Laboratoire d'Informatique de Grenoble

Semantic analysis of programs is essential in optimizing compilers and program verification systems. It encompasses data flow analysis, data type determination, generation of approximate invariant assertions, etc.

The model of abstract interpretation of programs developed by Cousot and Cousot [2nd ISOP, 1976], Cousot and Cousot [POPL 1977] and Cousot [PhD thesis 1978] is applied to the static determination of linear equality or inequality invariant relations among numerical variables of programs.

After years of research on ontology matching, it is reasonable to consider several questions: is the field of ontology matching still making progress? Is this progress significant enough to pursue further research? If so, what are the particularly promising directions? To answer these questions, we review the state of the art of ontology matching and analyze the results of recent ontology matching evaluations. These results show a measurable improvement in the field, the speed of which is albeit slowing down. We conjecture that significant improvements can be obtained only by addressing important challenges for ontology matching. We present such challenges with insights on how to approach them, thereby aiming to direct research into the most promising tracks and to facilitate the progress of the field.

By 2050, about one third of the French population will be over 65. Our laboratory's current research focuses on the monitoring of elderly people at home, to detect a loss of autonomy as early as possible. Our aim is to quantify criteria such as the international activities of daily living (ADL) or the French Autonomie Gerontologie Groupes Iso-Ressources (AGGIR) scales, by automatically classifying the different ADL performed by the subject during the day. A Health Smart Home is used for this. Our Health Smart Home includes, in a real flat, infrared presence sensors (location), door contacts (to control the use of some facilities), temperature and hygrometry sensor in the bathroom, and microphones (sound classification and speech recognition). A wearable kinematic sensor also informs postural transitions (using pattern recognition) and walk periods (frequency analysis). This data collected from the various sensors are then used to classify each temporal frame into one of the ADL that was previously acquired (seven activities: hygiene, toilet use, eating, resting, sleeping, communication, and dressing/undressing). This is done using support vector machines. We performed a 1-h experimentation with 13 young and healthy subjects to determine the models of the different activities, and then we tested the classification algorithm (cross validation) with real data.

International audience

We present here the final results of the first spectropolarimetric survey of a small sample of active M dwarfs, aimed at providing observational constraints on dynamo action on both sides of the full-convection threshold (spectral type M4). Our two previous studies were focused on early and mid M dwarfs. The present paper examines 11 fully convective late M dwarfs (spectral types M5-M8). Tomographic imaging techniques were applied to time-series of circularly polarized profiles of six stars, in order to infer their large-scale magnetic topologies. For three other stars we could not produce such magnetic maps, because of low variability of the Stokes V signatures, but were able to derive some properties of the magnetic fields.

Distributed computing is a very broad and active research area comprising fields such as cluster computing, computational grids, desktop grids and peer-to-peer (P2P) systems. Unfortunately, it is often impossible to obtain theoretical or analytical results to compare the performance of algorithms targeting such systems. One possibility is to conduct large numbers of back-to-back experiments on real platforms. While this is possible on tightly-coupled platforms, it is infeasible on modern distributed platforms as experiments are labor-intensive and results typically not reproducible. Consequently, one must resort to simulations, which enable reproducible results and also make it possible to explore wide ranges of platform and application scenarios. In this paper we describe the SimGrid framework, a simulation-based framework for evaluating cluster, grid and P2P algorithms and heuristics. This paper focuses on SimGrid v3, which greatly improves on previous versions thanks to a novel and validated modular simulation engine that achieves higher simulation speed without hindering simulation accuracy. Also, two new user interfaces were added to broaden the targeted research community. After surveying existing tools and methodologies we describe the key features and benefits of SimGrid.

The accurate detection and classification of moving objects is a critical aspect of advanced driver assistance systems. We believe that by including the object classification from multiple sensor detections as a key component of the object's representation and the perception process, we can improve the perceived model of the environment. First, we define a composite object representation to include class information in the core object's description. Second, we propose a complete perception fusion architecture based on the evidential framework to solve the detection and tracking of moving objects problem by integrating the composite representation and uncertainty management. Finally, we integrate our fusion approach in a real-time application inside a vehicle demonstrator from the interactIVe IP European project, which includes three main sensors: radar, lidar, and camera. We test our fusion approach using real data from different driving scenarios and focusing on four objects of interest: pedestrian, bike, car, and truck.

To support the large and various applications generated by the Internet of Things (IoT), Fog Computing was introduced to complement the Cloud Computing and offer Cloud-like services at the edge of the network with low latency and real-time responses. Large-scale, geographical distribution, and heterogeneity of edge computational nodes make service placement in such infrastructure a challenging issue. Diversity of user expectations and IoT devices characteristics also complicate the deployment problem. This article presents a survey of current research conducted on Service Placement Problem (SPP) in the Fog/Edge Computing. Based on a new classification scheme, a categorization of current proposals is given and identified issues and challenges are discussed.

