Agent-Based Models (ABM) are being increasingly used to evaluate urban systems, urban policies and environmental impacts. One prerequisite for using the ABM framework consists of generating a synthetic population representative of the actual population, featuring the appropriate attributes with respect to model objectives. A precise spatial positioning of the synthetic population agents is often key to ensuring ABM modeling quality. This paper considers the problem of allocating synthetic population agents to a finer spatial scale. Such an allocation process is performed from a higher-level statistical area where a synthetic population can be generated, i.e. a container statistical area (CSA), to several nested non-overlapping elementary statistical areas (ESA), where only marginals are available. This allocation step relies not only on common attributes between CSA and ESA, but also on additional discriminatory attributes, i.e. attributes of interest, estimated from external data sources. The case study examined herein is based on French census and fiscal data. Common attributes include 8 socio-demographic variables, totaling 17 modalities. An additional attribute of interest, i.e. income, has also been added. The allocation problem at hand is modeled as an integer quadratic programming problem. An exact algorithm is first applied to solve the problem; the applicability of this algorithm proves to be limited to small-size synthetic populations. A heuristic is proposed to handle the allocation of larger-size synthetic populations. Tests carried out on the case study show that this heuristic yields near optimal solutions; it is also computationally efficient and may fulfill the needs of a majority of users.

Energy moderation of the road transportation sector is required to limit climate change and to preserve resources. This work is focused on the moderation of vehicle consumption by optimizing the speed policy along an itinerary while taking into account vehicle dynamics, driver visibility and the road's longitudinal profile. First, a criterion is proposed in order to detect speed policies that are impeding drivers' eco-driving ability. Then, an energy evaluation is carried out and an optimization is proposed. A numerical application is performed on a speed limiting point with 20 usage cases and 5 longitudinal slope values. In the hypothesis of a longitudinal slope of zero, energy savings of 27.7 liter per day could be realized by a speed sign displacement of only 153.6 m. Potential energy savings can increase to up to 308.4 L per day for a −4% slope case, or up to 70.5 L per day for an ordinary −2% slope, with a sign displacement of only 391.5 m. This results in a total of 771,975 L of fuel savings over a 30 year infrastructure life cycle period. Therefore a methodology has been developed to help road managers optimize their speed policies with the aim of moderating vehicle consumption.

The objective of this study was to produce, by an enzymatic hydrolysis process at a pilot scale, a saithe (Pollachius virens) hydrolysate with a high antioxidant activity. Design of experiment methodology, based on laboratory-scale experiments, was used to obtain a behavioral reduced model that allows one to determine the optimal operating conditions maximizing the antioxidant activity. Two specifications were studied: the degree of hydrolysis and the antioxidant activity. The effects of the following hydrolysis parameters (temperature, pH, enzyme concentration, and operating time) were studied and presented as response surfaces. From these results, a multifactorial optimization was performed and the Pareto optimal set of efficient solutions was evaluated. The optimal conditions were tested at laboratory scale and then validated by comparison with tests carried out on a pilot plant.

[...]

[...]

[...]

Seaweeds are potentially excellent sources of bio- active metabolites that could represent useful leads in the de- velopment of new functional ingredients in pharmaceutical and cosmetic industries. In the last decade, new marine bioprocess technologies have allowed the isolation of sub- stances with biological properties. The brown alga Sargassum muticum (Yendo) Fensholt (Ochrophyta) was en- zymatically hydrolyzed to prepare water-soluble extracts by using six different commercially available carbohydrate- degrading enzymes and two proteases. Evaluation of 2,2- diphenyl-1-picrylhydrazyl (DPPH) and ferric reducing antiox- idant power (FRAP) antioxidant, tyrosinase, elastase, and bio- film inhibition, antibacterial and antiviral activities as well as evaluation of cytotoxicity were realized for each extract. Total phenolic content was measured for extract characterization, and solid-phase extraction was useful to purify the enzymatic extract. Soluble total phenolic content of S. muticum Viscozyme extract was highest with 6.4% of dry weight. Enzymatic Celluclast and Viscozyme extracts had the lowest value of DPPH IC50 indicating a strong antiradical activity, 0.6 mg mL−1, in comparison with other enzymes. The ferric reducing antioxidant power ranged between 48.7 μM Fe 2+ Eq, digested with Viscozyme, and 60.8 μM Fe 2+ Eq, digested with Amyloglucosidase. Tyrosinase inhibition activity of S. muticum Neutrase extract was 41.3% higher compared to other enzymes. Elastase inhibition activity of S. muticum Shearzyme extract had highest activity (32.8%). All enzymat- ic extracts showed no cytotoxic effect towards the kidney Vero cells. Meanwhile, only S. muticum Neutrase and Alcalase ex- tracts exhibited potential antiviral activity. In addition, S. muticum Viscozyme and Shearzyme extracts showed prom- ising activity in suppressing the biofilm formation against Pseudomonas aeruginosa and Escherichia coli, respectively. Purification of S. muticum Viscozyme extracts by solid-phase extraction managed to concentrate the phenolic content and improve the bioactivity. These results indicate the promising potential of enzyme-assisted followed by solid-phase extrac- tion in recovering phenolic content and in improving its bioactivity.

[...]

[...]

[...]

[...]

[...]

[...]

[...]

[...]

[...]

[...]

[...]

The photoadaptive responses of macroalgal communities to abiotic stresses have been studied, and a number of UV-absorbing molecules have been identified. Among these compounds, photoprotective compounds such as mycosporine-like amino acids and carotenoids have been isolated from various red macroalgal species. However, several substances still need to be characterized. We describe the preparation of photoprotective extracts obtained from Solieria chordalis. Two solvents, 2-octyl dodecanol and octyldodecyl ester of L-pyrrolidone carboxylic acid, were selected based on their cosmetic functions for performing an ultrasound-assisted extraction. The efficiency of extraction was monitored by spectrophotometry and in vitro photoprotective activity measurements. 2-Octyl dodecanol and octyldodecyl ester of L-pyrrolidone carboxylic acid extracts showed maximum absorption wavelengths ranging from 280 to 340 nm and 270 to 350 nm, respectively. The anti UV-B capacity for protecting a synthetic chlorophyll solution was assessed by measuring its pseudo first-order degradation kinetics at room temperature. Under irradiation at 312 nm, chlorophyll introduced in the 2-octyl dodecanol S. chordalis extract showed the slowest degradation kinetics with a half-life t(1/2) of 121.0 min. Several compounds were detected in the seaweed

[...]

[...]

Proliferation of the introduced brown macro alga Sargassum muticum is known as a natural and hard to control phenomenon occurring along the Atlantic coasts. The phenomenon causes serious troubles for local ecosystems including the alteration of ecosystem structure, the reduction in indigenous biodiversity and economic losses (tourism, aquaculture). However, despite the serious troubles caused by S. muticum, this species contains highly remarkable bioactive metabolites. This macro alga is at present under-exploited and the valorization of its metabolites to give a positive value to this seaweed could be a solution of ecosystemic service. Biorefinery process could be one solution to valorize S. muticum. Comprehensive knowledge concerning the biochemical composition of S. muticum and the impact of environmental factors, particularly seasons, on its composition is a prerequisite before its valorization. In this study, the biochemical composition of S. muticum was evaluated by using classic colorimetric methods based on chemical analysis and Fourier Transform Infra-Red spectrometry (FTIR), and was used as a rapid and safe method that could bring advantages in screening studies and a more comprehensive management and use of seaweed products. Our results are globally in accordance, notably for phenolic compounds, showing the relevance of the use of infrared spectrometry. Moreover, based on the absorption bands of some specific and valuable compounds shown by FTIR, there was a seasonal variation in the polysaccharides, i.e., uronic acids and sulphated compounds, together with phenolic contents of S. muticum.

Ultrafiltration membranes can be advantageously used to improve the bioactivity of a saithe protein hydrolysate containing peptides having a size lower than 7 kDa, by fractionating or concentrating some specific molecular weight peptide classes. The aim of this study is to evaluate the behaviour of a 4 kDa membrane in modified polyethersulfone in conditions close to industrial exploitation. This work consists in evaluating the effect of the initial peptide content (30, 90 and 150 mg/mL), the transmembrane pressure (10 and 30 bar), in total recycling modes and the volume reduction factor (until 5) on membrane performances, in terms of permeation flux and selectivity. The peptide content and mass repartition profiles are measured by size exclusion chromatography. It is observed that the pressure and the peptide content, which can act as complementary or antagonist factors, can be adjusted to modify the transmission of peptides and consequently the apparent molecular weight cut-off (MWCO) of the membrane. These results have been directly applied to study the refining of a peptidic fraction lower than 1 kDa having a potential bioactivity (i.e. antioxidant and hypotensive). The change in refining performances with the volume reduction factor (VRF) are discussed, including the recovery yield and the purity of the interesting fraction in permeate and the energy requirements.

Sargassum muticum is an invasive macroalga that uncontrollable widely grows on the European Atlantic coasts. This macroalga is particularly rich in phenolics (phlorotan-nins) presenting biological activities such as antioxidant, radical-scavenging, antibacte-rial, photoprotective properties, etc. Among the diversity of metabolites produced by the species, phenolic compounds could constitute one of the potential ways of valorization of Sargassum muticum.For the needs of the study, algae were harvested along a latitudinal gradient in Norway, Ireland, France, Spain and Portugal.After extraction, the highest phenolic contents and bioactivities were measured for the samples coming from Norway and Portugal. Consequently, both these countries have been retained in the present work.The extracts are composed of a “pool” of phenolic compounds of molecular weight varying, according to the literature, from 126 Da (phloroglucinol) to 650 kDa. But the most common compounds are comprised between 10 and 100 kDa. With the aim of concentrating and eventually fractionating (i.e. isolating the most active fractions) the phenolic extracts of Sargassum muticum, a few membranes of Molecular Weight Cut-Off 10 kDa were chosen. These membranes were tested by using different technologies: ultrafiltration, centrifugal tube and dialysis.Results show that there is no significant activity in the permeates signifying that the retention rate is about 100 % for the active fractions. Concerning the retentates, results show clearly an increase of the total phenolic contents as well as antioxidant and radical scavenging activities. The best performances are obtained with the centrifugal tubes using 10 kDa membranes.

