Search

The design of energy-efficient, high-performance, and reliable Convolutional Neural Network (CNN) accelerators involves significant challenges due to complex power and thermal management issues. This paper introduces SAfEPaTh, a novel system-level approach for accurately estimating power and temperature in tile-based CNN accelerators. By addressing both steady-state and transient-state scenarios, SAfEPaTh effectively captures the dynamic effects of pipeline bubbles in interlayer pipelines, utilizing real CNN workloads for comprehensive evaluation. Unlike traditional methods, it eliminates the need for circuit-level simulations or on-chip measurements. Our methodology leverages TANIA, a cutting-edge hybrid digital-analog tile-based accelerator featuring analog-in-memory computing cores alongside digital cores. Through rigorous simulation results using the ResNet18 model, we demonstrate SAfEPaTh's capability to accurately estimate power and temperature within 500 seconds, encompassing CNN model accelerator mapping exploration and detailed power and thermal estimations. This efficiency and accuracy make SAfEPaTh an invaluable tool for designers, enabling them to optimize performance while adhering to stringent power and thermal constraints. Furthermore, SAfEPaTh's adaptability extends its utility across various CNN models and accelerator architectures, underscoring its broad applicability in the field. This study contributes significantly to the advancement of energy-efficient and reliable CNN accelerator designs, addressing critical challenges in dynamic power and thermal management.

Estimating the State of Health (SOH) of batteries is crucial for ensuring the reliable operation of battery systems. Since there is no practical way to instantaneously measure it at run time, a model is required for its estimation. Recently, several data-driven SOH models have been proposed, whose accuracy heavily relies on the quality of the datasets used for their training. Since these datasets are obtained from measurements, they are limited in the variety of the charge/discharge profiles. To address this scarcity issue, we propose generating datasets by simulating a traditional battery model (e.g., a circuit-equivalent one). The primary advantage of this approach is the ability to use a simulatable battery model to evaluate a potentially infinite number of workload profiles for training the data-driven model. Furthermore, this general concept can be applied using any simulatable battery model, providing a fine spectrum of accuracy/complexity tradeoffs. Our results indicate that using simulated data achieves reasonable accuracy in SOH estimation, with a 7.2% error relative to the simulated model, in exchange for a 27X memory reduction and a =2000X speedup., Comment: 6 pages, 5 figures, conference paper at the 2023 IEEE/ACM ISLPED

Modern smartwatches often include photoplethysmographic (PPG) sensors to measure heartbeats or blood pressure through complex algorithms that fuse PPG data with other signals. In this work, we propose a collaborative inference approach that uses both a smartwatch and a connected smartphone to maximize the performance of heart rate (HR) tracking while also maximizing the smartwatch's battery life. In particular, we first analyze the trade-offs between running on-device HR tracking or offloading the work to the mobile. Then, thanks to an additional step to evaluate the difficulty of the upcoming HR prediction, we demonstrate that we can smartly manage the workload between smartwatch and smartphone, maintaining a low mean absolute error (MAE) while reducing energy consumption. We benchmark our approach on a custom smartwatch prototype, including the STM32WB55 MCU and Bluetooth Low-Energy (BLE) communication, and a Raspberry Pi3 as a proxy for the smartphone. With our Collaborative Heart Rate Inference System (CHRIS), we obtain a set of Pareto-optimal configurations demonstrating the same MAE as State-of-Art (SoA) algorithms while consuming less energy. For instance, we can achieve approximately the same MAE of TimePPG-Small (5.54 BPM MAE vs. 5.60 BPM MAE) while reducing the energy by 2.03x, with a configuration that offloads 80\% of the predictions to the phone. Furthermore, accepting a performance degradation to 7.16 BPM of MAE, we can achieve an energy consumption of 179 uJ per prediction, 3.03x less than running TimePPG-Small on the smartwatch, and 1.82x less than streaming all the input data to the phone., Comment: Published at 2023 Design, Automation \& Test in Europe Conference \& Exhibition (DATE)