Users and Internet service providers (ISPs) are constantly affected by denial-of-service (DoS) attacks. This cyber threat continues to grow even with the development of new protection technologies. Developing mechanisms to detect this threat is a current challenge in network security. This article presents a machine learning- (ML-) based DoS detection system. The proposed approach makes inferences based on signatures previously extracted from samples of network traffic. The experiments were performed using four modern benchmark datasets. The results show an online detection rate (DR) of attacks above 96%, with high precision (PREC) and low false alarm rate (FAR) using a sampling rate (SR) of 20% of network traffic.

Bike-sharing systems are becoming important for urban transportation. In these systems, users arrive at a station, pick up a bike, use it for a while, and then return it to another station of their choice. Each station has a finite capacity: it cannot host more bikes than its capacity. We propose a stochastic model of an homogeneous bike-sharing system and study the effect of the randomness of user choices on the number of problematic stations, i.e., stations that, at a given time, have no bikes available or no available spots for bikes to be returned to. We quantify the influence of the station capacities, and we compute the fleet size that is optimal in terms of minimizing the proportion of problematic stations. Even in a homogeneous city, the system exhibits a poor performance: the minimal proportion of problematic stations is of the order of the inverse of the capacity. We show that simple incentives, such as suggesting users to return to the least loaded station among two stations, improve the situation by an exponential factor. We also compute the rate at which bikes have to be redistributed by trucks for a given quality of service. This rate is of the order of the inverse of the station capacity. For all cases considered, the fleet size that corresponds to the best performance is half of the total number of spots plus a few more, the value of the few more can be computed in closed-form as a function of the system parameters. It corresponds to the average number of bikes in circulation.

Byzantine Fault Tolerant state machine replication (BFT) protocols are replication protocols that tolerate arbitrary faults of a fraction of the replicas. Although significant efforts have been recently made, existing BFT protocols do not provide acceptable performance when faults occur. As we show in this paper, this comes from the fact that all existing BFT protocols targeting high throughput use a special replica, called the primary, which indicates to other replicas the order in which requests should be processed. This primary can be smartly malicious and degrade the performance of the system without being detected by correct replicas. In this paper, we propose a new approach, called RBFT for Redundant-BFT: we execute multiple instances of the same BFT protocol, each with a primary replica executing on a different machine. All the instances order the requests, but only the requests ordered by one of the instances, called the master instance, are actually executed. The performance of the different instances is closely monitored, in order to check that the master instance provides adequate performance. If that is not the case, the primary replica of the master instance is considered malicious and replaced. We implemented RBFT and compared its performance to that of other existing robust protocols. Our evaluation shows that RBFT achieves similar performance as the most robust protocols when there is no failure and that, under faults, its maximum performance degradation is about 3%, whereas it is at least equal to 78% for existing protocols.

We study in this paper the problem of jointly clustering and learning\nrepresentations. As several previous studies have shown, learning\nrepresentations that are both faithful to the data to be clustered and adapted\nto the clustering algorithm can lead to better clustering performance, all the\nmore so that the two tasks are performed jointly. We propose here such an\napproach for $k$-Means clustering based on a continuous reparametrization of\nthe objective function that leads to a truly joint solution. The behavior of\nour approach is illustrated on various datasets showing its efficacy in\nlearning representations for objects while clustering them.\n

In this paper, we propose IoTChain, a combination of the OSCAR architecture [1] and the ACE authorization framework [2] to provide an E2E solution for the secure authorized access to IoT resources. IoTChain consists of two components, an authorization blockchain based on the ACE framework and the OSCAR object security model, extended with a group key scheme. The blockchain provides a flexible and trustless way to handle authorization while OSCAR uses the public ledger to set up multicast groups for authorized clients. To evaluate the feasibility of our architecture, we have implemented the authorization blockchain on top of a private Ethereum network. We report on several experiments that assess the performance of different architecture components.

Pansharpening aims to fuse a multispectral (MS) image with an associated panchromatic (PAN) image, producing a composite image with the spectral resolution of the former and the spatial resolution of the latter. Traditional pansharpening methods can be ascribed to a unified detail injection context, which views the injected MS details as the integration of PAN details and bandwise injection gains. In this paper, we design a new detail injection based convolutional neural network (DiCNN) framework for pansharpening with the MS details being directly formulated in end-to-end manners, where the first detail injection based CNN (DiCNN1) mines MS details through the PAN image and the MS image, and the second one (DiCNN2) utilizes only the PAN image. The main advantage of the proposed DiCNNs is that they provide explicit physical interpretations and can achieve fast convergence while achieving high pansharpening quality. Furthermore, the effectiveness of the proposed approaches is also analyzed from a relatively theoretical point of view. Our methods are evaluated via experiments on real MS image datasets, achieving excellent performance when compared to other state-of-the-art methods.