A comparative study between "alternative" extraction processes such as centrifugal partition extraction (CPE), supercritical fluid extraction (SFE) and pressurized liquid extraction (PLE) and classical solid/liquid used in the laboratory are currently focusing on the efficiency (selectivity and productivity) to obtain bioactive phenolic compounds from the phaeophyte Sargassum muticum model. The choice of the best process was based on several measurements: (i) the total phenolic content measured by the colorimetric Folin-Ciocalteu assay, (ii) radical scavenger and antioxidant activities assessed by the DPPH (2,2-diphenyl-1-picrylhydrazyl) radical scavenging assay, and the β-carotene bleaching method and finally (iii) the method productivity. Irrespective of the solvent used in the processes, alternative methods are always sharply more effective than classical ones. With the exception of SFE which does not allow extracting the totality of the active phenolic compounds, two of the other extraction methods were particularly promising. First, CPE afforded the most important yields in concentrated phenolic compounds (PC) (22.90±0.65% DW) also displaying the best activities (0.52±0.02 and 0.58±0.19 mg/mL for IC50 and AAC700, respectively). Secondly, PLE using an EtOH:water mixture 75:25 (v/v) allowed a good PC extraction (10.18±0.25% DW) with huge efficiency. Despite a lesser activity of the extracts (0.77±0.01 and 1.59±0.15 mg/mL for IC50 and AAC700, respectively) PLE is a green process and potentially complies European norms requirements for the prospective valorization of phenolic compounds from S. muticum in Brittany.

Sargassum muticum, an invasive brown macroalga presently distributed along European Atlantic coasts from southern Portugal to the south coast of Norway, was studied on a large geographical scale for its production of phenolic compounds with potential industrial applications and their chemical and biological activities. S. muticum can produce high biomass in Europe, which could be exploited to supply such compounds. S. muticum was collected in Portugal, Spain, France, Ireland and Norway (three sites/country) to examine the effect of the latitudinal cline and related environmental factors. Assays focused particularly on polyphenols and their activities. Crude acetone-water extracts were purified using solid phase extraction (SPE) and antioxidant and antimicrobial activities of crude extracts and semi-purified fractions measured. Total phenolic content was assessed by colorimetric Folin-Ciocalteu assay and reactive oxygen species activities by 2,2-diphenyl-1-picrylhydrazyl, reducing power, β-carotene bleaching method and xanthine oxidase assay. Antibacterial activities were tested on terrestrial and marine strains to evaluate potential use in biomedical and aquaculture fields. Purified active phlorotannins, isolated by SPE, were identified using NMR. Phenolic contents differ clearly among countries and among sites within countries. Quality did not change between countries, however, although there were some slight differences in phlorethol type. Additionally, some fractions, especially from the extreme north and south, were very active. We discuss this in relation to environmental conditions and the interest of these compounds. S. muticum represents a potential natural source of bioactive compounds and its collection could offer an interesting opportunity for the future management of this species in Europe.

The production by enzymatic treatment of fish protein hydrolysates (FPH) is a promising route to add value to fisheries proteinic co-products (fish frames, heads etc.). Indeed, FPH possess good nutritional properties and biological activities for food and feed uses. Pressure-driven membrane separations such as ultrafiltration (UF) and nanofiltration (NF) can be used after the hydrolysis to increase the specific activities of the FPH. This paper discusses the impact of a two-step UF/NF process producing four different fractions on two industrial FPH with different hydrolysis degrees. Fractionation is carried out in “realistic” conditions for an industrial process, on highly concentrated FPH solutions (about 100 g of dry matter/L) at a high volume reduction factor. For each step, UF or NF, the variation of the permeation flux in the course of the fractionation is discussed according to the FPH hydrolysis degree and the membranes cut-offs. The values of performance indicators defined in terms of nitrogen content are also examined, including the concentration factor (CF), the relative recovery in the retentate (ηR) and the mean and final retention factors (RFm and RFf). Computed values of these indicators are validated through the setting of volume and mass balances around each step. The impact of fractionation on the FPH peptidic population is shown. Peptidic populations are described in terms of chromatographic profiles (SEC–FPLC). The UF fractionation produces a permeate enriched with respect to the FPH smaller than a molecular weight of about 600–750 Dalton, and a retentate enriched in large peptides (above the same MW). A similar behaviour is found for the NF fractionation. Comparing the impact of the UF fractionation on the two hydrolysates allows to conclude that the membrane cut-off is well-suited when comprised between the MWs of the biggest and the most abundant peptides in the FPH.

Five organic tubular nanofiltration membranes of various material and molecular weight cut-off (MWCO) were tested for the enrichment of a fish (saithe, Pollachius virens) hydrolysate in low molecular weight antioxidant peptides (< 2 kDa). The effect of membrane material and MWCO on permeation flux, hydrolysate adsorption extent, membrane cleaning and peptides retention was investigated in low fouling conditions in order to limit concentration polarization effects. Water permeation fluxes appeared sensitive both to the MWCO and the membrane material. Membranes in polyethersulfone (PES), a highly hydrophobic material, exhibited lower fluxes than modified polyethersulfone (m-PES) and polysulfone (PS) membranes. The 4 kDa m-PES membrane was found to be the most adapted to the objective in terms of flux and retention factor. The 4 kDa m-PES and 8 kDa PS membranes could also be used in series to recover medium-size peptides. (c) 2008 Elsevier Ltd. All rights reserved.

[...]

[...]

Two recent techniques based on chemical footprinting analysis, HRMAS NMR and FTIR spectroscopy, were tested on a brown macroalgal model. These powerful and easily-to-use techniques allowed us to discriminate Sargassum muticum specimens collected in five different countries along Atlantic coasts, from Portugal to Norway. HRMAS NMR and FTIR permitted the obtaining of an overview of metabolites produced by the alga. Based on spectra analysis, results allowed us to successfully group the samples according to their geographical origin. HRMAS NMR and FTIR spectroscopy respectively point out the relation between the geographical localization and the chemical composition and demonstrated macromolecules variations regarding to environmental stress. Then, our results are discussed in regard of the powerful of these techniques together with the variability of the main molecules produced by Sargassum muticum along the Atlantic coasts.

BACKGROUND: Numerous studies have demonstrated that in vitro controlled enzymatic hydrolysis of fish and shellfish proteins leads to bioactive peptides. Ultrafiltration (UF) and/or nanofiltration (NF) can be used to refine hydrolysates and also to fractionate them in order to obtain a peptide population enriched in selected sizes. This study was designed to highlight the impact of controlled UF and NF on the stability of biological activities of an industrial fish protein hydrolysate (FPH) and to understand whether fractionation could improve its content in bioactive peptides. RESULTS: The starting fish protein hydrolysate exhibited a balanced amino acid composition, a reproducible molecular weight (MW) profile, and a low sodium chloride content, allowing the study of its biological activity. Successive fractionation on UF and NF membranes allowed concentration of peptides of selected sizes, without, however, carrying out sharp separations, some MW classes being found in several fractions. Peptides containing Pro, Hyp, Asp and Glu were concentrated in the UF and NF retentates compared to the unfractionated hydrolysate and UF permeate, respectively. Gastrin/cholecystokinin-like peptides were present in the starting FPH, UF and NF fractions, but fractionation did not increase their concentration. In contrast, quantification of calcitonin gene-related peptide (CGRP)-like peptides demonstrated an increase in CGRP-like activities in the UF permeate, relative to the starting FPH. The starting hydrolysate also showed a potent antioxidant and radical scavenging activity, and a moderate angiotensin-converting enzyme (ACE)-1 inhibitory activity, which were not increased by UF and NF fractionation. CONCLUSION: Fractionation of an FPH using membrane separation, with a molecular weight cut-off adapted to the peptide composition, may provide an effective means to concentrate CGRP-like peptides and peptides enriched in selected amino acids. The peptide size distribution observed after UF and NF fractionation demonstrates that it is misleading to characterize the fractions obtained by membrane filtration according to the MW cut-off of the membrane only, as is currently done in the literature. Copyright © 2010 Society of Chemical Industry

This paper deals with the identification of physical parameters of a one-degree-of-freedom electromechanical system that operates in closed loop. Two models are considered: the inverse dynamic model which is linear in relation to the physical parameters to be identified and the direct dynamic model which is linear in relation to a nonlinear combination of the physical parameters. Three methods are considered and compared: the traditional method which makes use of the inverse dynamic model, tailormade data prefiltering and the least-squares method, an instrumental variable approach which makes use of the direct and inverse dynamic models and another instrumental variable approach which makes use of the direct dynamic model only. The experimental results show that the instrumental variable method based on the use of the inverse and direct dynamic models seems to be more appropriate than the two others because it is robust against noise and the physical parameters are directly identified.

Dans cet article, une présentation des propriétés des polynômes de Tchebychev et une formulation dérivée utilisant ces polynômes sont effectuées. La méthode d'identification présentée consiste à appliquer une transformation linéaire sur le système d'équations qui régit le processus. Pour cela, l'opérateur mathématique utilisé s'appuie sur une décomposition du signal dans une base formée de polynômes orthogonaux. Nous montrons que cette projection se comporte comme un filtre passe-bande, de telle sorte qu'une seule opération est nécessaire pour le pré-traitement des données avant le processus d'identification, à savoir choisir la dimension de la base des polynômes Tchebychev. Une identification expérimentale en boucle fermée d'un robot à 2 axes avec cette méthode est effectuée dans une dernière partie. Les résultats obtenus sont comparés et analysés avec d'autres techniques utilisées en robotique.

This paper deals with joint stiffness identification with only actual motor force/torque data instead of motor andload positions. The parameters are estimated by using the DIDIM method which needs only input data. This method was previously validated on a 6 DOF rigid robot and is now extended to flexible systems. The criterion to be minimized is the quadratic error between the measured actual motor force/torque and the simulated one. The optimal parameters are calculated with the Nelder Mead simplex algorithm. An experimental setup exhibits the experimental identification results and shows the effectiveness of our approach.

This paper addresses the important topic of joint flexibility identification. Three dynamic models depending on measurements availability are compared. The parameters are estimated by using the ordinary least squares of an over linear system obtained from the sampling of the dynamic model along a closed loop tracking trajectory. An experimental setup exhibits the experimental identification results.

Ce papier présente une nouvelle méthode à erreur de sortie en deux étapes pour l'identification en boucle fermée des paramètres inertiels, de frottements et de raideur d'un axe de robot avec flexibilité articulaire localisée. Cette méthode utilise uniquement l'effort et la position du moteur, sans aucune mesure de flexibilité. La première étape identifie la fréquence du mode rigide asservi et la fréquence du mode flexible à entrée bloquée, en utilisant un modèle haute fréquence double intégrateur du robot. L'identification utilise une méthode à erreur de sortie sur la position du moteur. Ces deux paramètres de fréquence optimaux sont ensuite utilisés dans la deuxième étape pour l'identification des paramètres dynamiques du robot par une méthode à erreur de sortie sur l'effort moteur. Cette méthode récente (DIDIM) qui utilise les modèles d'identification directe et inverse du robot permet d'obtenir une convergence des paramètres très rapide et insensible aux conditions initiales. Une validation expérimentale sur un robot un axe montre l'efficacité de la nouvelle méthode comparée à deux méthodes à erreur de sortie classiques.

