r.t. traditional rule-based methods. A further gain is obtained from network slimming, homogeneous quantization, and optimized runtime for meeting memory and latency constraints of a typical real-time processing environment. We experiment with 8-bit and ternary quantization, benchmarking their accuracy and inference latency against a single-precision floating-point. We show that the ternary network closely matches the performance of its full-precision equivalent and outperforms the state-of-the-art rule-based algorithm. Finally, we report the inference latency on different hardware platforms and discuss future applications.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