Search

In this paper, we propose a novel deep learning architecture to classify unsafe driving maneuvers from dashcam and IMU data. Such architecture processes the output of an object detection algorithm in combination with raw video frames and GPS/IMU data. At the core of the architecture there is a novel Spatio-Temporal Attention Selector (STAS) module, which (1) extracts features describing the evolution of each object in the scene over time and (2) leverages multi-head dot product attention to select the relevant ones, i.e., the dangerous ones or the ones in danger, to perform classification. We also introduce a simple but effective methodology to increase the benefit of fine-tuning the backbone network. Our method is shown to achieve higher performance than other approaches in the literature applying attention over single frames.

With the rapid development of sensor technologies, time series data collected by multiple and spatially distributed sensors have been widely used in different research fields. Examples of such data include geo-tagged temperature data collected by temperature sensors, air pollutant monitoring data, and traffic data collected by road traffic sensors. Due to sensor failure, communication errors and storage loss, etc., data collected by sensors inevitably includes missing data. However, models commonly used in the analysis of such large-scale data often rely on complete data sets. This paper proposes a model for the imputation of missing data of traffic flow, which combines a self-attention mechanism, an auto-encoder, and a generative adversarial network, into a self-attention generative adversarial imputation net (SA-GAIN). The introduction of the self-attention mechanism can help the proposed model to effectively capture correlations between spatially-distributed sensors at different time points. Adversarial training through two neural networks, called generators and discriminators, allows the proposed model to generate imputed data close to the real data. In comparison with different imputation models, the proposed model shows the best performance in imputing missing data.

Personal informatics systems can help people promote their health and well-being. Recent studies have shown that such systems can be used to infer relevant health indicators such as, e.g., stress, anxiety, and sleeping habits. While automatic detection of sleep has been studied extensively, there is a lack of studies exploring how population and personalized models influence the performance of sleep detection. In this article, we address this challenge by investigating the recognition of sleep/wake stages and high/low sleep quality with a focus on the impact of personalized models. To evaluate our approach, we collect a dataset of physiological signals and self-reports about sleep/wake times and sleep quality score. The dataset contains 6557 hours of sensor data collected using wristbands from 16 participants over one month. Our results show that personalized models perform significantly better than population models for sleep quality recognition, and are comparably good for sleep stage detection. The balanced accuracy for sleep/wake and high/low sleep quality are 92.2% and 61.51%, which are significantly higher than baseline classifiers.

The study of human behaviors in cognitive sciences provides clues to understand and describe people’s personal and interpersonal functioning. In particular, the temporal analysis of behavioral dynamics can be a powerful tool to reveal events, correlations and causalities but also to discover abnormal behaviors. However, the annotation of these dynamics can be expensive in terms of temporal and human resources. To tackle this challenge, this paper proposes a methodology to semi-automatically annotate behavioral data. Behavioral dynamics can be expressed as sequences of simple dynamical processes: transitions between such processes are generally known as change-points. This paper describes the necessary steps to detect and classify change-points in behavioral data by using a dataset collected in a real use-case scenario. This dataset includes motor observations from children with typical development and with neuro-developmental disorders. Abnormal movements which are present in such disorders are useful to validate the system in conditions that are challenging even for experienced annotators. Results show that the system: can be effective in the semi-automated annotation task; can be efficient in presence of abnormal behaviors; may achieve good performance when trained with limited manually annotated data.

Radio resource management (RRM) is critical in 5G mobile communications due to its ubiquity on every radio device and its low latency constraints. The rapidly evolving RRM algorithms with low latency requirements combined with the dense and massive 5G base station deployment ask for an on-the-edge RRM acceleration system with a tradeoff between flexibility, efficiency, and cost-making application-specific instruction-set processors (ASIPs) an optimal choice. In this work, we start from a baseline, simple RISC-V core and introduce instruction extensions coupled with software optimizations for maximizing the throughput of a selected set of recently proposed RRM algorithms based on models using multilayer perceptrons (MLPs) and recurrent neural networks (RNNs). Furthermore, we scale from a single-ASIP to a multi-ASIP acceleration system to further improve RRM throughput. For the single-ASIP system, we demonstrate an energy efficiency of 218 GMAC/s/W and a throughput of 566 MMAC/s corresponding to an improvement of $10\times $ and $10.6\times $ , respectively, over the single-core system with a baseline RV32IMC core. For the multi-ASIP system, we analyze the parallel speedup dependency on the input and output feature map (FM) size for fully connected and LSTM layers, achieving up to $10.2\times $ speedup with 16 cores over a single extended RI5CY core for single LSTM layers and a speedup of $13.8\times $ for a single fully connected layer. On the full RRM benchmark suite, we achieve an average overall speedup of $16.4\times $ , $25.2\times $ , $31.9\times $ , and $38.8\times $ on two, four, eight, and 16 cores, respectively, compared to our single-core RV32IMC baseline implementation.