[...]

This paper addresses the topic of robot identification. The usual identification method makes use of the inverse dynamic model (IDM) and the least squares (LS) technique while robot is tracking exciting tra- jectories. Assuming an appropriate bandpass filtering, good results can be obtained. However, the users are in doubt whether the columns of the observation matrix (the regressors) are uncorrelated (exo- genous) or correlated (endogenous) with the error terms. The exogeneity condition is rarely verified in a formal way whereas it is a fundamental condition to obtain unbiased LS estimates. In Econometrics, the Durbin-Wu-Hausman test (DWH-test) is a formal statistic for investigating whether the regressors are exogenous or endogenous. However, the DWH-test cannot be straightforwardly used for robot identifi- cation because it is assumed that the set of instruments is valid. In this paper, a Revised DWH-test suitable for robot identification is proposed. The revised DWH-test validates/invalidates the instruments chosen by the user and validates the exogeneity assumption through the calculation of the QR factor- ization of the augmented observation matrix combined with a F-test if required. The experimental results obtained with a 6 degrees-of-freedom (DOF) industrial robot validate the proposed statistic.

The current transport system is not sustainable because of its high consumption of finite resources (mainly oil) and its impact on the environment (air quality, global warming, etc.). In this context, rail has several advantages: lower consumption and emissions than road and air modes, larger passenger flow, etc. That is why, many rail infrastructures are planned in the world. Today, the impact of energy is not taken into account during the project design. At the project phase of new rail infrastructures, it is nowadays important to characterize accurately the energy that will be induced by its operation phase, in addition to other more classical criteria as construction costs and travel time. Current literature consumption models used to estimate railways operation phase are obsolete or not enough accurate for taking into account the newest train or railways technologies. In this paper, an updated model of consumption for high-speed is proposed, based on experimental data obtained from full-scale tests performed on a new high-speed line. The assessment of the model is achieved by identifying train parameters and measured power consumptions for more than one hundred train routes. Perspectives are then discussed to use this updated model for accurately assess the energy impact of future railway infrastructures.

Offline robot dynamic identification methods are mostly based on the use of the inverse dynamic model, which is linear with respect to the dynamic parameters. This model is sampled while the robot is tracking reference trajectories that excite the system dynamics. This allows using linear least-squares techniques to estimate the parameters. The efficiency of this method has been proved through the experimental identification of many prototypes and industrial robots. However, this method requires the joint force/torque and position measurements and the estimate of the joint velocity and acceleration, through the bandpass filtering of the joint position at high sampling rates. The proposed new method called DIDIM requires only the joint force/torque measurement, which avoids the calculation of the velocity and acceleration by bandpass filtering of the measured position. It is a closed-loop output error method where the usual joint position output is replaced by the joint force/torque. It is based on a closed-loop simulation of the robot using the direct dynamic model, the same structure of the control law, and the same reference trajectory for both the actual and the simulated robot. The optimal parameters minimize the 2-norm of the error between the actual force/torque and the simulated force/torque. This is a nonlinear least-squares problem which is dramatically simplified using the inverse dynamic model to obtain an analytical expression of the simulated force/torque, linear in the parameters. A validation experiment on a two degree-of-freedom direct drive rigid robot shows that the new method is efficient.

Usually, the identification of the dynamic parameters of robot makes use of the inverse dynamic model which is linear with respect to the parameters. This model is sampled while the robot is tracking exciting trajectories. This allows using linear least squares (LS) techniques to estimate the parameters. The efficiency of this method has been proved through experimental identifications of a lot of prototypes and industrial robots. However, it is known that LS estimators are sensitive to outliers and leverage points. Thus, it may be helpful to verify their reliability. This is possible by using robust regressions and residual analysis. Then, we compare the results with those obtained with classical LS regression. This paper deals with this issue and introduces the experimental identification and residual analysis of an one degree of freedom (DOF) haptic interface using the Huber's estimator. To verify the pertinence of our analyses, this comparison is also performed on a medical interface consisting of a complex mechanical structure.

The identification of the dynamic parameters of robot is based on the use of the inverse dynamic model which is linear with respect to the parameters. This model is sampled while the robot is tracking exciting trajectories, in order to get an over determined linear system. The linear least squares solution of this system calculates the estimated parameters. The efficiency of this method has been proved through the experimental identification of a lot of prototypes and industrial robots. However, this method needs joint torque and position measurements and the estimation of the joint velocities and accelerations through the pass band filtering of the joint position at high sample rate. So, the observation matrix is noisy. Moreover identification process takes place when the robot is controlled by feedback. These violations of assumption imply that the LS solution is biased. The Refined Instrumental Variable (SRIV) approach deals with this problem of noisy observation matrix and can be statistically optimal. This paper focuses on this technique which will be applied to a 2 degrees of freedom (DOF) prototype developed by the IRCCyN Robotic team.

The current transport system is not sustainable because of its high consumption of finite resources (mainly oil) and its impact on the environment (air quality, global warming, etc.). In this context, rail has several advantages: lower consumption and emissions than road and air modes, larger passenger flow, etc. That is why, many rail infrastructures are planned in the world. Today, the impact of energy is not taken into account during the project design. At the project phase of new rail infrastructures, it is nowadays important to characterize accurately the energy that will be induced by its operation phase, in addition to other more classical criteria as construction costs and travel time. Current literature consumption models used to estimate railways operation phase are obsolete or not enough accurate for taking into account the newest train or railways technologies. In this paper, an updated model of consumption for high-speed is proposed, based on experimental data obtained from full-scale tests performed on a new high-speed line. The assessment of the model is achieved by identifying train parameters and measured power consumptions for more than one hundred train routes. Perspectives are then discussed to use this updated model for accurately assess the energy impact of future railway infrastructures.

This brief deals with the identification of industrial robots. The usual identification method is based on the use of the inverse dynamic identification model and the least-squares technique. Another approach is the direct and inverse dynamic identification models (DIDIM) method, which is a closed-loop output error (CLOE) method minimizing the quadratic error between the actual and simulated joint torques. In this brief, the DIDIM method is compared with the usual CLOE method, which minimizes the quadratic error between the actual and simulated joint positions. The comparison is carried out on six degrees of freedom industrial robot. The experimental results show that the DIDIM method needs three iterations to identify 60 parameters while the usual CLOE method needs 20 iterations.

This paper deals with the important topic of industrial robot identification. The usual identification method is based on the inverse dynamic identification model and the least squares technique. This method has been successfully applied on several industrial robots. Good results can be obtained, provided a well tuned derivative band-pass filtering of joint positions is used to calculate the joint velocities and accelerations. However, one cannot be sure whether or not the band-pass filtering is well tuned. An alternative is the instrumental variable (IV) method, which is robust to data filtering and is statistically optimal. In this paper, a generic IV approach suitable for robot identification is proposed. The instrument set is the inverse dynamic model built from simulated data calculated from simulation of the direct dynamic model. The simulation is based on previous estimates and assumes the same reference trajectories and the same control structure for both actual and simulated robots. Finally, gains of the simulated controller are updated according to IV estimates to obtain a valid instrument set at each step of the algorithm. The proposed approach validates the inverse and direct dynamic models simultaneously, is not sensitive to initial conditions, and converges rapidly. Experimental results obtained on a six-degrees-of-freedom industrial robot show the effectiveness of this approach: 60 dynamic parameters are identified in three itérations.

This paper deals with the important topic of rigid industrial robots identification. The usual identification method is based on the use of the inverse dynamic model and the least-squares technique. In order to obtain good results, a well-tuned derivative bandpass filtering of joint positions is needed to calculate the joint velocities and accelerations. However, we can doubt whether the bandpass filter is well-tuned or not. Another approach is the instrumental variable (IV) method which is robust to data filtering and which is statistically optimal. In this paper, an IV approach relevant for identification of rigid industrial robots is introduced. The set of instruments is the inverse dynamic model built from simulated data which are calculated from the simulation of the direct dynamic model. The simulation assumes the same reference trajectories and the same control structure for both the actual and the simulated robot and is based on the previous IV estimates. Furthermore, to obtain a rapid convergence, the gains of the simulated controller are updated according to IV estimates. Thus, the proposed approach validates the inverse and direct dynamic models simultaneously and is not sensitive to initial conditions. The experimental results obtained with a 2 degrees of freedom (DOF) planar prototype and with a 6 DOF industrial robot show the effectiveness of our approach: it is possible to identify 60 parameters in 3 iterations and in 11 s.

This paper deals with the topic of industrial robots identification. The usual identification method is based on the use of the inverse dynamic model (IDM) and least squares (LS) technique. Good results can be obtained provided that a well-tuned bandpass filtering is used. However, we are always in doubt if regressors are exogenous i.e. statistically uncorrelated with error terms. Surprisingly, in papers dealing with identification of real-world systems, exogeneity assumption is never verified whereas it is a fundamental condition to obtain unbiased estimates. In Econometrics, the Durbin-Wu-Hausman test (DWH-test) is a theoretical method for investigating whether regressors are exogenous or not. The DWH-test makes of the Two Stage Lesat Squares estimator (2SLS) and an augmented LS regression. However, this test cannot be used as is for robots identification: instruments set is supposed to be valid and restrictive statistical assumptions are made while they are quite implausible in practice. In this paper, we aim at bridging the gap between Econometrics and Control engineering practices by introducing a revisited version relevant for robots identification. An experimental validation performed on a 2 degrees of freedom (DOF) robot shows the effectiveness and the usefulness of this revisited DWH-test.

This paper deals with the topic of robots dynamics identification. In this paper, we focus on the instrumental variable (IV) technique. For robots, the set of instruments is the inverse dynamic model built from simulated data which are calculated from the integration of the direct dynamic model, assuming the same reference trajectories and the same control law structure for both actual and simulated robots. The integration of the direct dynamic model is based on previous IV estimates. This defines an iterative algorithm. Furthermore, gains of the simulated controller are updated to get a fast convergence. Experimental results obtained on a six degrees of freedom robot manufactured by Stäubli show the effectiveness of our approach: 60 dynamic parameters are estimated in only two iterations.

This paper addresses physical parameters identification of mathematical models that are linear in relation to these physical parameters. We can obtain good results with the least squares technique, provided that a well-tuned data filtering is used, and by using instrumental variable (IV) methods, which deal with the problem of noisy observation matrix. However, IV theory is based on instruments validity. In econometrics, statistical tests evaluating instruments quality have been developed. They make use of the two-stage least squares estimator and the concentration parameter introduced by Basmann. In this paper we show how to extend econometric theory to control engineering. An algorithm evaluating instruments quality is presented and experimentally validated on a two-degrees-of-freedom SCARA robot.

