Search

This article proposes a bridgeless push–pull resonant ac–dc converter featuring naturally balanced switching loss at the primary side. The proposed converter employs a current-fed push–pull structure and two rectifier diodes at the primary side and a series resonance circuit at the secondary side. By incorporating two rectifier diodes, a diode bridge can be removed from the grid side of the proposed converter, and thereby that the number of components and primary-side conduction loss are reduced. One major advantage of the proposed converter is balanced loss distribution over one cycle of grid voltage, which guarantees high power transfer capability with enhanced overall switch utilization. The series-resonant circuit provides zero-current switching turn- off at the output diode, thereby reducing the reverse recovery loss. The operating principles and theoretical derivations of the converter are discussed in detail. The validity and performance of the proposed converter are verified using a prototype with 1.65 kW output power.

This work discusses the main aspects of vulnerability and degradation of accuracy of an image classification engine implemented into SRAM-based FPGAs under faults. The image classification engine is an all-convolutional neural-network (CNN) trained with a dataset of traffic sign recognition benchmark. The Caffe and Ristretto frameworks were used for CNN training and fine-tuning while the ZynqNet inference engine was adopted as hardware implementation on a Xilinx 28 nm SRAM-based FPGA. The CNN under test was generated using an evolutive approach based on genetic algorithm. The methodologies for qualifying this CNN under faults is presented and both heavy-ions accelerated irradiation and emulated fault injection were performed. To cross validate results from radiation and fault injection, different implementations of the same CNN were tested using reduced arithmetic precision and protection of user data by Hamming codes, in combination with configuration memory healing by the scrubbing mechanism available in Xilinx FPGA. Some of these alternative implementations increased significantly the mission time of the CNN, when compared to the original ZynqNet operating on 32 bits floating point number, and the experiment suggests areas for further improvements on the fault injection methodology in use.

Teachers’ beliefs, professional development, access to hardware/software, and instructional contexts all influence teachers’ use of technology (Ertmer et al., 2012; Gonzalez-Carriedo & Esprivalo Ha...

This paper presents a uniform optimal power sharing strategy to coordinate the wind turbines (WTs) in a wind farm (WF) to provide occasional and continuous frequency response (FR). The coordination of WTs is formulated as an optimisation problem that takes into account the WT dynamics and tries to reduce the long term loss of energy yield caused by the provision of FR. This is achieved by maximising the total kinetic energy of the WF over time while reducing wear and tear of WTs. The proposed optimal power sharing strategy relies on periodic communication between each WT and a WF controller. Local linear approximations are employed to predict the system behaviour and the solution of the optimisation problem is obtained using the proposed centralised and/or distributed algorithm. The distributed algorithm only requires one‐way communication between the WF controller and local WTs, reducing the communication overheads. Simulation studies are carried out on a WF model to demonstrate the effectiveness of the proposed strategy. The results show the strategy enables reduction of yield loss over previous methods while avoiding over torque operation during FR provision.

Marvell’s ThunderX2 has been the first Arm-based processor with deployments in large-scale HPC production systems, challenging the dominance that x86 processors had in the last decades. While x86 processors and its software stack have been characterized in detail, the behavior of Arm counterparts is not well known, limiting its adoption. This work methodically characterizes performance and power efficiency of the ThunderX2 running different HPC workloads compiled with two state-of-the-art compilers, GCC and Arm HPC Compiler. We study the maturity of available compilers and find that the Arm HPC Compiler is able to apply additional optimizations, resulting in better performance than GCC. In addition, we also compare both performance and power with respect to an Intel Skylake processor. Despite the faster single thread performance of Skylake, ThunderX2 is able to match performance on multi-threaded workloads due to its superior memory bandwidth. However, power efficiency of ThunderX2 is far from matching Skylake-based processors when AVX512 extensions are used.

This article evaluates the RISC-V Rocket processor embedded in a Commercial Off-The-Shelf (COTS) SRAM-based field-programmable gate array (FPGA) under heavy-ions-induced faults and emulation fault injection. We also analyze the efficiency of using mitigation techniques based on hardware redundancy and scrubbing. Results demonstrated an improvement of $3\times $ in the cross section when scrubbing and coarse grain triple modular redundancy are used. The Rocket processor presented analogous sensitivity to radiation effects as the state-of-the-art soft processors. Due to the complexity of the system-on-chip, not only the Rocket core but also its peripherals should be protected with proper solutions. Such solutions should address the specific vulnerabilities of each component to improve the overall system reliability while maintaining the trade-off with performance.

The increasing penetration level of plug-in electric vehicles (PEVs), as well as distributed generators (DGs), imposes significant challenges on power system planning. This work presents a coordinated approach for the planning of PEV charging facilities and DGs, including both the locations and the capacities, with the consideration of the transportation - power network couplings. First, the PEV charging demand is characterized by a temporal-SoC (State of Charge) analysis, and the DG generation uncertainties are modeled by K-means clustering using historical data. The M/M/s/N queuing model is used to formulate the dynamics of charging stations and then obtain the optimal station capacity, including the number of chargers and waiting spaces. Furthermore, the placement of charging stations is optimized by the Floyd Algorithm to minimize the total distance to obtain charging service. Finally, the sitting and sizing of DGs are optimized over multiple objectives, including active power losses, reactive power losses, and voltage deviation. The solution is evaluated through a case study of the IEEE 53-bus test feeder coupled with a 25-node transportation network. It is shown that the proposed solution enables the active and reactive power losses as well as the voltage deviation to be reduced by 37.6%, 44.3%, and 33.6%, respectively, after the optimal integration of PEV charging stations and DGs. The scalability and effectiveness of the solution are further validated in an IEEE 123-bus test feeder coupled with the same transportation network, and the result confirms the effectiveness and scalability of the proposed solution.