OpenMP has become a reference standard for the design of parallel applications. This standard is evolving very fast, thus offering ever new opportunities to the application programmers. However, OpenMP runtime environments are often not fully aligned to the actual requirements imposed by the evolution of such standard. Among the main lacks, we find: (a) a limited capability to effectively cope with task priorities, and (b) the inadequacy in guaranteeing core properties while processing tasks such as the so called extit{work-conservativeness}---the ability of the OpenMP runtime environment to fully exploit the underlying multi-processor/multi-core machine through the avoidance of thread-blocking phases. In this article we present the design of extensions to the GNU OpenMP ({ t GOMP}) implementation, integrated into { t gcc}, which allow the effective management of tasks and their priorities. Our proposal is based on a user-space library---modularly combined with the one already offered by { t GOMP}---and an external kernel-level Linux module---offering the opportunity to exploit raising hardware facilities for the purpose of task/priority management. We also provide experimental results showing the effectiveness of our proposal, achieved by running either OpenMP common benchmarks or a new benchmark application (named {sc Hashtag-Text}) that we explicitly devised in order to stress the OpenMP runtime environment in relation to the above-mentioned task/priority management aspects.

Nowadays, industrial robotics requires that robots and humans share the same workspace and collaborate to a certain extent. In such a scenario, the safety is the minimum requirement and, for this reason, many off-the-shelf collaborative robots are now available on the market which, basically, are able to limit the contact forces in the case of impact. Differently from many different works, this article presents a solution to human multi-robot safe interaction in which multiple mobile manipulators are in charge of performing a cooperative task in a workspace shared with human operators. The safety of the interaction is assessed by a safety field that considers the whole system and is general enough concerning its expression. Based on the value of this field, the cooperative task trajectory is properly modified so as to ensure a safe interaction while trying to preserve as much as possible the nominal task, which is instead completely aborted whenever the safety of the interaction cannot be guaranteed. The solution is first designed within a centralized architecture and, then, upon this, a distributed implementation is presented, which, in general, aims to exhibit the same performance as the centralized counterpart. Finally, both simulations and experiments on real industrial robots corroborate the designed solution.

Humans drive in a holistic fashion which entails, in particular, understanding dynamic road events and their evolution. Injecting these capabilities in autonomous vehicles can thus take situational awareness and decision making closer to human-level performance. To this purpose, we introduce the ROad event Awareness Dataset (ROAD) for Autonomous Driving, to our knowledge the first of its kind. ROAD is designed to test an autonomous vehicle's ability to detect road events, defined as triplets composed by an active agent, the action(s) it performs and the corresponding scene locations. ROAD comprises videos originally from the Oxford RobotCar Dataset, annotated with bounding boxes showing the location in the image plane of each road event. We benchmark various detection tasks, proposing as a baseline a new incremental algorithm for online road event awareness termed 3D-RetinaNet. We also report the performance on the ROAD tasks of Slowfast and YOLOv5 detectors, as well as that of the winners of the ICCV2021 ROAD challenge, which highlight the challenges faced by situation awareness in autonomous driving. ROAD is designed to allow scholars to investigate exciting tasks such as complex (road) activity detection, future event anticipation and continual learning. The dataset is available at https://github.com/gurkirt/road-dataset; the baseline can be found at https://github.com/gurkirt/3D-RetinaNet., IEEE Transactions on Pattern Analysis and Machine Intelligence, 45 (1), ISSN:0162-8828, ISSN:1939-3539