Energy moderation of the road transportation sector is required to limit climate change and to preserve resources. This work is focused on the moderation of vehicle consumption by optimizing the speed policy along an itinerary while taking into account vehicle dynamics, driver visibility and the road's longitudinal profile. First, a criterion is proposed in order to detect speed policies that are impeding drivers' eco-driving ability. Then, an energy evaluation is carried out and an optimization is proposed. A numerical application is performed on a speed limiting point with 20 usage cases and 5 longitudinal slope values. In the hypothesis of a longitudinal slope of zero, energy savings of 27.7 liter per day could be realized by a speed sign displacement of only 153.6 m. Potential energy savings can increase to up to 308.4 L per day for a −4% slope case, or up to 70.5 L per day for an ordinary −2% slope, with a sign displacement of only 391.5 m. This results in a total of 771,975 L of fuel savings over a 30 year infrastructure life cycle period. Therefore a methodology has been developed to help road managers optimize their speed policies with the aim of moderating vehicle consumption.

This paper deals with the identification of physical parameters of a one-degree-of-freedom electromechanical system that operates in closed loop. Two models are considered: the inverse dynamic model which is linear in relation to the physical parameters to be identified and the direct dynamic model which is linear in relation to a nonlinear combination of the physical parameters. Three methods are considered and compared: the traditional method which makes use of the inverse dynamic model, tailormade data prefiltering and the least-squares method, an instrumental variable approach which makes use of the direct and inverse dynamic models and another instrumental variable approach which makes use of the direct dynamic model only. The experimental results show that the instrumental variable method based on the use of the inverse and direct dynamic models seems to be more appropriate than the two others because it is robust against noise and the physical parameters are directly identified.

Dans cet article, une présentation des propriétés des polynômes de Tchebychev et une formulation dérivée utilisant ces polynômes sont effectuées. La méthode d'identification présentée consiste à appliquer une transformation linéaire sur le système d'équations qui régit le processus. Pour cela, l'opérateur mathématique utilisé s'appuie sur une décomposition du signal dans une base formée de polynômes orthogonaux. Nous montrons que cette projection se comporte comme un filtre passe-bande, de telle sorte qu'une seule opération est nécessaire pour le pré-traitement des données avant le processus d'identification, à savoir choisir la dimension de la base des polynômes Tchebychev. Une identification expérimentale en boucle fermée d'un robot à 2 axes avec cette méthode est effectuée dans une dernière partie. Les résultats obtenus sont comparés et analysés avec d'autres techniques utilisées en robotique.

This paper deals with joint stiffness identification with only actual motor force/torque data instead of motor andload positions. The parameters are estimated by using the DIDIM method which needs only input data. This method was previously validated on a 6 DOF rigid robot and is now extended to flexible systems. The criterion to be minimized is the quadratic error between the measured actual motor force/torque and the simulated one. The optimal parameters are calculated with the Nelder Mead simplex algorithm. An experimental setup exhibits the experimental identification results and shows the effectiveness of our approach.

This paper addresses the important topic of joint flexibility identification. Three dynamic models depending on measurements availability are compared. The parameters are estimated by using the ordinary least squares of an over linear system obtained from the sampling of the dynamic model along a closed loop tracking trajectory. An experimental setup exhibits the experimental identification results.

Ce papier présente une nouvelle méthode à erreur de sortie en deux étapes pour l'identification en boucle fermée des paramètres inertiels, de frottements et de raideur d'un axe de robot avec flexibilité articulaire localisée. Cette méthode utilise uniquement l'effort et la position du moteur, sans aucune mesure de flexibilité. La première étape identifie la fréquence du mode rigide asservi et la fréquence du mode flexible à entrée bloquée, en utilisant un modèle haute fréquence double intégrateur du robot. L'identification utilise une méthode à erreur de sortie sur la position du moteur. Ces deux paramètres de fréquence optimaux sont ensuite utilisés dans la deuxième étape pour l'identification des paramètres dynamiques du robot par une méthode à erreur de sortie sur l'effort moteur. Cette méthode récente (DIDIM) qui utilise les modèles d'identification directe et inverse du robot permet d'obtenir une convergence des paramètres très rapide et insensible aux conditions initiales. Une validation expérimentale sur un robot un axe montre l'efficacité de la nouvelle méthode comparée à deux méthodes à erreur de sortie classiques.

[...]

This paper addresses the topic of robot identification. The usual identification method makes use of the inverse dynamic model (IDM) and the least squares (LS) technique while robot is tracking exciting tra- jectories. Assuming an appropriate bandpass filtering, good results can be obtained. However, the users are in doubt whether the columns of the observation matrix (the regressors) are uncorrelated (exo- genous) or correlated (endogenous) with the error terms. The exogeneity condition is rarely verified in a formal way whereas it is a fundamental condition to obtain unbiased LS estimates. In Econometrics, the Durbin-Wu-Hausman test (DWH-test) is a formal statistic for investigating whether the regressors are exogenous or endogenous. However, the DWH-test cannot be straightforwardly used for robot identifi- cation because it is assumed that the set of instruments is valid. In this paper, a Revised DWH-test suitable for robot identification is proposed. The revised DWH-test validates/invalidates the instruments chosen by the user and validates the exogeneity assumption through the calculation of the QR factor- ization of the augmented observation matrix combined with a F-test if required. The experimental results obtained with a 6 degrees-of-freedom (DOF) industrial robot validate the proposed statistic.

The current transport system is not sustainable because of its high consumption of finite resources (mainly oil) and its impact on the environment (air quality, global warming, etc.). In this context, rail has several advantages: lower consumption and emissions than road and air modes, larger passenger flow, etc. That is why, many rail infrastructures are planned in the world. Today, the impact of energy is not taken into account during the project design. At the project phase of new rail infrastructures, it is nowadays important to characterize accurately the energy that will be induced by its operation phase, in addition to other more classical criteria as construction costs and travel time. Current literature consumption models used to estimate railways operation phase are obsolete or not enough accurate for taking into account the newest train or railways technologies. In this paper, an updated model of consumption for high-speed is proposed, based on experimental data obtained from full-scale tests performed on a new high-speed line. The assessment of the model is achieved by identifying train parameters and measured power consumptions for more than one hundred train routes. Perspectives are then discussed to use this updated model for accurately assess the energy impact of future railway infrastructures.

Offline robot dynamic identification methods are mostly based on the use of the inverse dynamic model, which is linear with respect to the dynamic parameters. This model is sampled while the robot is tracking reference trajectories that excite the system dynamics. This allows using linear least-squares techniques to estimate the parameters. The efficiency of this method has been proved through the experimental identification of many prototypes and industrial robots. However, this method requires the joint force/torque and position measurements and the estimate of the joint velocity and acceleration, through the bandpass filtering of the joint position at high sampling rates. The proposed new method called DIDIM requires only the joint force/torque measurement, which avoids the calculation of the velocity and acceleration by bandpass filtering of the measured position. It is a closed-loop output error method where the usual joint position output is replaced by the joint force/torque. It is based on a closed-loop simulation of the robot using the direct dynamic model, the same structure of the control law, and the same reference trajectory for both the actual and the simulated robot. The optimal parameters minimize the 2-norm of the error between the actual force/torque and the simulated force/torque. This is a nonlinear least-squares problem which is dramatically simplified using the inverse dynamic model to obtain an analytical expression of the simulated force/torque, linear in the parameters. A validation experiment on a two degree-of-freedom direct drive rigid robot shows that the new method is efficient.

Usually, the identification of the dynamic parameters of robot makes use of the inverse dynamic model which is linear with respect to the parameters. This model is sampled while the robot is tracking exciting trajectories. This allows using linear least squares (LS) techniques to estimate the parameters. The efficiency of this method has been proved through experimental identifications of a lot of prototypes and industrial robots. However, it is known that LS estimators are sensitive to outliers and leverage points. Thus, it may be helpful to verify their reliability. This is possible by using robust regressions and residual analysis. Then, we compare the results with those obtained with classical LS regression. This paper deals with this issue and introduces the experimental identification and residual analysis of an one degree of freedom (DOF) haptic interface using the Huber's estimator. To verify the pertinence of our analyses, this comparison is also performed on a medical interface consisting of a complex mechanical structure.

The identification of the dynamic parameters of robot is based on the use of the inverse dynamic model which is linear with respect to the parameters. This model is sampled while the robot is tracking exciting trajectories, in order to get an over determined linear system. The linear least squares solution of this system calculates the estimated parameters. The efficiency of this method has been proved through the experimental identification of a lot of prototypes and industrial robots. However, this method needs joint torque and position measurements and the estimation of the joint velocities and accelerations through the pass band filtering of the joint position at high sample rate. So, the observation matrix is noisy. Moreover identification process takes place when the robot is controlled by feedback. These violations of assumption imply that the LS solution is biased. The Refined Instrumental Variable (SRIV) approach deals with this problem of noisy observation matrix and can be statistically optimal. This paper focuses on this technique which will be applied to a 2 degrees of freedom (DOF) prototype developed by the IRCCyN Robotic team.

The current transport system is not sustainable because of its high consumption of finite resources (mainly oil) and its impact on the environment (air quality, global warming, etc.). In this context, rail has several advantages: lower consumption and emissions than road and air modes, larger passenger flow, etc. That is why, many rail infrastructures are planned in the world. Today, the impact of energy is not taken into account during the project design. At the project phase of new rail infrastructures, it is nowadays important to characterize accurately the energy that will be induced by its operation phase, in addition to other more classical criteria as construction costs and travel time. Current literature consumption models used to estimate railways operation phase are obsolete or not enough accurate for taking into account the newest train or railways technologies. In this paper, an updated model of consumption for high-speed is proposed, based on experimental data obtained from full-scale tests performed on a new high-speed line. The assessment of the model is achieved by identifying train parameters and measured power consumptions for more than one hundred train routes. Perspectives are then discussed to use this updated model for accurately assess the energy impact of future railway infrastructures.

This brief deals with the identification of industrial robots. The usual identification method is based on the use of the inverse dynamic identification model and the least-squares technique. Another approach is the direct and inverse dynamic identification models (DIDIM) method, which is a closed-loop output error (CLOE) method minimizing the quadratic error between the actual and simulated joint torques. In this brief, the DIDIM method is compared with the usual CLOE method, which minimizes the quadratic error between the actual and simulated joint positions. The comparison is carried out on six degrees of freedom industrial robot. The experimental results show that the DIDIM method needs three iterations to identify 60 parameters while the usual CLOE method needs 20 iterations.