Memory-based Collaborative Filtering (CF) has been a widely used approach for personalised recommendation with considerable success in many applications. An important issue regarding memory-based CF lies in similarity computation: the sparsity of the rating matrix leads to similarity computations based on few co-rated items between users, resulting in high sensitive predictions. Additionally, the ‘sparse’ similarity computation has high computational cost, due to the dimensionality of the item space. In this paper, we pursue both these issues. We propose a new model to compute similarity by representing users (or items) through their distances to preselected users, named landmarks. Such user modelling allows the introduction of more ratings into similarity computations through transitive relations created by the landmarks. Unlike conventional memory-based CF, the proposal builds a new user space defined by distances to landmarks, avoiding sensitivity in similarity computations. Findings from our experiments show that the proposed modelling achieves better accuracy than the ‘sparse’ similarity representation in all tested datasets, and has also yielded competitive accuracy results against the compared model-based CF algorithms . Furthermore, the proposed implementation has beaten all compared methods in terms of computational performance, becoming a promising alternative to memory-based CF algorithms for large datasets.

The modular multilevel converter is the state-of-the-art topology for voltage source converter HVDC. Despite its advantages, this converter handles large internal low-frequency energy ripples, and the capacitance that supports these dynamics is a key design parameter that affects the operating region of the converter. Different strategies can be found in the literature to increase the feasible region of operation of the converter. Nevertheless, they are typically open loop in nature and use precalculated control references. This paper presents an alternative based on model predictive control that steers the system through optimal control trajectories that are calculated online. This provides feedback and corrective control action in real time. The predictive controller used for this purpose is presented and a linear time-varying approximation is used to reduce the computational burden of the algorithm. The feasible boundaries of the converter are sought and the final performance of the control algorithm is evaluated through detailed simulations using a switching model of the converter.

Comunicació presentada a INTERSPEECH 2021, celebrat del 30 d'agost al 3 de setembre de 2021 a Brno, Txèquia. The aim of this contribution is to automatically detect COVID19 patients by analysing the acoustic information embedded in coughs. COVID-19 affects the respiratory system, and, consequently, respiratory-related signals have the potential to contain salient information for the task at hand. We focus on analysing the spectrogram representations of cough samples with the aim to investigate whether COVID-19 alters the frequency content of these signals. Furthermore, this work also assesses the impact of gender in the automatic detection of COVID-19. To extract deep-learnt representations of the spectrograms, we compare the performance of a cough-specific, and a Resnet18 pre-trained Convolutional Neural Network (CNN). Additionally, our approach explores the use of contextual attention, so the model can learn to highlight the most relevant deep-learnt features extracted by the CNN. We conduct our experiments on the dataset released for the Cough Sound Track of the DICOVA 2021 Challenge. The best performance on the test set is obtained using the Resnet18 pre-trained CNN with contextual attention, which scored an Area Under the Curve (AUC) of 70.91 % at 80 % sensitivity. This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreements No. 826506 (sustAGE) and No. 770376 (TROMPA). Further funding has been received from the FI Predoctoral Grant 2018FI-B01015 from AGAUR, Generalitat de Catalunya.

The Galileo High Accuracy Service (HAS) is a new capability of the European Global Navigation Satellite System that is currently under development. The Galileo HAS will start providing satellite orbit and clock corrections (i.e. non-dispersive effects) and soon it will also correct dispersive effects such as inter-frequency biases and, in its full capability, ionospheric delay. We analyse here an ionospheric correction system based on the fast precise point positioning (Fast-PPP) and its potential application to the Galileo HAS. The aim of this contribution is to present some recent upgrades to the Fast-PPP model, with the emphasis on the model geometry and the data used. The results show the benefits of integer ambiguity resolution to obtain unambiguous carrier phase measurements as input to compute the Fast-PPP model. Seven permanent stations are used to assess the errors of the Fast-PPP ionospheric corrections, with baseline distances ranging from 100 to 1000 km from the reference receivers used to compute the Fast-PPP corrections. The 99% of the GPS and Galileo errors in well-sounded areas and in mid-latitude stations are below one total electron content unit. In addition, large errors are bounded by the error prediction of the Fast-PPP model, in the form of the variance of the estimation of the ionospheric corrections. Therefore, we conclude that Fast-PPP is able to provide ionospheric corrections with the required ionospheric accuracy, and realistic confidence bounds, for the Galileo HAS. Open Access funding provided thanks to the CRUECSIC agreement with Springer Nature. The present work was supported in part by the European Space Agency contract IONO4HAS 4000128823/19/NL/AS, by the project RTI2018-094295-B-I00 funded by the MCIN/AEI 10.13039/501100011033 which is co-founded by the FEDER programme and by the Horizon 2020 Marie Skłodowska-Curie Individual Global Fellowship 797461 NAVSCIN.

Open-domain conversational QA (ODCQA) calls for effective question rewriting (QR), as the questions in a conversation typically lack proper context for the QA model to interpret. In this paper, we compare two types of QR approaches, generative and expansive QR, in end-to-end ODCQA systems with recently released QReCC and OR-QuAC benchmarks. While it is common practice to apply the same QR approach for both the retriever and the reader in the QA system, our results show such strategy is generally suboptimal and suggest expansive QR is better for the sparse retriever and generative QR is better for the reader. Furthermore, while conversation history modeling with dense representations outperforms QR, we show the advantages to apply both jointly, as QR boosts the performance especially when limited history turns are considered.

There are many natural questions that are best answered with a list. We address the problem of answering such questions using lists that occur on the Web, i.e. List Question Answering (ListQA). The diverse formats of lists on the Web makes this task challenging. We describe state-of-the-art methods for list extraction and ranking, that also consider the text surrounding the lists as context. Due to the lack of realistic public datasets for ListQA, we present three novel datasets that together are realistic, reproducible and test out-of-domain generalization. We benchmark the above steps on these datasets, with and without context. On the hardest setting (realistic and out-of-domain), we achieve an end-to-end Precision@1 of 51.28% and HITs@5 of 79.38%, effectively demonstrating the difficulty of the task and quantifying the immediate opportunity for improvement. We highlight some future directions through error analysis and release the datasets for further research.