Recurrent neural networks such as Long Short-Term Memories (LSTMs) learn temporal dependencies by keeping an internal state, making them ideal for time-series problems such as speech recognition. However, the output-to-input feedback creates distinctive memory bandwidth and scalability challenges in designing accelerators for RNNs. We present Muntaniala, an RNN accelerator architecture for LSTM inference with a silicon-measured energy-efficiency of 3.25 TOP/s/W and performance of 30.53 GOP/s in UMC 65nm technology. The scalable design of Muntaniala allows running large RNN models by combining multiple tiles in a systolic array. We keep all parameters stationary on every die in the array, drastically reducing the I/O communication to only loading new features and sharing partial results with other dies. For quantifying the overall system power, including I/O power, we built Vau da Muntanialas, to the best of our knowledge, the first demonstration of a systolic multi-chip-on-PCB array of RNN accelerator. Our multi-die prototype performs LSTM inference with 192 hidden states in 330 μs with a total system power of 9.0 mW at 10 MHz consuming 2.95 μJ. Targeting the 8/16-bit quantization implemented in Muntaniala, we show a phoneme error rate (PER) drop of approximately 3% with respect to floating-point (FP) on a 3L-384NH-123NI LSTM network on the TIMIT dataset. ISSN:1549-8328 ISSN:1057-7122 ISSN:1558-0806

This paper investigates a new lane reservation problem with task merging that consists of optimally determining which lanes in a transportation network have to be reserved and designing reserved lane-based routes in the network for time-crucial transport tasks. Part of the tasks whose destinations are geographically close is merged to reduce the number of vehicles and transport costs. Reserved lanes can reduce the travel time of task vehicles passing through them, while they will generate negative impact on normal traffic, such as traffic delay to the vehicles on adjacent non-reserved lanes. The objective is to minimize the total negative impact of all reserved lanes. For this problem, two new integer linear programming (ILP) models are first developed. The complexity of the problem is proved to be NP-hard. Since commercial solver (like CPLEX) is time-consuming for solving it when the problem size increases, a fast and effective improved differential evolution algorithm (IDEA) is developed based on explored problem properties. Extensive experimental results for a real-life case and benchmark instances of up to 500 nodes in the network and 30 transport tasks show the favorable performance of the IDEA, as compared to CPLEX, differential evolution algorithm and genetic algorithm. Management insights are also drawn to support practical decision-making.

The fourth industrial revolution, also known as Industry 4.0, is reshaping the way individuals live and work while providing a substantial influence on the manufacturing scenario. The key enabling technology that has made Industry 4.0 a concrete reality is without doubt collaborative robotics, which is also evolving as a fundamental pillar of the next revolution, the so-called Industry 5.0. The improvement of employees' safety and well-being, together with the increase of profitability and productivity, are indeed the main goals of human-robot collaboration (HRC) in the industrial setting. The robotic controller design and the analysis of existing decision and control techniques are crucially needed to develop innovative models and state-of-the-art methodologies for a safe, ergonomic, and efficient HRC. To this aim, this paper presents an accurate review of the most recent and relevant contributions to the related literature, focusing on the control perspective. All the surveyed works are carefully selected and categorized by target (i.e., safety, ergonomics, and efficiency), and then by problem and type of control, in presence or absence of optimization. Finally, the discussion of the achieved results and the analysis of the emerging challenges in this research field are reported, highlighting the identified gaps and the promising future developments in the context of the digital evolution.

For a driving simulator to be a valid tool for research, vehicle development, or driver training, it is crucial that it elicits similar driver behavior as the corresponding real vehicle. To assess such behavioral validity, the use of quantitative driver models has been suggested but not previously reported. Here, a task-general conceptual driver model is proposed, along with a taxonomy defining levels of behavioral validity. Based on these theoretical concepts, it is argued that driver models without explicit representations of sensory or neuromuscular dynamics should be sufficient for a model-based assessment of driving simulators in most contexts. As a task-specific example, two parsimonious driver steering models of this nature are developed and tested on a dataset of real and simulated driving in near-limit, low-friction circumstances, indicating a clear preference of one model over the other. By means of closed-loop simulations, it is demonstrated that the parameters of this preferred model can generally be accurately estimated from unperturbed driver steering data, using a simple, open-loop fitting method, as long as the vehicle positioning data are reliable. Some recurring patterns between the two studied tasks are noted in how the model’s parameters, fitted to human steering, are affected by the presence or absence of steering torques and motion cues in the simulator.