This paper deals with the important topic of industrial robot identification. The usual identification method is based on the inverse dynamic identification model and the least squares technique. This method has been successfully applied on several industrial robots. Good results can be obtained, provided a well tuned derivative band-pass filtering of joint positions is used to calculate the joint velocities and accelerations. However, one cannot be sure whether or not the band-pass filtering is well tuned. An alternative is the instrumental variable (IV) method, which is robust to data filtering and is statistically optimal. In this paper, a generic IV approach suitable for robot identification is proposed. The instrument set is the inverse dynamic model built from simulated data calculated from simulation of the direct dynamic model. The simulation is based on previous estimates and assumes the same reference trajectories and the same control structure for both actual and simulated robots. Finally, gains of the simulated controller are updated according to IV estimates to obtain a valid instrument set at each step of the algorithm. The proposed approach validates the inverse and direct dynamic models simultaneously, is not sensitive to initial conditions, and converges rapidly. Experimental results obtained on a six-degrees-of-freedom industrial robot show the effectiveness of this approach: 60 dynamic parameters are identified in three itérations.

This paper deals with the important topic of rigid industrial robots identification. The usual identification method is based on the use of the inverse dynamic model and the least-squares technique. In order to obtain good results, a well-tuned derivative bandpass filtering of joint positions is needed to calculate the joint velocities and accelerations. However, we can doubt whether the bandpass filter is well-tuned or not. Another approach is the instrumental variable (IV) method which is robust to data filtering and which is statistically optimal. In this paper, an IV approach relevant for identification of rigid industrial robots is introduced. The set of instruments is the inverse dynamic model built from simulated data which are calculated from the simulation of the direct dynamic model. The simulation assumes the same reference trajectories and the same control structure for both the actual and the simulated robot and is based on the previous IV estimates. Furthermore, to obtain a rapid convergence, the gains of the simulated controller are updated according to IV estimates. Thus, the proposed approach validates the inverse and direct dynamic models simultaneously and is not sensitive to initial conditions. The experimental results obtained with a 2 degrees of freedom (DOF) planar prototype and with a 6 DOF industrial robot show the effectiveness of our approach: it is possible to identify 60 parameters in 3 iterations and in 11 s.

This paper deals with the topic of robots dynamics identification. In this paper, we focus on the instrumental variable (IV) technique. For robots, the set of instruments is the inverse dynamic model built from simulated data which are calculated from the integration of the direct dynamic model, assuming the same reference trajectories and the same control law structure for both actual and simulated robots. The integration of the direct dynamic model is based on previous IV estimates. This defines an iterative algorithm. Furthermore, gains of the simulated controller are updated to get a fast convergence. Experimental results obtained on a six degrees of freedom robot manufactured by Stäubli show the effectiveness of our approach: 60 dynamic parameters are estimated in only two iterations.

This paper deals with the topic of industrial robots identification. The usual identification method is based on the use of the inverse dynamic model (IDM) and least squares (LS) technique. Good results can be obtained provided that a well-tuned bandpass filtering is used. However, we are always in doubt if regressors are exogenous i.e. statistically uncorrelated with error terms. Surprisingly, in papers dealing with identification of real-world systems, exogeneity assumption is never verified whereas it is a fundamental condition to obtain unbiased estimates. In Econometrics, the Durbin-Wu-Hausman test (DWH-test) is a theoretical method for investigating whether regressors are exogenous or not. The DWH-test makes of the Two Stage Lesat Squares estimator (2SLS) and an augmented LS regression. However, this test cannot be used as is for robots identification: instruments set is supposed to be valid and restrictive statistical assumptions are made while they are quite implausible in practice. In this paper, we aim at bridging the gap between Econometrics and Control engineering practices by introducing a revisited version relevant for robots identification. An experimental validation performed on a 2 degrees of freedom (DOF) robot shows the effectiveness and the usefulness of this revisited DWH-test.

This paper addresses physical parameters identification of mathematical models that are linear in relation to these physical parameters. We can obtain good results with the least squares technique, provided that a well-tuned data filtering is used, and by using instrumental variable (IV) methods, which deal with the problem of noisy observation matrix. However, IV theory is based on instruments validity. In econometrics, statistical tests evaluating instruments quality have been developed. They make use of the two-stage least squares estimator and the concentration parameter introduced by Basmann. In this paper we show how to extend econometric theory to control engineering. An algorithm evaluating instruments quality is presented and experimentally validated on a two-degrees-of-freedom SCARA robot.

Ce papier présente une nouvelle méthode à erreur de sortie en deux étapes pour l'identification en boucle fermée des paramètres inertiels, de frottements et de raideur d'un axe de robot avec flexibilité articulaire localisée. Cette méthode utilise uniquement l'effort et la position du moteur, sans aucune mesure de flexibilité. La première étape identifie la fréquence du mode rigide asservi et la fréquence du mode flexible à entrée bloquée, en utilisant un modèle haute fréquence double intégrateur du robot. L'identification utilise une méthode à erreur de sortie sur la position du moteur. Ces deux paramètres de fréquence optimaux sont ensuite utilisés dans la deuxième étape pour l'identification des paramètres dynamiques du robot par une méthode à erreur de sortie sur l'effort moteur. Cette méthode récente (DIDIM) qui utilise les modèles d'identification directe et inverse du robot permet d'obtenir une convergence des paramètres très rapide et insensible aux conditions initiales. Une validation expérimentale sur un robot un axe montre l'efficacité de la nouvelle méthode comparée à deux méthodes à erreur de sortie classiques.

This paper deals with joint stiffness identification with a new Closed-Loop Output Error (CLOE) method which minimizes the quadratic error between the actual motor force/torque and the simulated one. This method is based on the DIDIM (Direct and Inverse Identification Model) procedure which has been validated on rigid robots and which is now applied to a flexible joint robot. DIDIM method requires a gain updating in the simulated robot in order to keep the bandwidth of the rigid controlled degree of freedom (dof ) and to keep the natural frequency of the flexible dof, close to the actual ones, at each step of the recursive Gauss Newton non linear programming algorithm. This gain updating requires a first step of estimating the natural frequency of the flexible dof before applying DIDIM method in a second step. An experimental setup exhibits identification results and shows the effectiveness of our approach.

Tyre-road estimation methods have been the objective of many research programmes throughout the world. Most of these methods aim at estimating the friction components such as tyre longitudinal slip rate κ and friction coefficient μ in the contact patch area. In order to estimate the maximum available friction coefficient μmax , these methods generally use a probabilistic relationship between the grip obtained for low tyre excitations (such as constant speed driving) and the grip obtained for high tyre excitations (such as emergency braking manoeuvre). Confirmation or invalidation of this relationship from experimental results is the purpose of this paper. Experiments have been carried out on a reference track including several test boards corresponding to a wide textural spectrum. The main advantage of these experiments lies in the use of a vehicle allowing us to accurately build point-by-point relationship between κ and μ. This relationship has been determined for different tyres and pavement textures. Finally, the curves obtained are analysed to check the validity of the relationship between the current friction coefficient used by the car during normal driving conditions and μmax.

This paper addresses the important topic of joint flexibility identification. Three dynamic models depending on measurements availability are compared. The parameters are estimated by using the ordinary least squares of an over linear system obtained from the sampling of the dynamic model along a closed loop tracking trajectory. An experimental setup exhibits the experimental identification results.

L'identification des paramètres dynamiques des robots est généralement basée sur l'utilisation du modèle dynamique inverse (MDI) qui est linéaire par rapport aux paramètres dynamiques à identifier et les moindres carrés linéaires (MCL). Cette méthode nécessite une mesure précise d'une image du couple et de la position à une fréquence assez élevée d'une part, et de l'estimation des vitesses et accélérations par un filtrage passe bande de la position d'autre part. La matrice d'observation étant bruitée et l'identification étant effectuée en boucle fermée, nous avons adapté la méthode de la variable instrumentale (IV) pour remédier au problème de la matrice d'observation bruitée. Récemment, une méthode d'identification originale ne nécessitant que la mesure des couples (grandeurs d'entrée) a été appliquée avec succès sur un robot industriel à 6 degrés de liberté (DDL). Cette méthode appelée DIDIM est une méthode erreur de sortie en boucle fermée (CLOE) basée sur une simulation en boucle fermée du robot. Les résultats expérimentaux ont montré que les résultats IV et DIDIM sont numériquement égaux. Dans ce papier, nous comparons ces deux méthodes grâce à une analyse basée sur les doubles moindres carrés (2SLS). Cette étude est expérimentalement validée sur un robot 2 DDL plan.

Off-line robot dynamic identification methods are mostly based on the use of the inverse dynamic model, which is linear with respect to the dynamic parameters. This model is sampled while the robot is tracking reference trajectories that excite the system dynamics. This allows using linear least-squares techniques to estimate the parameters. This method requires the joint force/torque and position measurements and the estimate of the joint velocity and acceleration, through the bandpass filtering of the joint position at high sampling rates. Recently, the DIDIM method has been proposed and validated on a 2 degree-of-freedom robot. DIDIM method requires only the joint force/torque measurement. It is a closed-loop output error (CLOE) method based on a closed-loop simulation of the robot using the direct dynamic model, the same structure of the control law, and the same reference trajectory for both the actual and the simulated robot. The optimal parameters minimize the 2-norm of the error between the actual force/torque and the simulated force/torque. A validation experiment on a 6 dof Staubli TX40 robot shows that DIDIM method is very efficient on industrial robots.

Off-line robot dynamic identification methods are mostly based on the use of the inverse dynamic model, which is linear with respect to the dynamic parameters. This model is calculated with torque and position sampled data while the robot is tracking reference trajectories that excite the system dynamics. This allows using linear least-squares techniques to estimate the parameters. This method requires the joint force/torque and position measurements and the estimate of the joint velocity and acceleration, through the bandpass filtering of the joint position at high sampling rates. A new method called DIDIM (Direct and Inverse Dynamic Identification Models) has been proposed and validated on a 2 degree-of- freedom robot [1]. DIDIM method requires only the joint force/torque measurement. It is based on a closed-loop simulation of the robot using the direct dynamic model, the same structure of the control law, and the same reference trajectory for both the actual and the simulated robot. The optimal parameters minimize the 2-norm of the error between the actual force/torque and the simulated force/torque. A validation experiment on a 6 dof Staubli TX40 robot shows that DIDIM method is very efficient on industrial robots.

This paper deals with joint stiffness identification with only actual motor force/torque data instead of motor and load positions. The parameters are estimated by using the DIDIM method which needs only input data. This method was previously validated on a 6 DOF rigid robot and is now extended to flexible systems. The criterion to be minimized is the quadratic error between the measured actual motor force/torque and the simulated one. The optimal parameters are calculated with the Nelder - Mead simplex algorithm. An experimental setup exhibits the experimental identification results and shows the effectiveness of our approach.