Real-time velocity estimation is a core task in autonomous driving, which is carried out based on available raw sensors such as wheel odometry and motor currents. When the system dynamics and observations can be modeled together as a fully known linear Gaussian state space (SS) model, the celebrated Kalman filter (KF) is a low complexity optimal solution. However, both linearity of the underlying SS model and accurate knowledge of it are often not encountered in practice. This work proposes to estimate the velocity using a hybrid data-driven (DD) implementation of the KF for non-linear systems, coined KalmanNet. KalmanNet integrates a compact recurrent neural network in the flow of the classical KF, retaining low computational complexity, high data efficiency, and interpretability, while enabling operation in non-linear SS models with partial information. We apply KalmanNet on an autonomous racing car as part of the Formula Student (FS) Driverless competition. Our results demonstrate the ability of KalmanNet to outperform a state-of-the-art implementation of the KF that uses a postulated SS model, while being applicable on the vehicle control unit used by the car.

Simple climate models (SCMs) are numerical representations of the Earth's gas cycles and climate system. SCMs are easy to use and computationally inexpensive, making them an ideal tool in both scientific and decision-making contexts (e.g., complex climate model emulation, parameter estimation experiments, climate metric calculations, and probabilistic analyses). Despite their prolific use, the fundamental responses of SCMs are often not directly characterized. In this study, we use fundamental impulse tests of three chemical species (CO2, CH4, and black carbon – BC) to understand the fundamental gas cycle and climate system responses of several comprehensive (Hector v2.0, MAGICC 5.3, MAGICC 6.0) and idealized (FAIR v1.0, AR5-IR) SCMs. We find that while idealized SCMs are widely used, they fail to capture the magnitude and timescales of global mean climate responses under emissions perturbations, which can produce biased temperature results. Comprehensive SCMs, which have physically based nonlinear forcing and carbon cycle representations, show improved responses compared to idealized SCMs. Even the comprehensive SCMs, however, fail to capture the response timescales to BC emission perturbations seen recently in two general circulation models. Some comprehensive SCMs also generally respond faster than more complex models to a 4×CO2 concentration perturbation, although this was not evident for lower perturbation levels. These results suggest where improvements should be made to SCMs. Further, we demonstrate here a set of fundamental tests that we recommend as a standard evaluation suite for any SCM. Fundamental impulse tests allow users to understand differences in model responses and the impact of model selection on results.

A pulse-width-modulation (PWM) bidirectional ac-dc converter (i.e., active rectifier) with dc-faultblocking and self-pre-charge capabilities is proposed in this paper for low-voltage dc (LVDC) applications. The proposed converter, which is named as ”H5 + converter”, consists of an H4 bridge, a bidirectional switch, and a transient-voltage-surge (TVS) diode. The bidirectional switch and the TVS diode enable the dc-fault blocking and dc-bus self-pre-charge, while preserving the low common mode voltage noise and low leakage current of the converter. Additionally, the proposed H5+ converter has advanced features under a dc-side short circuit fault, such as fault diagnosis and fault recovery. Operation principles of the proposed converter are presented and analyzed. A down scaled prototype is built. Experiment results are shown and analyzed, including steady-state waveforms, common-mode performance, start-up dynamics, as well as dc fault blocking, fault diagnosis, and fault recovery. Moreover, the proposed converter is compared with other two dc-fault blocking converters for LVDC applications, in terms of converter capabilities, required devices, and power switch losses.

In the context of the Future Circular Collider (FCC) study, industrial and academic partners are developing novel Nb3Sn superconducting wires with a wide variety of layouts, production methods, and compositions, with the aim of achieving challenging performance targets including a non-copper critical current density of 1500 A mm–2 at 16 T and 4.2 K. There is a clear need for a systematic and quantitative approach to analyze these conductors, identifying correlations between performance, microstructure, and wire design, in order to optimize designs and heat treatments, and to identify the most promising directions for future trials. Image analysis methods have been developed to provide a quantitative description of key geometrical characteristics of a wire with an impact on Nb3Sn phase formation and superconducting performance. In this paper, these methods are introduced, examples are presented of their application to prototype conductors produced for the FCC study, and opportunities for improving the performance of these prototype conductors are identified. Finally, initial steps toward models of diffusion and phase transformations are reported, and the potential for establishing a quantitative, analytical approach to wire design is evaluated, identifying topics requiring further research.

This paper presents comparative results from fault injection (FI) and heavy ions accelerated irradiation on a Xilinx 7 series static RAM (SRAM)-based field-programmable gate array (FPGA) for a soft-core microprocessor mitigated by triple modular redundancy (TMR) with different levels of granularity. The Arm Cortex-M0 soft-core processor executing two software applications is employed as a case study. The TMR implementation is automatically generated from synthesized netlist and includes coarse and fine grain variants. Apart from the TMR mitigation, the configuration memory scrubbing is used as implemented by the engine natively available on Xilinx 7 series FPGAs. Experiments with FI and heavy ions allow analyzing the effectiveness of the automated TMR mitigation combined with memory scrubbing and also to analyze the consistency of reliability metrics from FI and heavy ions. The dynamic cross section of the design was improved up to 4.5 times according to the implemented TMR granularity and when associated with the configuration memory scrubbing.

DC microgrids are expected to play an important role in maximising the benefits of distributed energy resources in future low carbon smart power systems. One of the remaining complex challenges is the requirement for effective DC protection solutions. The advancement of DC protection is hindered by the lack of good understanding and development of reliable and effective earthing schemes which can enable safe and secure operation of DC microgrids in both on-grid and off-grid modes. Therefore, this study discusses different DC microgrid earthing opportunities, and comprehensively evaluates through detailed simulation studies the influence of different earthing methods on the fault behaviour of DC microgrid. A transient model of an active DC microgrid is developed in PSCAD/EMTDC and used for the paper studies.