In the last years, the need to de-identify privacy-sensitive information within Electronic Health Records (EHRs) has become increasingly felt and extremely relevant to encourage the sharing and publication of their content in accordance with the restrictions imposed by both national and supranational privacy authorities. In the field of Natural Language Processing (NLP), several deep learning techniques for Named Entity Recognition (NER) have been applied to face this issue, significantly improving the effectiveness in identifying sensitive information in EHRs written in English. However, the lack of data sets in other languages has strongly limited their applicability and performance evaluation. To this aim, a new de-identification data set in Italian has been developed in this work, starting from the 115 COVID-19 EHRs provided by the Italian Society of Radiology (SIRM): 65 were used for training and development, the remaining 50 were used for testing. The data set was labelled following the guidelines of the i2b2 2014 de-identification track. As additional contribution, combined with the best performing Bi-LSTM + CRF sequence labeling architecture, a stacked word representation form, not yet experimented for the Italian clinical de-identification scenario, has been tested, based both on a contextualized linguistic model to manage word polysemy and its morpho-syntactic variations and on sub-word embeddings to better capture latent syntactic and semantic similarities. Finally, other cutting-edge approaches were compared with the proposed model, which achieved the best performance highlighting the goodness of the promoted approach.

Manipulation in contrast to grasping is a trajectorial task that needs to use dexterous hands. Improving the dexterity of robot hands, increases the controller complexity and thus requires to use the concept of postural synergies. Inspired from postural synergies, this research proposes a new framework called kernelized synergies that focuses on the re-usability of same subspace for precision grasping and dexterous manipulation. In this work, the computed subspace of postural synergies is parameterized by kernelized movement primitives to preserve its grasping and manipulation characteristics and allows its reuse for new objects. The grasp stability of proposed framework is assessed with the force closure quality index, as a cost function. For performance evaluation, the proposed framework is initially tested on two different simulated robot hand models using the Syngrasp toolbox and experimentally, four complex grasping and manipulation tasks are performed and reported. Results confirm the hand agnostic approach of proposed framework and its generalization to distinct objects irrespective of their dimensions.

Central banks and governments all over the world are increasingly exploring digital versions of fiat money, known as retail Central Bank Digital Currencies (CBDCs). Most initiatives rely on Distributed Ledger Technologies and are presented as alternatives to physical cash. Consequently, anonymity-related regulatory questions have naturally started to arise in terms of Anti-Money Laundering and Counter-Terrorist Financing compliance. Against this backdrop, this paper provides a techno-legal taxonomy of approaches to balance privacy and transparency in CBDCs without thwarting accountability, but it also underlines cross-sectoral impacts. The contribution heeds regulation-by-design as its core methodological foundation, with Privacy-Enhancing Technologies as the relevant use case. Thus, it highlights that not only technology aids legal purposes, but also that some regulatory requirements ought to be designed into technology for one to reach agreed-upon results and/or standards. ispartof: IEEE Transactions On Network and Service Management vol:19 issue:2 pages:1776-1788 status: Published online

Machine learning is based on training and can be implemented through Reinforcement Learning. This technique is of clear skinnerian derivation: the actions that lead to the achievement of the objective are rewarded, while those that lead to defeat are punished, according to the sequential and rewarding logic that presides over the Task Analysis. While, however, in task analysis techniques carried out with children/students, personal or cultural parameters are taken into account, albeit superficially, this does not happen in Reinforcement Learning, where learning proceeds by trial and error and control and supervision remain the prerogative of the human being. The “choices” of machines, therefore, are conditioned by the programmer’s culture. One possibility is offered by Deep Reinforcement Learning, based on Convolutional Neural Networks that help the algorithm to plan, understand and strategically elaborate the actions to be taken, even if it remains risky, according to the authors, to delegate the interpretative and predictive aspect to the machines.