Off-line robot dynamic identification methods are mostly based on the use of the inverse dynamic model, which is linear with respect to the dynamic parameters. This model is sampled while the robot is tracking reference trajectories that excite the system dynamics. This allows using linear least-squares techniques to estimate the parameters. This method requires the joint force/torque and position measurements and the estimate of the joint velocity and acceleration, through the bandpass filtering of the joint position at high sampling rates. A new method called DIDIM has been proposed and validated on a 2 degree-of-freedom robot. DIDIM method requires only the joint force/torque measurement. It is a closed-loop output error (CLOE) method based on a closed-loop simulation of the robot using the direct dynamic model, the same structure of the control law, and the same reference trajectory for both the actual and the simulated robot. The optimal parameters minimize the 2-norm of the error between the actual force/torque and the simulated force/torque. A validation experiment on a 6 dof Staubli TX40 robot shows that DIDIM method is very efficient on industrial robots.

Usually, the identification of the dynamic parameters of robot is based on the use of the inverse dynamic model which is linear with respect to the parameters. This model is sampled while the robot is tracking exciting trajectories. This allows using linear least squares techniques to estimate the parameters. The efficiency of this method has been proved through the experimental identification of a lot of prototypes and industrial robots. However, this method needs joint torque and position measurements and the estimation of the joint velocities and accelerations through the pass band filtering of the joint position at high sample rate. The new method overcomes these drawbacks. It is based on a closed loop simulation using the direct dynamic model. The optimal parameters minimize the 2 norms of the error between the actual torque and the simulated torque assuming the same control law and the same tracking trajectory. This non linear least squares problem dramatically is simplified using the inverse model to calculate the simulated torque.

The identification of the dynamic parameters of robot is based on the use of the inverse dynamic identification model which is linear with respect to the parameters. This model is sampled while the robot is tracking “exciting” trajectories, in order to get an over determined linear system. The linear least squares solution of this system calculates the estimated parameters. The efficiency of this method has been proved through the experimental identification of a lot of prototypes and industrial robots. However, this method needs joint torque and position measurements and the estimation of the joint velocities and accelerations through the bandpass filtering of the joint position at high sample rate. So, the observation matrix is noisy. Moreover identification process takes place when the robot is controlled by feedback. These violations of assumption imply that the LS estimator is not consistent. This paper focuses on the Refined Instrumental Variable (RIV) approach to over-come this problem of noisy observation matrix. This technique is applied to a 2 degrees of freedom (DOF) prototype devel-oped by the IRCCyN Robotic team.

L'identification des paramètres dynamiques des robots est généralement basée sur l'utilisation du modèle inverse qui est linéaire par rapport aux paramètres dynamiques à identifier. Afin d'obtenir un système surdéterminé, ce modèle est échantillonné pendant que le robot est piloté avec des mouvements excitants. La solution optimale est obtenue à l'aide des moindres carrés linéaires (MCL). L'efficacité de cette approche a été prouvée à travers de l'identification expérimentale de nombreux prototypes et robots industriels. Elle a été étendue à d'autres systèmes tels que les compacteurs, les voitures, les moteurs et les interfaces haptiques. Cependant, cette méthode nécessite une mesure précise d'une image du couple et de la position à une fréquence assez élevée. Elle nécessite également l'estimation des vitesses et accélérations par un filtrage passe bande de la position. La matrice d'observation est donc bruitée. De plus, l'identification est effectuée en boucle fermée du fait du caractère double intégrateur instable des robots. La méthode de la variable instrumentale (IV) traite du problème de la matrice d'observation bruitée et peut être statistiquement optimale. Ce papier se focalise sur cette technique et propose une extension de cette méthode. Elle est appliquée sur un robot 2 degrés de liberté (DDL) plan développé par l'équipe robotique de l'IRCCyN.

Off-line robot dynamic identification methods are mostly based on the use of the inverse dynamic model, which is linear with respect to the dynamic parameters. This model is sampled while the robot is tracking reference trajectories that excite the system dynamics. This allows using linear least-squares techniques to estimate the parameters. The efficiency of this method has been proved through the experimental identification of many prototypes and industrial robots. However, this method requires the joint force/torque and position measurements and the estimate of the joint velocity and acceleration, through the bandpass filtering of the joint position at high sampling rates. The proposed new method requires only the joint force/torque measurement. It is a closed-loop output error method where the usual joint position output is replaced by the joint force/torque. It is based on a closed-loop simulation of the robot using the direct dynamic model, the same structure of the control law, and the same reference trajectory for both the actual and the simulated robot. The optimal parameters minimize the 2-norm of the error between the actual force/torque and the simulated force/torque. This is a non-linear least-squares problem which is dramatically simplified using the inverse dynamic model to obtain an analytical expression of the simulated force/torque, linear in the parameters. A validation experiment on a 2 degree-of-freedom direct drive robot shows that the new method is efficient.

Off-line robot dynamic identification methods are mostly based on the use of the inverse dynamic model, which is linear with respect to the dynamic parameters. This model is sampled while the robot is tracking reference trajectories that excite the system dynamics. This allows using linear least-squares techniques to estimate the parameters. This method requires the joint force/torque and position measurements and the estimate of the joint velocity and acceleration, through the bandpass filtering of the joint position at high sampling rates. A new method called DIDIM has been proposed and validated on a 2 degree-of-freedom robot. DIDIM method requires only the joint force/torque measurement. It is based on a closed-loop simulation of the robot using the direct dynamic model, the same structure of the control law, and the same reference trajectory for both the actual and the simulated robot. The optimal parameters minimize the 2-norm of the error between the actual force/torque and the simulated force/torque. A validation experiment on a 6 dof Staubli TX40 robot shows that DIDIM method is very efficient on industrial robots

Ce chapitre de livre, accessible par internet, décrit une méthode pour connaître l'influence de l'erreur de mesure sur le résultat final. Elle revient à utiliser une méthode classique utilisée dans l'évaluation de la robustesse des simulations numériques vis à vis de la troncature liée au codage des réels par les ordinateurs. (la méthode CESTAC :Contrôle et Estimation Stochastique des Arrondis de Calculs).

The identification of the dynamic parameters of robot is based on the use of the inverse dynamic model which is linear with respect to the parameters. This model is sampled while the robot is tracking exciting trajectories, in order to get an over determined linear system. The linear least squares solution of this system calculates the estimated parameters. The efficiency of this method has been proved through the experimental identification of a lot of prototypes and industrial robots. However, this method needs joint torque and position measurements and the estimation of the joint velocities and accelerations through the pass band filtering of the joint position at high sample rate. So, the observation matrix is noisy. Moreover identification process takes place when the robot is controlled by feedback. These violations of assumption imply that the LS solution is biased. The Simple Refined Instrumental Variable (SRIV) approach deals with this problem of noisy observation matrix and can be statistically optimal. This paper focuses on this technique which will be applied to a 2 degrees of freedom (DOF) prototype developed by the IRCCyN Robotic team.

[...]

The Mobility Allowance Shuttle Transit (MAST) system is a type of public transportation where vehicles may deviate from a fixed route to serve clients who wish to get on or off within a service area. In this area, the system provides an on-demand service. We use Mixed-Integer Linear Programming to tackle the MAST system and consider often neglected realistic constraints, as those linked to vehicle capacity and fleet availability. We perform a sensitivity analysis to understand how the system performance varies as a result of changing client priorities.

In a context of decreasing resources and of climate change, lowering road vehicles consumption is a key point to meet CO2 reduction requirements. In addition to car technological advances, eco-driving is part of the solution but the road infrastructure should ensure its development. This work aims to demonstrate that road energy demand and associated pollutant emissions can be reduced by working out minor optimization of road infrastructure itself. For this to happen, a simple eco-driving potential criterion is built upon infrastructure parameters such as slopes and sight distances. This criterion aims to detect Misplaced Speed-sectioning Positions (MSP) with regard to the Starting Point of Deceleration (SPD); with speed-sectioning being the succession of speed changes along a given route. An enhanced energy waste formulation is then developed to quantify the vehicles energy waste due to misplaced road-signs. Thirdly, a traffic simulation constitutes a framework for energy evaluation; considering a full flow of vehicles, based on real traffic data, and by modeling several driver behaviors. Simulation results show that a significant fuel reduction of up to 5.5% can be achieved locally, simply by moving a road sign, for rural areas and without degrading road safety.

Eco-driving is an efficient way to preserve energy resources and limit environment impacts. It is first of all a question of driver behaviour and motivation, vehicle characteristics, but infrastructure parameters could favour - or impede- ecodriving. This study aims to qualify this infrastructure influence, for the speed sectioning parameter of roads. A research evaluating the adequacy between speed changes along a route, vehicle dynamics and the road gradient constitute the frame of this work. Indeed this adequacy has been proven to be descriptive of the ecodriving potentiality of a route, as detailed in a previous experimental evaluation in France (Coiret et al [8]). It is reinforced here with the help of experimental evaluation in Bosnia, with a metro-logical reinforcement : differential gps, temporal tagging system at points of interest, roof anemometer for the influence of wind, is now used to have more accurate data. In particular the better altitude measurements increase results quality, with achievable reductions of 5 % of energy consumptions for given route sections with sectioning optimisation (ie : simple speed limit panels repositioning) We have also researched crucial points on how to compare different countries and their legislation, geography, and traffic conditions. Bosnia and Herzegovina and France are in the context, to indicate that an assessment of quality level in eco-driving of the two countries is taken in consideration. Perspectives are given in terms of traffic software generalization of the experimental results and in terms of speed sectioning revisions in order to reduce linked energy losses and CO2 emissions. A route model could then be evaluated, for which speed-sectioning and positioning of amenities (speed bumper, traffic lights) could be optimized in energy use of infrastructure, with a concern not to alter the security features.