In this work, we describe a new method for time series classification (TSC) that consists of modeling time series as probability density functions (PDFs) and applies the divergence (the Integrated Squared Error) between two PDFs as a similarity measure for classification via k-Nearest Neighbours (kNN). The proposed method starts by projecting the original time series data into the reconstructed phase space (RPS) via time delay embedding. From these data points in RPS, the corresponding underlying PDF is estimated by Kernel Density Estimation (KDE). Then, a similarity matrix is built by using the Integrated Squared Error (ISE) as a distance measure between two PDFs, from which kNN algorithms can be eventually applied for classification. Two experiments were conducted in order to evaluate this proposal. The first one investigated the impact of the time delay embedding parameters on classification accuracy. We concluded that the embedding dimension is the most influential parameter, for which the results have shown to be highly sensitive. The second experiment provides a comparative analysis of the proposed method against the main state-of-the-art methods for TSC on several well-known benchmark datasets. The results were quite encouraging, and our proposal was able to outperform the compared TSC methods in the majority of the datasets.

International audience; This work presents two fault injection and dependability test methodologies exploring the fault tolerance of successive approximation algorithms. This type of approximate computing algorithm can present an inherent fault tolerance, converging to a final correct output even under faults affecting processed data. A set of algorithms was implemented as embedded software in the ARM Cortex A9 processor of Xilinx Zynq-7000 series board. Experiments consist of exposing the decapsulated processor to laser beams targeting the data cache memory and emulation fault injections at the register file. Results show that successive approximation is effective in protecting the output from faults injected at the data cache memory, but not from the ones injected at the register file. The experiments also show that most of the silent data corruption errors provoked by data cache fault injections are not significant and can be accepted as correct by merely tolerating a result variation of as little as 1%.

The paper presents the performance of the modular multilevel converter (MMC) tapping cells associated with an ac filter in terms of efficiency and power density in a hybrid LV ac/dc microgrid application and compares it to the performance of the conventional topology used in LV application, i.e., the two-level voltage source converter (2L-VSC). A biobjective optimization based on the Genetic Algorithm is hence developed, providing details on designing the components of the LCL filter, the MMC, and the 2L-VSC. The MMC reaches an efficiency of 99.4% when the main dc-grid is left floating. However, due to its modularity and scalability, offering multiport connections option, the MMC tapping cells has the disadvantage of low power density. Exploring the filtering capability of the equivalent arm inductance of the MMC seen from the ac grid side, optimization design results show that higher switching frequencies allow a significant volume reduction of the inductive components of the MMC/ LCL filter while higher switching frequencies have little impact on the switching losses of the MMC. This has the benefit of reducing the overall footprint of the converter and encouraging the use of the MMC in LV application.

Affordable, reliable, and readily accessible electric power underpins our modern society in a multitude of ways, and in this context, the smart grid is becoming an increasingly important factor in power generation, transmission, and distribution. Current analytical tools for the planning, operation, and circuit design in power systems derive from the antecedent technology area of circuit theory, which is both nonobvious for modern data analysts and assumes balanced conditions and a steady state, even though future power networks will routinely experience transient and steady-state unbalances. Next-generation analytical tools should therefore be fully equipped for dynamically unbalanced systems to approach the physical limits of power networks; data analytics is both well suited and necessary for this endeavor but is nonobvious for power engineers. Hence, to fully exploit their evident and promised advantages, an analysis of the smart grid requires close collaboration and convergence between power engineers and experts in signal processing and machine learning, whereby analytical tools expressed in a common language would be a natural step forward.

Active network management (ANM) using power electronic devices will become an essential tool for distribution network operators to deal with the variability of a large number of low-carbon technologies. To enable ANM, this paper proposes a control scheme based on deep reinforcement learning, as an alternative to traditional optimisation. The algorithm uses only a small number of network measurements and can learn approximations of optimal control actions, identified in offline simulations, via a neural network. Once trained, the control scheme chooses power converter set-points that can, for instance, even out loadings on different substations in real-time without the computational burden of high-level optimisation. The performance of the proposed control algorithm is validated against the optimal power flow (OPF) using data from real low-voltage networks. The results show that the solution and benefits are comparable to those obtained by the OPF.

Notwithstanding the significant advancements in the field of deep learning, the basic long short-term memory (LSTM) or Gated Recurrent Unit (GRU) units have largely remained unchanged and unexplored. There are several possibilities in advancing the state-of-art by rightly adapting and enhancing the various elements of these units. Activation functions are one such key element. In this work, we explore using diverse activation functions within GRU and bi-directional GRU (BiGRU) cells in the context of speech emotion recognition (SER). We also propose a novel Attention ReLU GRU (AR-GRU) that employs attention-based Rectified Linear Unit (AReLU) activation within GRU and BiGRU cells. We demonstrate the effectiveness of AR-GRU on one exemplary application using the recently proposed network for SER namely Interaction-Aware Attention Network (IAAN). Our proposed method utilising AR-GRU within this network yields significant performance gain and achieves an unweighted accuracy of 68.3% (2% over the baseline) and weighted accuracy of 66.9 % (2.2 % absolute over the baseline) in four class emotion recognition on the IEMOCAP database.

Many applications employ irregular and sparse memory accesses that cannot take advantage of existing cache hierarchies in high performance processors. To solve this problem, Data Layout Transformation (DLT) techniques rearrange sparse data into a dense representation, improving locality and cache utilization. However, prior proposals in this space fail to provide a design that (i) scales with multi-core systems, (ii) hides rearrangement latency, and (iii) provides the necessary interfaces to ease programmability. In this work we present PLANAR, a programmable near-memory accelerator that rearranges sparse data into dense. By placing PLANAR devices at the memory controller level we enable a design that scales well with multi-core systems, hides operation latency by performing non-blocking fine-grain data rearrangements, and eases programmability by supporting virtual memory and conventional memory allocation mechanisms. Our evaluation shows that PLANAR leads to significant reductions in data movement and dynamic energy, providing an average 4.58× speedup. A. Barredo has been partially supported by the Spanish Government under Formación del Personal Investigadorfellowship number BES-2017-080635. A. Armejach has been partially supported by the Spanish Ministry of Economy, Industry and Competitiveness under Juan de la Cierva postdoctoral fellowship number IJCI-2017-33945. M. Moretó has been partially supported by the Spanish Ministry of Economy, Industry and Competitiveness under Ramon y Cajal fellowship number RYC-2016-21104. This work has been partially supported by the European Union’s Horizon 2020 research and innovation program under the Mont-Blanc 2020 project (grant agreement 779877), by the Spanish Ministry of Science and Innovation (PID2019-107255GB-C21/AEI/10.13039/501100011033), by the Generalitat de Catalunya (contracts 2017-SGR-1414 and 2017-SGR-1328), and by the Arm-BSC Centre of Excellence.