Ultra-low-resolution Infrared (IR) array sensors offer a low-cost, energy-efficient, and privacy-preserving solution for people counting, with applications such as occupancy monitoring. Previous work has shown that Deep Learning (DL) can yield superior performance on this task. However, the literature was missing an extensive comparative analysis of various efficient DL architectures for IR array-based people counting, that considers not only their accuracy, but also the cost of deploying them on memory- and energy-constrained Internet of Things (IoT) edge nodes. In this work, we address this need by comparing 6 different DL architectures on a novel dataset composed of IR images collected from a commercial 8x8 array, which we made openly available. With a wide architectural exploration of each model type, we obtain a rich set of Pareto-optimal solutions, spanning cross-validated balanced accuracy scores in the 55.70-82.70% range. When deployed on a commercial Microcontroller (MCU) by STMicroelectronics, the STM32L4A6ZG, these models occupy 0.41-9.28kB of memory, and require 1.10-7.74ms per inference, while consuming 17.18-120.43 $\mu$J of energy. Our models are significantly more accurate than a previous deterministic method (up to +39.9%), while being up to 3.53x faster and more energy efficient. Further, our models' accuracy is comparable to state-of-the-art DL solutions on similar resolution sensors, despite a much lower complexity. All our models enable continuous, real-time inference on a MCU-based IoT node, with years of autonomous operation without battery recharging., Comment: This article has been accepted for publication in IEEE Internet of Things Journal; Fixed typos

Neural Architecture Search (NAS) is quickly becoming the go-to approach to optimize the structure of Deep Learning (DL) models for complex tasks such as Image Classification or Object Detection. However, many other relevant applications of DL, especially at the edge, are based on time-series processing and require models with unique features, for which NAS is less explored. This work focuses in particular on Temporal Convolutional Networks (TCNs), a convolutional model for time-series processing that has recently emerged as a promising alternative to more complex recurrent architectures. We propose the first NAS tool that explicitly targets the optimization of the most peculiar architectural parameters of TCNs, namely dilation, receptive-field and number of features in each layer. The proposed approach searches for networks that offer good trade-offs between accuracy and number of parameters/operations, enabling an efficient deployment on embedded platforms. We test the proposed NAS on four real-world, edge-relevant tasks, involving audio and bio-signals. Results show that, starting from a single seed network, our method is capable of obtaining a rich collection of Pareto optimal architectures, among which we obtain models with the same accuracy as the seed, and 15.9-152x fewer parameters. Compared to three state-of-the-art NAS tools, ProxylessNAS, MorphNet and FBNetV2, our method explores a larger search space for TCNs (up to 10^12x) and obtains superior solutions, while requiring low GPU memory and search time. We deploy our NAS outputs on two distinct edge devices, the multicore GreenWaves Technology GAP8 IoT processor and the single-core STMicroelectronics STM32H7 microcontroller. With respect to the state-of-the-art hand-tuned models, we reduce latency and energy of up to 5.5x and 3.8x on the two targets respectively, without any accuracy loss., Comment: Accepted for publication at the IEEE Transactions on Computers

The widespread adoption of Electric Vehicles (EVs) is limited by their reliance on batteries with presently low energy and power densities compared to liquid fuels and are subject to aging and performance deterioration over time. For this reason, monitoring the battery State Of Charge (SOC) and State Of Health (SOH) during the EV lifetime is a very relevant problem. This work proposes a battery digital twin structure designed to accurately reflect battery dynamics at the run time. To ensure a high degree of correctness concerning non-linear phenomena, the digital twin relies on data-driven models trained on traces of battery evolution over time: a SOH model, repeatedly executed to estimate the degradation of maximum battery capacity, and a SOC model, retrained periodically to reflect the impact of aging. The proposed digital twin structure will be exemplified on a public dataset to motivate its adoption and prove its effectiveness, with high accuracy and inference and retraining times compatible with onboard execution., Comment: Accepted as a conference paper at the 2022 IEEE International Conference on Omni-Layer Intelligent Systems (COINS)