Le point de départ de notre recherche était la comparaison environnementale de différentes techniques de construction d'infrastructure de transport. Ce travail est matérialisé par le logiciel Ecorce. Il s'agissait de sélectionner et d'améliorer les techniques de construction pour réduire les impacts environnementaux. Cependant, les impacts environnementaux d'une infrastructure de transport sont, pour des critères comme les GES ou l'énergie, imputables pour une grande partie à la phase d'usage. L'idée initiale de notre recherche fut donc d'évaluer et de proposer des solutions afin de limiter les impacts de la phase d'usage dès la conception du projet par l'optimisation des matériaux ou des tracés. Aborder la phase d'usage étend considérablement les domaines à étudier. Pour arriver à des résultats concrets, nous avons restreint notre champ d'investigation à deux critères : les GES et l'énergie. Par ailleurs, nous avons travaillé en collaboration : avec la SNCF réseau sur la partie ferroviaire, avec le Laboratoire Transport et Environnement sur la partie routière. Ainsi, dans le domaine ferroviaire, nous avons proposé une méthodologie d'évaluation appelée PEAM (Project Energy Assessment Method) que nous avons appliquée pour évaluer et comparer deux tracés sur le projet de ligne à grande vitesse Montpellier-Perpignan. Dans le domaine ferroviaire, nous sommes allés au bout des objectifs initiaux du livrable, car nous avons aussi considéré la ligne classique existante, nous nous sommes donc situés au niveau d'un réseau. Dans le domaine routier, nous avons proposé une méthodologie appelée Sloop (Slope optimisation) afin d'optimiser le profil longitudinal d'un tracé routier pour limiter l'énergie ou les émission de GES pendant les phases de construction, de maintenance et d'usage. Nous l'avons appliqué sur un projet récent de modification du tracé de la RN7 afin d'évaluer son potentiel. Ainsi, sur ce cas, nous avons montré que nous pouvions baisser de 10% les émissions de GES en phase d'usage en proposant un profil longitudinal optimisé et réaliste au regard des critères de conception routière. Nous avons également contribué à démontrer l'influence de la texture sur la consommation de carburant en étudiant l'impact de la résistance au roulement du reÐtement de chaussée. Ceci ouvre des perspectives importantes concernant la gestion des réseaux d'infrastructure. En effet, certaines techniques bas coût comme les enduits ont des macrotextures importantes qui pourraient être à l'origine de surconsommation. Alors que le domaine routier était notre domaine de prédilection, nous sommes allés moins loin que dans le domaine ferroviaire en n'atteignant pas l'échelle du réseau comme le titre initial du livrable le présuppose. Cela tient à la différence de culture entre les deux domaines. Le propriétaire de lgv tient compte explicitement de l'énergie qu'il va facturer à l'entreprise ferroviaire. Le propriétaire d'un réseau routier n'est pas dans cette situation. Il a, donc, été, pour nous, plus facile de susciter des collaborations avec des entreprises du domaine ferroviaire que dans le domaine routier. Cependant, nos résultats récents nous rendent optimistes pour nouer de nouvelles collaborations dans ce secteur. Nous avons proposé une suite à ce travail pour le contrat d'objectif et de performance 2017-2021. Cette proposition a été validée par l'Ifsttar et ses tutelles. Nous utiliserons nos acquis concernant l'évaluation de la phase d'usage pour orienter nos travaux vers l'écoexploitation des infrastructures de transport. Il s'agira de proposer un outil d'évaluation des impacts environnementaux qui prendra en compte la phase d'usage ce qui permettra de synthétiser nos résultats et de les proposer via un logiciel adapté à la profession. Nous travaillerons également sur la potentialité d'éco-conduite des infrastructures routières selon l'adéquation entre tracé et vitesse d'usage à travers une thèse qui a commencé fin 2016 en cotutelle avec l'université de Sarajevo. Nous continuerons à travailler sur la résistance au roulement et sur l'influence des paramètres géométrique des tracés sur la consommation au sein du projet Danois Rose. Nous avons prévu une ouverture au mode doux en étudiant l'usage des pistes cyclables en développant un appareil permettant de mesurer la « cyclabilité » des itinéraires à Îlo, la « cyclabilité » étant la mesure de l'énergie d'usage. Lors de la prise en compte de la phase d'usage, nous avons concentré nos travaux sur deux critères : les GES et l'énergie. Sans se disperser, nous nous intéresserons progressivement à d'autres indicateurs comme les émissions de particules fines.Enfin, nous sommes, actuellement, limités dans nos évaluations par nos hypothèses simplistes sur le trafic. Nous devrons monter des collaborations afin de construire différents scénarios d'évolution permettant de fournir des évaluations plus riches tenant compte, notamment, de l'arriÎe du Îhicule autonome.

Ce rapport est le deuxième rapport prévu au sein de ce contrat, écrit par l'équipe universitaire à destination de la SNCF dans le cadre de la convention de co-encadrement SNCF/UniversitéGustaveEiffel (anciennement Ifsttar) accompagne cette thèse (No Ifsttar : RP1-E19045) sous la responsabilité scientifique de Jean Calio, expert Data, Mobilité et Territoires et Développement Durable de la SNCF et de Pierre-Olivier Vandanjon.

Rouen métropole a sollicité les capacités de l'Ifsttar dans le domaine du traitement du signal appliqué à l'usage des infrastructures à travers un courriel envoyé le 1er mars 2017 par M Toynan, responsable du service Ouvrage d'art de la métropole de Rouen, à MM Coiret et Vandanjon, chercheurs à l'Ifsttar. Cette sollicitation concerne un ouvrage en projet pour le franchissement de la Seine. L'ouvrage sera de type léger, sous la forme d'une passerelle dédiée aux modes doux (cycles, piétons). Il complètera l'offre de franchissement existante et reliera efficacement des pôles générateurs de déplacements dans l'agglomération, et notamment l'Ecoquartier Luciline en cours de réalisation en rive droite et l'Ecoquartier Flaubert en projet en rive gauche dans le cadre du projet global Seine Cité. L'ouvrage est donc envisagé dans le secteur Ouest de l'agglomération rouennaise, entre les ponts Guillaume Le Conquérant et Gustave Flaubert, soit en traverse de la partie de Seine accueillant régulièrement la grande Armada. Les appuis principaux de la structure seraient situés dans l'alignement des hangars existants. La structure provisoire dite > sera déconstruite préalablement aux travaux de passerelle. A ce jour, il n'est pas envisagé d'implanter d'appui définitif en Seine. Néanmoins, le recours à des appuis intermédiaires en phase réalisation semble incontournable. De même, la définition d'un gabarit de passe navigable influe sur la définition de la limite basse d'intrados d'ouvrage et donc sur la définition des accès. Seule une connaissance plus précise des trajectoires fluviales entre le Pont Guillaume le Conquérant en amont et le Pont Flaubert en aval permettra donc de gagner en précision sur la définition des implantations possibles d'appuis provisoires puis sur la définition du gabarit fluvial définitif. À cet effet, l'Ifsttar a été sollicité pour une étude statistique de trajectoire de navires, selon un cahier des charges qui est donné en annexe au présent document. L'Ifsttar a répondu à ce cahier des charges en rédigeant une proposition technique, également en annexe à ce document, dont le produit est un rapport d'étude sur les trajectoires fluviales sur la section considérée. Une réunion de lancement a eu lieu le 11 mai 2017 dans les locaux de Rouen métropole. Ce document est la troisième version de ce rapport d'étude. Une première version a été envoyée le 17 juin afin d'échanger avec la Métropole Rouen Normandie. Ce document a été le support d'une réunion téléphonique avec la métropole de Rouen le 27 juin afin, notamment, de préparer une réunion entre la métropole et le grand port maritime de Rouen. Nous avons abouti à une deuxième version envoyée le 3 juillet 2017 afin de fournir des données opérationnelles pour la métropole. Suite à des relectures internes, nous avons modifié cette deuxième version en améliorant sa forme afin d'aboutir à cette version qui est remise lors de la réunion de présentation des résultats du 11 juillet 2017. Ce rapport est découpé en 10 chapitres. Tout d'abord, la structure des données, que nous avons obtenues par le port maritime, est expliquée. Les grandes étapes du traitement de ces données sont présentées dans le chapitre suivant. Puis, les résultats de ce traitement sont affichées sous forme de graphiques pour chaque trimestre de l'année 2016 dans les chapitres 4 à 8. Enfin, les deux dernier chapitres discutent les résultats et conclut le rapport.

Le projet VibraSimu, financé par la direction scientifique de l'IFSTTAR, a été conduit de 2017 à 2019 par trois laboratoires de l'établissement : le Lepsis, Ease et le LPC. Le projet vise à mesurer les vibrations d'un vélo, à les modéliser et les reproduire sur simulateur immersif et enfin, à en évaluer l'impact sur le ressenti des participants. Le projet offre ainsi une meilleure connaissance des vibrations transmises au vélo et au cycliste lors de la circulation sur différents revêtements et améliore le simulateur vélo de l'institut en produisant les vibrations liées aux irrégularités de la chaussée. Il montre également la limite des dispositifs grand public (smartphones) pour mesurer ces vibrations.

L'invention concerne un procédé d'évaluation (1) de l'usure d'un pneu (2) apte à être exécuté sur un calculateur (6). Selon l'invention, le procédé (1) consiste à : déterminer, dans une première phase (10), une première fonction (F5) représentative de l'usure courante du pneu (2), comparer, dans une deuxième phase (12), au moins un point de cette première fonction (Fs) représentative de l'usure courante du pneu (2) avec au moins une borne (B) correspondant à une usure nominale limite du pneu (2) d'une fonction de référence (F0) modélisant l'usure du pneu (2), émettre un signal d'alerte (Z) propre à l'usure limite du pneu (2) lorsqu'au moins un point de la première fonction (F5) dépasse la au moins une borne (B). L'invention concerne également un dispositif convenant à la mise en oe½uvre dudit procédé (1) selon l'invention.

This article describes the generation of a detailed two-layered synthetic population of households and individuals for French municipalities. Using French census data, four synthetic reconstruction methods associated with two probabilistic integerization methods are applied. The paper offers an in-depth description of each method through a common framework. A comparison of these methods is then carried out on the basis of various criteria. Results show that the tested algorithms produce realistic synthetic populations with the most efficient synthetic reconstruction methods assessed being the Hierarchical Iterative Proportional Fitting and the relative entropy minimization algorithms. Combined with the Truncation Replication Sampling allocation method for performing integerization, these algorithms generate household-level and individual-level data whose values lie closest to those of the actual population.

Analyzing spatial structures of transportation data at various scales can be of prime interest to transportation planning and governance. In recent years, multi-scale spatial analysis methods have been developed and used in fields like ecology and geography, but only a few studies have applied these methods to transportation data. However, such methods can provide an efficient exploratory tool for: identifying those scales at which transportation data vary spatially; modeling the spatial structures at each scale; and determining the processes at work that explain these spatial structures. This paper describes and demonstrates how a multi-scale spatial analysis method, namely distance-based Moran's eigenvector maps (dbMEM), can be applied to study the spatial layout of car ownership. For this analysis, we rely on aggregated census data for small statistical areas within France's Loire-Atlantique administrative region. At first, 176 spatial vectors representing spatial patterns with a positive autocorrelation are constructed. Among the 176 vectors, only 23 significant ones are retained after performing a regression with car ownership as the dependent variable. Next, we divide these spatial vectors into three sub-models representing three spatial scales: broad scale, medium scale, and fine scale. Lastly, we identify a set of sociodemographic factors capable of explaining the spatial variation at each scale, i.e.: the broad-scale variation is mainly explained by population density, couples with children and income variables; the medium scale by couples with children, share of individuals in the 25-54 year age range and income; and the fine scale by couples with children and income variables.