Deep learning methods have proven to be very effective for the task of image denoising even when clean reference images are not available. In particular, Noise2Noise, which requires pairs of noisy images during the training phase, has been shown to yield results as good as approaches using pairs of noisy and clean images (Noise2Clean). However, the performance of Noise2Noise drops when the amount of training data is reduced, limiting its capability in practical scenarios.In this work, an analysis of the Noise2Noise learning strategy is done using real noise and synthetic datasets. This paper demonstrates, using diverse network architectures and loss functions, that the duplicity of information in the noisy pairs can be exploited to reach increased denoising performance of Noise2Noise. Additionally, the issue of overfitting in Noise2Noise is analyzed, given its relevance when training with limited data, and an interpretable early termination criterion is proposed.

This paper presents the first successful experiment implementing whole-body model predictive control with state feedback on a torque-control humanoid robot. We demonstrate that our control scheme is able to do whole-body target tracking, control the balance in front of strong external perturbations and avoid collision with an external object. The key elements for this success are threefold. First, optimal control over a receding horizon is implemented with Crocoddyl, an optimal control library based on differential dynamics programming, providing state-feedback control in less than 10 ms. Second, a warm start strategy based on memory of motion has been implemented to overcome the sensitivity of the optimal control solver to initial conditions. Finally, the optimal trajectories are executed by a low-level torque controller, feedbacking on direct torque measurement at high frequency. This paper provides the details of the method, along with analytical benchmarks with the real humanoid robot Talos.A video of the experiment is available at https://peertube.laas.fr/videos/watch/cbc25927-337c-4635-a1bc-153b9aeb4135

With great potential for being applied to Internet of Things (IoT) applications, the concept of cloud-based Snapshot Real Time Kinematics (SRTK) was proposed and its feasibility under zero-baseline configuration was confirmed recently by the authors. This article first introduces the general workflow of the SRTK engine, as well as a discussion on the challenges of achieving an SRTK fix using actual snapshot data. This work also describes a novel solution to ensure a nanosecond level absolute timing accuracy in order to compute highly precise satellite coordinates, which is required for SRTK. Parameters such as signal bandwidth, integration time and baseline distances have an impact on the SRTK performance. To characterize this impact, different combinations of these settings are analyzed through experimental tests. The results show that the use of higher signal bandwidths and longer integration times result in higher SRTK fix rates, while the more significant impact on the performance comes from the baseline distance. The results also show that the SRTK fix rate can reach more than 93% by using snapshots with a data size as small as 255 kB. The positioning accuracy is at centimeter level when phase ambiguities are resolved at a baseline distance less or equal to 15 km. This research was funded by Albora Technologies and Universitat Politècnica de Catalunya with industrial PhD grant number DI 082 from the Generalitat de Catalunya; This research was partially funded by the Spanish Ministry of Science and Innovation project RTI2018-094295-B-I00. P.C. has been partially supported by the NSF under Awards CNS-1815349 and ECCS-1845833

The volume of air traffic in the National Air Space has been growing at a very fast pace during recent years, and increasing demand for air travel in coming years is predicted. Strategic Planning is a necessary tool to guarantee a safe and efficient increase of Air Traffic. An Aggregate Model for Strategic Planning of Air Traffic Management is the framework of this project.Aggregate models require the creation of a network that is representative of the expected flow and which will serve as a platform to solve flow optimization problems. We have integrated an automatic method for route clustering - Smax method - to generate the required network. Automatic clustering is necessary in order to efficiently cluster many individual scenarios.Some available alternatives to determine the number of clusters are tested. The studied alternatives provide success rates that oscillates between 49% and 68%. In addition to the relatively low success rates, these methods require one user-input parameter, and they are highly sensitive to it.A different method, based on the Silhouette Score and the Dip Test measures, is developed. The Smax method requires no user-input parameters, and it consistently provides rates of success that approximately oscillate between the 72% and the 81%.The presented clustering approach noticeably improves the rate of correct clustering cases for the specific scenario of route clustering.

Global navigation satellite system (GNSS) satellites are equipped with very stable atomic clocks that can be used for assess-ing the models and strategies involved in the estimation processes, where the clock estimates should present high stability. For instance, GNSS products (including satellite and receiver clocks) are computed on daily basis, i.e., with the data of each day being processed independently from other days. This choice produces the well-known day-boundary discontinuities (DBDs) on clock estimates that stem from the estimation process, rather than to the nature of the atomic clock itself. The aim of the present contribution is to propose a strategy to estimate the satellite and receiver clock offsets that is capable to reduce the DBDs observed in the products of different analysis centers (ACs) within the International GNSS Service (IGS), ultimately improving the accuracy of clock estimates. Our approach relies on the use of unambiguous, undifferenced and uncombined carrier phase measurements collected by a network of permanent receivers on ground. The strategy consid-ers the carrier phase hardware delays and assumes their possible variations along time. Our daily data processing aims to maintaining the natural continuity over days of the carrier phase measurements after integer ambiguity resolution (IAR), even if IAR is performed on daily batches. We compare our clock estimations with those computed by different IGS ACs, evaluating the linear behavior of the satellite atomic clocks on the day change. The results show the removal of DBD on clock estimates computed with the continuous and unambiguous carrier phase measurements. This DBD improvement may benefit the statistical characterization of long-term phenomena correlated with the on-board clocks. The present work was supported in part by the European Space Agency contract (REL-GAL) N.4000122402/17/NL/IB, by the Spanish Ministry of Science, Innovation and Universities project RTI2018-094295-B-I00 and by the Horizon 2020 Marie Skłodowska-Curie Individual Global Fellowship 797461 NAVSCIN. The authors acknowledge the use of data and products provided by the International GNSS Service