Low-resolution infrared (IR) array sensors offer a low-cost, low-power, and privacy-preserving alternative to optical cameras and smartphones/wearables for social distance monitoring in indoor spaces, permitting the recognition of basic shapes, without revealing the personal details of individuals. In this work, we demonstrate that an accurate detection of social distance violations can be achieved processing the raw output of a 8x8 IR array sensor with a small-sized Convolutional Neural Network (CNN). Furthermore, the CNN can be executed directly on a Microcontroller (MCU)-based sensor node. With results on a newly collected open dataset, we show that our best CNN achieves 86.3% balanced accuracy, significantly outperforming the 61% achieved by a state-of-the-art deterministic algorithm. Changing the architectural parameters of the CNN, we obtain a rich Pareto set of models, spanning 70.5-86.3% accuracy and 0.18-75k parameters. Deployed on a STM32L476RG MCU, these models have a latency of 0.73-5.33ms, with an energy consumption per inference of 9.38-68.57{\mu}J., Comment: Accepted as a conference paper at the 2022 IEEE International Symposium on Circuits and Systems (ISCAS)

Collaborative Inference (CI) optimizes the latency and energy consumption of deep learning inference through the inter-operation of edge and cloud devices. Albeit beneficial for other tasks, CI has never been applied to the sequence- to-sequence mapping problem at the heart of Neural Machine Translation (NMT). In this work, we address the specific issues of collaborative NMT, such as estimating the latency required to generate the (unknown) output sequence, and show how existing CI methods can be adapted to these applications. Our experiments show that CI can reduce the latency of NMT by up to 44% compared to a non-collaborative approach., Comment: Accepted as a conference paper at the 2022 IEEE International Symposium on Circuits and Systems (ISCAS)

The spontaneous muonium to antimuonium conversion is one of the interesting charged lepton flavor violation processes. It serves as a clear indication of new physics and plays an important role in constraining the parameter space beyond Standard Model. MACE is a proposed experiment to probe such a phenomenon and expected to enhance the sensitivity to the conversion probability by more than two orders of magnitude from the current best upper constraint obtained by the PSI experiment two decades ago. Recent developments in the theoretical and experimental aspects to search for such a rare process are summarized., Comment: 13 pages, 10 figures, 3 tables, contribution to Snowmass 2021

Short bunch proton beams are of great significance for the applications of white neutron beams and muon beams. The accelerator complex of China Spallation Neutron Source (CSNS) was designed to support the applications mainly based on neutron scattering techniques where the proton pulse length is not very sensitive. Some theoretical and experimental studies have been performed to see if one can extract a short-bunch proton beam by bunch rotation from the rapid cycling synchrotron (RCS) at CSNS. The experimental results at RCS have evidently displayed the bunch lengthening and rotation process, which demonstrates the effectiveness of this method even with a very short available time for the RF gymnastic processes and a high-intensity beam. With a beam power of 50 kW and normal longitudinal emittance at the injection, the proton beam with a bunch length of about 53% with respect to the one in the normal operation mode was obtained and transported to the spallation target. With a reduced longitudinal emittance at injection and the beam power of 30 kW, the shortest extraction bunch length obtained is about 26% of the one in the normal operation mode. Different machine settings have also been tested to show the impact of the desynchronization between the RF and magnetic fields, the influence of the non-adiabatic risetime and the adiabatic decay time of the RF voltage on the extraction bunch length. The experimental results are well consistent with the theoretical and simulated ones. It is interesting to observe that space charge has a beneficial effect on the bunch lengthening which will result in a shorter bunch at the extraction with the later bunch rotation. The controlled desynchronization method between the RF and magnetic fields in an RCS was also proven successful., Comment: 19 pages, 14 figures