We present initial results from time-series imaging at infrared wavelengths of 0.9 deg^2 in the Orion Nebula Cluster (ONC). During Fall 2009 we obtained 81 epochs of Spitzer 3.6 and 4.5 μm data over 40 consecutive days. We extracted light curves with ~3% photometric accuracy for ~2000 ONC members ranging from several solar masses down to well below the hydrogen-burning mass limit. For many of the stars, we also have time-series photometry obtained at optical (I_c) and/or near-infrared (JK_s ) wavelengths. Our data set can be mined to determine stellar rotation periods, identify new pre-main-sequence eclipsing binaries, search for new substellar Orion members, and help better determine the frequency of circumstellar disks as a function of stellar mass in the ONC. Our primary focus is the unique ability of 3.6 and 4.5 μm variability information to improve our understanding of inner disk processes and structure in the Class I and II young stellar objects (YSOs). In this paper, we provide a brief overview of the YSOVAR Orion data obtained in Fall 2009 and highlight our light curves for AA-Tau analogs—YSOs with narrow dips in flux, most probably due to disk density structures passing through our line of sight. Detailed follow-up observations are needed in order to better quantify the nature of the obscuring bodies and what this implies for the structure of the inner disks of YSOs.

Medical image segmentation is an important tool for current clinical applications. It is the backbone of numerous clinical diagnosis methods, oncological treatments and computer-integrated surgeries. A new class of machine learning algorithm, deep learning algorithms, outperforms the results of classical segmentation in terms of accuracy. However, these techniques are complex and can have a high range of variability, calling the reproducibility of the results into question. In this article, through a literature review, we propose an original overview of the sources of variability to better understand the challenges and issues of reproducibility related to deep learning for medical image segmentation. Finally, we propose 3 main recommendations to address these potential issues: (1) an adequate description of the framework of deep learning, (2) a suitable analysis of the different sources of variability in the framework of deep learning, and (3) an efficient system for evaluating the segmentation results.

Abstract — Most of present work for autonomous navigation in dynamic environment doesn’t take into account the dynamics of the obstacles or the limits of the perception system. To face these problems we applied the Probabilistic Velocity Obstacle (PV O) approach [1] to a dynamic occupancy grid. The paper presents a method to estimate the probability of collision where uncertainty in position, shape and velocity of the obstacles, occlusions and limited sensor range contribute directly to the computation. A simple navigation algorithm is then presented in order to apply the method to collision avoidance and goal driven control. Simulation results show that the robot is able to adapt its behaviour to the level of available knowledge and navigate safely among obstacles with a constant linear velocity. Extensions to non-linear, non-constant velocities are proposed. I.

This paper presents a comparison of the expected lifetime for Internet of Things (IoT) devices operating in several wireless networks: the IEEE 802.15.4/e, Bluetooth low energy (BLE), the IEEE 802.11 power saving mode, the IEEE 802.11ah, and in new emerging long-range technologies, such as LoRa and SIGFOX. To compare all technologies on an equal basis, we have developed an analyzer that computes the energy consumption for a given protocol based on the power required in a given state (Sleep, Idle, Tx, and Rx) and the duration of each state. We consider the case of an energy constrained node that uploads data to a sink, analyzing the physical (PHY) layer under medium access control (MAC) constraints, and assuming IPv6 traffic whenever possible. This paper considers the energy spent in retransmissions due to corrupted frames and collisions as well as the impact of imperfect clocks. The comparison shows that the BLE offers the best lifetime for all traffic intensities in its capacity range. LoRa achieves long lifetimes behind 802.15.4 and BLE for ultra low traffic intensity; SIGFOX only matches LoRa for very small data sizes. Moreover, considering the energy consumption due to retransmissions of lost data packets only decreases the lifetimes without changing their relative ranking. We believe that these comparisons will give all users of IoT technologies indications about the technology that best fits their needs from the energy consumption point of view. Our analyzer will also help IoT network designers to select the right MAC parameters to optimize the energy consumption for a given application.

This paper proposes a first attempt to build an end-to-end speech-to-text\ntranslation system, which does not use source language transcription during\nlearning or decoding. We propose a model for direct speech-to-text translation,\nwhich gives promising results on a small French-English synthetic corpus.\nRelaxing the need for source language transcription would drastically change\nthe data collection methodology in speech translation, especially in\nunder-resourced scenarios. For instance, in the former project DARPA TRANSTAC\n(speech translation from spoken Arabic dialects), a large effort was devoted to\nthe collection of speech transcripts (and a prerequisite to obtain transcripts\nwas often a detailed transcription guide for languages with little standardized\nspelling). Now, if end-to-end approaches for speech-to-text translation are\nsuccessful, one might consider collecting data by asking bilingual speakers to\ndirectly utter speech in the source language from target language text\nutterances. Such an approach has the advantage to be applicable to any\nunwritten (source) language.\n