This article presents an energy regulator based on a Model Predictive Control (MPC) algorithm for a Voltage Source Converter (VSC). The MPC is formulated to optimise the converter performance according to the weights defined in an objective function that trades off additional features, such as current harmonic distortion, reactive power tracking and DC bus voltage oscillation. Differently from most approaches found in the research literature, the MPC proposed here considers the coupling dynamics between the AC and DC sides of the VSC. This study is focused on the example case of a single-phase VSC, which presents a nonlinear relationship between its AC and DC sides and a sustained double-line frequency power disturbance in its DC bus. To reduce the burden of the MPC, the controller is formulated to benefit from the slow energy dynamics of the system. Thus, the cascaded structure typically used in the control of VSCs is kept and the MPC is set as an energy regulator at a reduced sampling frequency while the current control relies on a fast inner controller. The computational burden of the algorithm is further reduced by using a linear time-varying approximation. The controller is presented in detail and experimental validation showing the performance of the algorithm is provided.

The GNSS LABoratory tool (gLAB) is an interactive educational suite of applications for processing data from the Global Navigation Satellite System (GNSS). gLAB is composed of several data analysis modules that compute the solution of the problem of determining a position by means of GNSS measurements. The present work aimed to improve the pre-fit outlier detection function of gLAB since outliers, if undetected, deteriorate the obtained position coordinates. The methodology exploits robust statistical tools for regression provided by the Flexible Statistics and Data Analysis (FSDA) toolbox, an extension of MATLAB for the analysis of complex datasets. Our results show how the robust analysis FSDA technique improves the capability of detecting actual outliers in GNSS measurements, with respect to the present gLAB pre-fit outlier detection function. This study concludes that robust statistical analysis techniques, when applied to the pre-fit layer of gLAB, improve the overall reliability and accuracy of the positioning solution. This work was supported in part by the Spanish Ministry of Science, Innovation and Universities project RTI2018-094295-B-I00 and in part by the Horizon 2020 Marie Skłodowska-Curie Individual Global Fellowship 797461 NAVSCIN.

In recent years there has been a surge of interest in designing custom accelerators for power-efficient high-performance computing. However, available tools to simulate low-level RTL designs often neglect the target system in which the design will operate. This hinders proper testing and debugging of functionalities, and does not allow co-designing the accelerator to obtain a balanced and efficient architecture. In this paper, we introduce gem5 + rtl, a flexible framework that enables simulation of RTL models inside a full-system software simulator. We present the framework’s functionality that allows easy integration of RTL models on a simulated system-on-chip (SoC) that is able to boot Linux and run complex multi-threaded and multi-programmed workloads. We demonstrate the framework with two relevant use cases that integrate a multi-core SoC with a Performance Monitoring Unit (PMU) and the NVIDIA Deep Learning Accelerator (NVDLA), showcasing how the framework enables testing RTL model features and how it can enable co-design taking into account the entire SoC. This research was supported by the European Union Regional Development Fund within the framework of the ERDF Operational Program of Catalonia 2014-2020 with a grant of 50% of total cost eligible under the DRAC project [001-P- 001723], by the Spanish goverment (grant RTI2018-095094- B-C21 CONSENT), by the Spanish Ministry of Science and Innovation (contracts PID2019-107255GB-C21) and by the Catalan Government (contracts 2017-SGR-1414, 2017-SGR705). This work has also been supported by the European Community’s Horizon 2020 Framework Programme under the Mont-Blanc 2020 and EPI projects (grant agreements n. 779877 and n. 826647); and by the Arm-BSC Center of Excellence. G. López-Paradís has been partially supported by the Agency for Management of University and Research Grants (AGAUR) of the Government of Catalonia under Ajuts per a la contractació de personal investigador novell fellowship No. 2021FI B00994. A. Armejach has been partially supported by the Spanish Ministry of Economy, Industry and Competitiveness under Juan de la Cierva postdoctoral fellowship number IJCI-2017-33945. M. Moretó has been partially supported by the Spanish Ministry of Economy, Industry and Competitiveness under Ramón y Cajal fellowship No. RYC-2016-21104.

The Mont-Blanc 2020 (MB2020) project has triggered the development of the next generation industrial processor for Big Data and High Performance Computing (HPC). MB2020 is paving the way to the future low-power European processor for exascale, defining the System-on-Chip (SoC) architecture and implementing new critical building blocks to be integrated in such an SoC. In this paper, we first present an overview of the MB2020 project, then we describe our experimental infrastructure, the requirements of relevant applications, and the IP blocks developed in the project. Finally, we present our emulation-based final demonstrator and explain how it integrates within our first generation of HPC processors. This work is supported by the European Community’s Horizon 2020 Framework Programme under the Mont-Blanc 2020 project, grant agreement n. 779877.

© 2021 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting /republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works The paper addresses a real-time optimization-based reference calculation integrated with a control structure for Modular Multilevel Converters (MMC) operating under normal and constrained situations (where it has reached current and/or voltage limitations, e.g. during system faults). The algorithm prioritizes to satisfy the Transmission System Operators (TSO) AC grid current demanded set-points. The constrained optimization problem is formulated based on the steady-state model of the MMC, whereby the prioritization is achieved through distinct weights defined in the Objective Function’s (OF) terms. The resultant optimization problem, however, is highly nonlinear requiring high computation burden to be solved in real-time. To overcome this issue, this paper applies a Linear Time-Varying (LTV) approximation, where the nonlinear dynamics of the system are represented as constant parameters, while a Linear Time-Invariant (LTI) system is used to formulate the optimization constraints. The converter's current references are determined in real-time by solving a constrained linearized optimization problem at each control time step, considering the TSO's demands, the current MMC operating point and its physical limitations. Finally, the linearized-optimization problem is integrated with the MMC controllers and evaluated under different network conditions, where the results indicated that method can be potentially employed to obtain the MMCs current references.