The ORBIT code has been widely used for beam dynamics simulations including injection and acceleration in high-intensity hadron synchrotrons. When the ORBIT's 1D longitudinal tracking was employed for the acceleration process in CSNS/RCS, the longitudinal emittance in eV-s was found decreasing substantially during acceleration, though the adiabatic condition is still met during this process. This is against the Liouville theorem that predicts the preservation of the emittance during acceleration. The recent machine study in the accelerator and the simulations with a self-made code demonstrate that the longitudinal emittance is almost invariant, which further indicates that the ORBIT longitudinal tracking might be incorrect. A detailed check-over in the ORBIT code source finds that the longitudinal finite difference equation used in the code is erroneous when applied to an acceleration process. The new code format PyORBIT has the same problem. After the small secondary factor is included in the code, ORBIT can produce results keeping the longitudinal emittance invariant. This paper presents some details about the study., Comment: 3 pages

Differential and angle-integrated cross sections for the $^{10}$B($n, {\alpha}$)$^{7}$Li, $^{10}$B($n, {\alpha}$$_{0}$)$^{7}$Li and $^{10}$B($n, {\alpha}$$_{1}$)$^{7}$Li$^{*}$ reactions have been measured at CSNS Back-n white neutron source. Two enriched (90%) $^{10}$B samples 5.0 cm in diameter and ~85.0 ${\mu}$g/cm$^{2}$ in thickness each with an aluminum backing were prepared, and back-to-back mounted at the sample holder. The charged particles were detected using the silicon-detector array of the Light-charged Particle Detector Array (LPDA) system. The neutron energy E$_{n}$ was determined by TOF (time-of-flight) method, and the valid ${\alpha}$ events were extracted from the E$_{n}$-Amplitude two-dimensional spectrum. With 15 silicon detectors, the differential cross sections of ${\alpha}$-particles were measured from 19.2{\deg} to 160.8{\deg}. Fitted with the Legendre polynomial series, the ($n, {\alpha}$) cross sections were obtained through integration. The absolute cross sections were normalized using the standard cross sections of the $^{10}$B($n, {\alpha}$)$^{7}$Li reaction in the 0.3 - 0.5 MeV neutron energy region. The measurement neutron energy range for the $^{10}$B($n, {\alpha}$)$^{7}$Li reaction is 1.0 eV $\le$ En < 2.5 MeV (67 energy points), and for the $^{10}$B($n, {\alpha}$$_{0}$)$^{7}$Li and $^{10}$B($n, {\alpha}$$_{1}$)$^{7}$Li$^{*}$ reactions is 1.0 eV $\le$ En < 1.0 MeV (59 energy points). The present results have been analyzed by the resonance reaction mechanism and the level structure of the $^{11}$B compound system, and compared with existing measurements and evaluations., Comment: 19 pages, 34 figures, to be published in Chinese Physics C

The 200-MeV electron linac of the National Synchrotron Radiation Laboratory (NSRL) located in Hefei is one of the earliest high-energy electron linear accelerators in China. The electrons are accelerated to 200 MeV by five acceleration tubes and are collimated by scrapers. The scraper aperture is smaller than the acceleration tube one, so some electrons hit the materials when passing through them. These lost electrons cause induced radioactivity mainly due to bremsstrahlung and photonuclear reaction. This paper describes a study of induced radioactivity for the NSRL Linac using FLUKA simulations and gamma-spectroscopy. The measurements showed that electrons were lost mainly at the scraper. So the induced radioactivity of the NSRL Linac is mainly produced here. The radionuclide types were simulated using the FLUKA Monte Carlo code and the results were compared against measurements made with a High Purity Germanium (HPGe) gamma spectrometer. The NSRL linac had been retired because of upgrading last year. The removed components were used to study induced radioactivity. The radionuclides confirmed by the measurement are: $^{57}$Ni, $^{52}$Mn, $^{51}$Cr, $^{58}$Co, $^{56}$Co, $^{57}$Co, $^{54}$Mn, $^{60}$Co and $^{22}$Na, the first eight nuclides of which are predicted by FLUKA simulation. The research will provide the theoretical basis for the similar accelerator decommissioning plan, and is significant for accelerator structure design, material selection and radiation protection design., Comment: 6 pages, 5 figures, Submitted to the Chinese Physics C