This paper presents a fully parametrized framework, entirely described in VHDL, to simplify the FPGA implementation of non-recurrent Artificial Neural Networks (ANNs), which works independently of the complexity of the networks in terms of number of neurons, layers and, to some extent, overall topology. More specifically, the network may consist of fully-connected, max-pooling or convolutional layers which can be arbitrarily combined. The ANN is used only for inference, while back-propagation is performed off-line during the ANN learning phase. Target of this work is to achieve fast-prototyping, small, low-power and cost-effective implementation of ANNs to be employed directly on the sensing nodes of IOT (i.e. Edge Computing). The performance of so-implemented ANNs is assessed for two real applications, namely hand movement recognition based on electromyographic signals and handwritten character recognition. Energy per operation is measured in the FPGA realization and compared with the corresponding ANN implemented on a microcontroller (mu C) to demonstrate the advantage of the FPGA based solution.

This paper reports the development of a manipulation system for electric wires, implemented by means of a commercial gripper installed on an industrial manipulator and equipped with cameras and suitably designed tactile sensors. The purpose of this system is the execution of wire insertion on commercial electromechanical components. The synergy between computer vision and tactile sensing is necessary because, in a real environment, the tight spaces very often prevent the possibility to use the vision system, also when the same task is performed by a human being. A novel technique to speed up the generation of training data sets for convolutional neural networks (CNNs) is proposed. Therefore, this technique is used to train a CNN in order to detect small objects (such as wire terminals). Moreover, aiming to prevent faults during the task and to interact with the environment safely, several machine learning approaches are used to produce an affordable output from the tactile sensor. The proposed approach shows how a cheap sensor embedded with suitable intelligence can provide information comparable to a more expensive force sensor. Note to Practitioners —This paper was motivated by the lack of commercial solution for the automatic cabling of switchgears. Existing approaches to this problem are in some way limited to specific large-scale products or simple layouts. This paper investigated a robust and flexible solution, based on the exploitation of multiple sensors and machine learning algorithms, for wire detection, grasping, and connection. The proposed approach is characterized by simple design and self-tuning capabilities, and it can be easily employed on a wide range of switchgear layouts thanks to the large workspace of the manipulator. Experimental results show that the proposed system is able to achieve a 95% success rate within a realistic admissible region. In the future research, we will integrate the proposed solution with an electromechanical component localization module and a terminal fastening system to evaluate the performance on the real production line.

Mobile crowdsensing harnesses the sensing power of modern smartphones to collect and analyze data beyond the scale of what was previously possible with traditional sensor networks. Given the participatory nature of mobile crowdsensing, it is imperative to incentivize mobile users to provide sensing services in a timely and reliable manner. Most importantly, given sensed information is often valid for a limited period of time, the capability of smartphone users to execute sensing tasks largely depends on their mobility pattern, which is often uncertain. For this reason, in this paper, we propose IncentMe, a framework that solves this core issue by leveraging game-theoretical reverse auction mechanism design. After demonstrating that the proposed problem is NP-hard, we derive two mechanisms that are parallelizable and achieve higher approximation ratio than existing work. IncentMe has been extensively evaluated on a road traffic monitoring application implemented using mobility traces of taxi cabs in San Francisco, Rome, and Beijing. Results demonstrate that the mechanisms in IncentMe outperform the state of the art work by improving the efficiency in recruiting participants by 30 percent.

Allo scopo di analizzare le domande a risposta aperta e i saggi brevi, e più in particolare, di riflettere su come costruire e valutare in modo valido e attendibile un compito di scrittura è necessario chiarire la differenza tra saggio breve e componimento. Questo articolo illustra i principali problemi che è possibile incontrare nell'uso di saggi brevi e per misurare i livelli di apprendimento in differenti ambiti disciplinari, mantenendo questa distinzione. Inoltre, l'articolo offre una serie di suggerimenti operativi per migliorare l'uso della tipologia di domande a risposta aperta argomentata. In order to analyze open-ended questions and essay items and, more particularly, to construct a good writing task and mark students’ answers in a valid and reliable way, one should firstly make a clear distinction between an essay test and the assessment of writing itself. This article illustrates the main issues in adopting essay tests to assess writing skills and to measure educational achievement in different subjects, maintaining this distinction. Furthermore, it offers a series of practical suggestions for improving the use of essay-type questions.