BackgroundAlthough frustration is a common emotional reaction during playing games, an excessive level of frustration can harm users’ experiences, discouraging them from undertaking further game interactions. The automatic detection of players’ frustration enables the development of adaptive systems, which through a real-time difficulty adjustment, would adapt the game to the user’s specific needs; thus, maximising players experience and guaranteeing the game success. To this end, we present our speech-based approach for the automatic detection of frustration during game interactions, a specific task still under-explored in research.MethodThe experiments were performed on the Multimodal Game Frustration Database (MGFD), an audiovisual dataset—collected within the Wizard-of-Oz framework—specially tailored to investigate verbal and facial expressions of frustration during game interactions. We explored the performance of a variety of acoustic feature sets, including Mel-Spectrograms and Mel-Frequency Cepstral Coefficients (MFCCs), as well as the low dimensional knowledge-based acoustic feature set eGeMAPS. Due to the always increasing improvements achieved by the use of Convolutional Neural Networks (CNNs) in speech recognition tasks, unlike the MGFD baseline—based on Long Short-Term Memory (LSTM) architecture and Support Vector Machine (SVM) classifier—in the present work we take into consideration typically used CNNs, including ResNets, VGG, and AlexNet. Furthermore, given the still open debate on the shallow vs deep networks suitability, we also examine the performance of two of the latest deep CNNs, i. e., WideResNets and EfficientNet.ResultsOur best result, achieved with WideResNets and Mel-Spectrogram features, increases the system performance from 58.8 % Unweighted Average Recall (UAR) to 93.1 % UAR for speech-based automatic frustration recognition.

Comunicació presentada al 15th IEEE International Conference on Automatic Face and Gesture Recognition (FG 2020), celebrat del 16 al 20 de novembre de 2020 a Buenos Aires, Argentina. There is a growing interest in affective computing research nowadays given its crucial role in bridging humans with computers. This progress has recently been accelerated due to the emergence of bigger dataset. One recent advance in this field is the use of adversarial learning to improve model learning through augmented samples. However, the use of latent features, which is feasible through adversarial learning, is not largely explored, yet. This technique may also improve the performance of affective models, as analogously demonstrated in related fields, such as computer vision. To expand this analysis, in this work, we explore the use of latent features through our proposed adversarial-based networks for valence and arousal recognition in the wild. Specifically, our models operate by aggregating several modalities to our discriminator, which is further conditioned to the extracted latent features by the generator. Our experiments on the recently released SEWA dataset suggest the progressive improvements of our results. Finally, we show our competitive results on the Affective Behavior Analysis in-the-Wild (ABAW) challenge dataset. This work is partly supported by the Spanish Ministry of Economy and Competitiveness under project grant TIN2017- 90124-P, the Maria de Maeztu Units of Excellence Programme (MDM-2015-0502), and the donation bahi2018-19 to the CMTech at UPF. Further funding has been received from the European Union’s Horizon 2020 research and innovation programme under grant agreement No. 826506 (sustAGE).

Measuring airways in chest computed tomography (CT) scans is important for characterizing diseases such as cystic fibrosis, yet very time-consuming to perform manually. Machine learning algorithms offer an alternative, but need large sets of annotated scans for good performance. We investigate whether crowdsourcing can be used to gather airway annotations. We generate image slices at known locations of airways in 24 subjects and request the crowd workers to outline the airway lumen and airway wall. After combining multiple crowd workers, we compare the measurements to those made by the experts in the original scans. Similar to our preliminary study, a large portion of the annotations were excluded, possibly due to workers misunderstanding the instructions. After excluding such annotations, moderate to strong correlations with the expert can be observed, although these correlations are slightly lower than inter-expert correlations. Furthermore, the results across subjects in this study are quite variable. Although the crowd has potential in annotating airways, further development is needed for it to be robust enough for gathering annotations in practice. For reproducibility, data and code are available online: http://github.com/adriapr/crowdairway.git.

With Europe dedicated to limiting climate change and greenhouse gas emissions, large shares of Renewable Energy Sources (RES) are being integrated in the national grids, phasing out conventional generation. The new challenges arising from the energy transition will require a better coordination between neighboring system operators to maintain system security. To this end, this paper studies the benefit of exchanging primary frequency reserves between asynchronous areas using the Supplementary Power Control (SPC) functionality of High-Voltage Direct-Current (HVDC) lines. First, we focus on the derivation of frequency metrics for asynchronous AC systems coupled by HVDC interconnectors. We compare two different control schemes for HVDC converters, which allow for unilateral or bilateral exchanges of reserves between neighboring systems. Second, we formulate frequency constraints and include them in a unit commitment problem to ensure the N-1 security criterion. A data-driven approach is proposed to better represent the frequency nadir constraint by means of cutting hyperplanes. Our results suggest that the exchange of primary reserves through HVDC can reduce up to 10% the cost of reserve procurement while maintaining the system N-1 secure., Submitted to "IEEE Transactions on Power Systems" on November 24, 2020 - Revised on April 14, 2021

The high prevalence of chronic pain in society raises the need to develop new digital tools that can automatically and objectively assess pain intensity in individuals. These tools can contribute to an optimisation of clinical resources, as they offer cost-effective solutions for early detection, continuous monitoring, and treatment personalisation by utilising Artificial Intelligence techniques. In this work, we present our contribution to the Pain Intensity Estimation from Facial Expressions task of the EMOPAIN 2020 Challenge. Specifically, we compare the performance of Recurrent Neural Networks trained with standard or Curriculum Learning (CL) approaches to predict the pain intensity level of individuals reported in an 11-point scale from facial expressions. The results obtained using the test partition support the use of CL-based approaches in the automatic prediction of pain from facial features. The best model trained using a CL approach achieved a Concordance Correlation Coefficient (CCC) of 0.196 in the test partition, while the model trained using a standard approach, without CL, achieved a CCC of 0.174. In terms of CCC, these results respectively represent an improvement of 0.136 and 0.114 on the best results of the baseline system reported by the Challenge organisers using the test partition.

Multimodal Sentiment Analysis in Real-life Media (MuSe) 2020 is a Challenge-based Workshop focusing on the tasks of sentiment recognition, as well as emotion-target engagement and trustworthiness detection by means of more comprehensively integrating the audio-visual and language modalities. The purpose of MuSe 2020 is to bring together communities from different disciplines; mainly, the audio-visual emotion recognition community (signal-based), and the sentiment analysis community (symbol-based). We present three distinct sub-challenges: MuSe-Wild, which focuses on continuous emotion (arousal and valence) prediction; MuSe-Topic, in which participants recognise 10 domain-specific topics as the target of 3-class (low, medium, high) emotions; and MuSe-Trust, in which the novel aspect of trustworthiness is to be predicted. In this paper, we provide detailed information on MuSe-CAR, the first of its kind in-the-wild database, which is utilised for the challenge, as well as the state-of-the-art features and modelling approaches applied. For each sub-challenge, a competitive baseline for participants is set; namely, on test we report for MuSe-Wild a combined (valence and arousal) CCC of .2568, for MuSe-Topic a score (computed as 0.34 * UAR + 0.66 * F1) of 76.78 % on the 10-class topic and 40.64 % on the 3-class emotion prediction, and for MuSe-Trust a CCC of .4359.

The present paper proposes a novel device defined as an intelligent electronic fuse (iFuse) meant to be connected in series with any current-bidirectional voltage-unidirectional active switch present in a given converter. The iFuse duty is to isolate its series- associated switch from the rest of the converter circuit immediately after detecting that said switch has failed in short circuit. Nonetheless, it maintains the reverse (free- wheeling) current path originally offered by the failed switch. The failure detection is performed when the failed switch causes a shoot-through event. Therefore, the iFuse is designed to be able to block the elevated current occurring in such event. The iFuse allows increasing the fault-tolerant capability and the reliability of power converters where such qualities are hindered by switch short-circuit failures, as in converters featuring parallelized switches, neutral-point-clamped multilevel topologies, or redundant legs. The feasibility of the iFuse device is verified through experimental tests, proving that the device is able to detect the failure of its associated switch and isolate it from the rest of the converter circuit in 6 μs, while stopping short-circuit currents of up to 1 kA without incurring in harmful di/dt values.

Off-grid micro-grids and solar home systems have become an enabler for the electrification of rural areas in developing countries. Their future integration into full-scale grids poses multiple technical challenges. This paper presents a concept of local DC mini-grid formed by the interconnection of nearby micro-grids. This is envisioned as a step towards the bottom-up development of a larger grid. The paper places special emphasis on the analysis of the controller structure, which is based on a variation of the common droop controller. Generally, decentralised droop schemes present inherent limitations to achieve, simultaneously, DC voltage regulation and accurate energy tracking. The decentralised controller proposed here improves conventional approaches by optimally adjusting the droop controller setpoints based on an Optimal Power Flow (OPF) algorithm that takes into account the characteristics of individual converters and the energy trading commands. This approach can benefit from the communication layer required by the energy trading system. The control design of the mini-grid is customised for its deployment on a variation of the dual active bridge (DAB) power electronic converter. Despite the use of a centralised OPF, the system stability is guaranteed by the decentralised low-level controller. The performance of the control approach and its sensitivity to unknown system parameters is measured and validated through detailed simulation studies.

Typical Modular Multilevel Converter (MMC)-based HVDC systems don't include capacitors directly connected to the DC-side in parallel with the HVDC lines. The dynamics of the HVDC line voltage depend on the equivalent capacitance of the HVDC system, which in turn depends on the length of the lines. Recent work in the literature has shown that this can become problematic for HVDC line voltage control in the case of short links. This paper describes the operation of a laboratory-scaled MMC-based HVDC short link platform devised to study this phenomenon. Special emphasis is put on the design and the scaling of a frequency-dependent cable emulator. Experimental results are compared with a benchmark high-voltage simulated model.

The FM-index is a data structure used in genomics for exact search of input sequences over large reference genomes. Algorithms based on the FM-index show an irregular memory access pattern, resulting in a memory bound problem. We analyze a recent implementation of the FM-index and highlight existing throughput-memory trade-offs, showing that memory requirements limit implementation of large k-steps. We propose COFI, a COmpressed FM-Index for large K-steps. COFI enables a 15-step FM-index using less than 16 GB for a human genome reference of 3 giga base pairs. An algorithm based on this new layout is evaluated on both a Knights Landing (KNL) and an Skylake-based system (SKX). We achieve average speed-ups of 1.46X and 1.39X, respectively, with respect to an state-of-the-art FM-index implementation that is already well optimized.

Overall, low-carbon technologies can have a positive effect on distribution networks, however large clusters can result in severe negative impacts. These problems can be alleviated by redistributing power flows through network reconfiguration or by using power electronic devices to soft-mesh the radial lines. Whilst the first offers limited control due to a reduced number of possible open/close switch commands, the second offers greater flexibility and continuous management of the power but depends on a more complex control scheme.The present paper uses high-resolution data, obtained from real low-voltage substations and feeders, to evaluate and compare the effectiveness of these two strategies. Specifically, a conventional network reconfiguration approach is compared to an alternative, based on soft-open-points and high-level optimization. The findings show that whilst both methods offer similar benefits, considering present day scenarios, the approach based on power electronic converters becomes advantageous in the long-term, as the volume of low-carbon technologies increases.