The French metropolis, Rouen Normandie, has a project of a new bridge requiring temporary pier in the river Seine during the building phase. The location of this pier in the river is constrained by mechanical reasons related to the construction of the bridge. The purpose of this study is to find a position of this pier in the mechanically constrained area which minimizes vessel traffic obstruction and therefore collision risk. This type of study is classically carried out by using complex and multiple vessel dynamics simulation software in order to assess the risk for a given vessel to crash into the pier. The novelty of this study is to propose a statistical study based on the database of Automatic Identification System (AIS) of the Vessel Traffic Service (VTS) that is an automatic tracking system relying on ships transceivers. The technical objective is to identify low risk areas according to vessel speeds and sizes. Both factors should have an impact on the maneuverability of ships to avoid pier collision risk. Among all trajectories, straight line trajectories are selected based on statistical methods. The computation of prediction intervals of these trajectories delimits navigation zone. If all the straight trajectories are taken into account, the navigation zone extends to the whole river surface, which is not an helping result. However, by focusing on large and high speed vessels trajectories the navigation zone of these weak maneuverable ships is more centered in the middle of the river and represents only 25% of the river area. The complement of this area delineates possible locations of the temporary pier of the bridge in the river that do not disrupt vessel traffic.

Road safety relies mostly on available tire to road grip regarding to the forces required by a vehicle in a given dynamic situation. For example safety issues can occur when deceleration required forces exceeding the tire/road grip potential. Issues are much critical while traveling in a curve with a coupled longitudinal/transverse grip demand. Instrumented vehicles can be used in order to evaluate the available tire to road grip in rolling situations, but it remains linked to a dedicated vehicle and models have to be use to extrapolate these evaluations to various vehicles. The aim of this work is to propose a sensor embedded in the road infrastructure to evaluate the longitudinal and transverse grip demand of any traveling vehicle. A 6 axis sensor is embedded in a road test track, under a piece of the real road upper layer. Measurement are made at a acquisition rate of 2500 Hz and forces measuring ranges are of 2500 daN (vertical axis) and 750 daN (tangential). Measurements are compared to on-board measurement with a vehicle instrumented with dynamometric wheels. Three situations are experienced in a 100 meters curve: traveling at a given speed and traveling while accelerating or braking moderately, and for three vehicle speeds of 40, 60 and 80 km/h. Force evaluation from the two systems are differing of only 5% for the vertical force and 10% for the tangential forces. The differences are lower for the higher grip demands. For example moderate braking from 70 km/h in curve lead to 0.23 and 0.25 longitudinal friction coefficients (LFC) for road sensor and dynamometric wheel, and to 0.17 and 0.19 transversal friction coefficients (TFC). In perspectives, evaluations of an excessive grip demand could result in solutions as lowering the designed speed or improving the road surface layer.

This paper presents a framework to tackle the problem, which has received little attention in the literature, of adding variables to a synthetic population from aggregate data. The work herein thus enriches the existing literature by proposing a new and e icient methodology to meet this practical need. The methodology integrates three distinct stages, the first of which theoretically models the problem as a multinomial distribution. The addition of a new variable is formulated as an entropy maximization using the variables available in both the synthetic population and aggregate data. Solving this problem (in our specific case study) is not possible due to the large number of constraints involved. The second stage then presents a heuristic yielding a practical solution to the problem. This heuristic combines Bayes' theorem with the cross-entropy minimization algorithm. However, given the large number of parameters to be estimated by the proposed heuristic, some of the results obtained prove to be invalid. To rectify this shortcoming, a post-processing method is applied during a third stage to ensure the consistency of our results. The methodology is described in great detail, and examples are provided for a better understanding of these three stages. Also, this methodology is applied to a real-world case study. An income is allocated to each of the 157 000 households in the French city of Nantes based on aggregate data from the FiLoSoFi database. Income constitutes an essential microsimulation variable for taking many social and economic aspects into account (e.g. household purchasing power, redistribution policy, tax policy). Special attention is also paid to the reproducibility of our results with the databases and R-scripts used, all of which are freely available. This method remains general and is indeed applicable to other variables with available aggregate data.

The current transport system is not sustainable because of its high consumption of finite resources (mainly oil) and its impact on the environment (air quality, global warming, etc.). The design phase of the transport infrastructure has to be reconsidered in order to take into account these environmental burdens. Very recently, a methodology called Sloop (Slope Optimization) has been proposed to optimize the road longitudinal profile according to either an energy consumption or a Global Warming Potential (GWP) criterion calculated for construction and use phases. However, this optimization requires a high computation effort. It is a severe limitation to deploy this application to significant road project . The purpose of this paper is to propose a tailor-made optimization process which improves significantly this methodology. For the construction phase assessment, the methodology is based on a model of earthworks which computes the differences of geometry between the natural terrain and the road longitudinal profile. The use phase is assessed by simulating the traffic. Traffic simulations are based on vehicle dynamic models validated with real experiments. Starting from an initial longitudinal profile proposed by road designers, our algorithm finds a better profile in terms of Energy or GWP. The degrees of freedom of the optimization problem are mainly the road slopes and secondly the radii of the circles which connect them. The main constrain is the connection of the profile with the existing road. The optimization is complex due to the important number of degrees of freedom and the computation cost of the use phase assessment. It is the reason why, we developed a specific algorithm based on the Sequential Quadratic Programming Method (SQP) with suited computations of the linear approximation.

Inaccuracy of buses is a common situation. A common practice is that a certain amount of slack is usually added to the schedule of bus operation, so that the bus can execute the schedule in most cases. On the other hand, slack means that buses sometimes have to wait for a while at the station or slow down while driving. Since the bus cannot accelerate the driving process by itself, this method cannot make the bus fully implement the schedule. Researchers invented the Transit signal priority (TSP) and conditional signal priority (CSP), the purpose of which is to give the bus signal priority to speed up when it is delayed to a certain extent. Some previous work has studied the driving process of buses with CSP. However, there is still room for further improvement in the mathematical description of the bus driving process based on CSP. In this article, we analyze the driving state of the bus under different CSP states, that is, positive and negative. Then a series of representative and operational assumptions are given. These assumptions can be used as the basis for future research on such topics. With the assumptions, we give a mathematical model of the bus driving process using CSP. According to some performance indicators of the bus driving process obtained in the modeling process, an optimization goal is established to comprehensively improve the driving effect of the bus. Mathematical analysis and numerical solution verify the applicability of the model.

Images captured by cameras have become ubiquitous. Being able to reconstruct 3D scenes only using these images would be a highly desirable capability. However, this is very challenging, as images are 2D snapshots of the world, therefore generating ambiguities when lifting them from 2D back to 3D. In this thesis, we focus on two methods for 3D scene reconstruction: single-image depth estimation, and primitive decomposition.Single-image depth estimation (SIDE) refers to the ability to reconstruct the visible 3D surface of a scene given only a single image as input. This is an ill-posed problem, since many 3D scenes can explain the observed image. In order to solve this problem, modern works rely on data-driven methods and are mainly deep-learning based. These methods therefore use large training datasets of RGB-D pairs, i.e. aligned color and depth images, along with deep neural networks, in order to learn a good prior to predict depth from a single image.Primitive decompositions can be used to represent a scene as an arrangement of elementary shapes. In the 1960s, Lawrence G. Roberts proposed to represent the 3D world as an arrangement of cuboids. This representation is particularly useful for its simplicity and compactness, which downstream applications such as robotics could leverage.The first contribution of this thesis introduces a solution to a notorious problem in single-image depth estimation; most methods suffer from smooth edges around occlusion boundaries, while they are supposed to be sharp. Our method, called SharpNet, introduces geometric constraints with synthetic data during training in order to predict sharp depth maps, guided by occlusion boundaries and surface normals.Our second contribution extends our pursuit of sharper depth edges in SIDE. Neural networks are notoriously biased towards low frequencies, implying that sharp edges, which correspond to high frequency details, will often be overlooked by deep-learning based methods. In this work we introduce a new depth refinement method that sharpens predicted depth maps, and that is able to estimate crisp high frequency details. This method predicts displacement fields, which are used to sharpen depth edges by moving pixels in 2D space. When put on top of baseline single-image depth estimation methods, our method consistently improves the sharpness of depth maps without sacrificing their accuracy.Our third contribution also aims to improve existing SIDE methods with a simple extension. Most of these previous work rely on U-shaped Encoder-Decoder architectures, often referred to as UNets. While our two first contributions focus on accuracy along occlusion boundaries, this third contribution focuses on efficiency. These occlusion boundaries are indeed usually sparse in natural scenes, which creates an imbalance that results in smooth predicted depth edges. In WaveletMonoDepth, we instead take advantage of this sparsity. Because depth edges are sparse, we can compute convolutions only in places with large depth variations i.e. occlusion boundaries. Using wavelet decomposition as an intermediate representation for depth maps, we obtain large gains in efficiency while suffering only a minimal loss in accuracy.Our final contribution explores primitive decompositions as a representation for 3D scenes. RGB-D cameras can be used to scan scenes and store them as 3D point clouds. However, this process is often noisy, and 3D point clouds are expensive to store. With MonteBoxFinder, we propose to represent noisy 3D point clouds with 3D cuboids, by first detecting many 3D cuboids candidates, then finding an arrangement that best fits the scene. This search problem is highly combinatorial, and an exhaustive search is often prohibitive. MonteBoxFinder therefore draws inspiration from Monte Carlo Tree Search methods, in order to efficiently find a good set of cuboids. These cuboids can then be used as ground truth to train single-image cuboid decomposition method.

[...]

This paper is divided into two parts. The first part is devoted to the study of a class of Approximate Message Passing (AMP) algorithms which are widely used in the fields of statistical physics, machine learning, or communication theory. The AMP algorithms studied in this part are those where the measurement matrix has independent elements, up to the symmetry constraint when this matrix is symmetric, with a variance profile that can be sparse. The AMP problem is solved by adapting the approach of Bayati, Lelarge, and Montanari (2015) to this matrix model. The Lotka-Volterra (LV) model is the standard model for studying the dynamical behavior of large dimensional ecological food chains. The second part of this paper is focused on the study of the statistical distribution of the globally stable equilibrium vector of a LV system in the situation where the random symmetric interaction matrix among the living species is sparse, and in the regime of large dimensions. This equilibirium vector is the solution of a Linear Complementarity Problem, which distribution is shown to be characterized through the AMP approach developed in the first part. In the large dimensional regime, this distribution is close to a mixture of a large number of truncated Gaussians.

[...]

Cette brochure recense l’ensemble des exercices des épreuves du brevet ayant trait au thème « Algorithmique et programmation » du programme de mathématiques de cycle 4, de 2017 à 2022. Pour chaque exercice, sont identifiés les thèmes mathématiques et informatiques abordés ainsi que les compétences informatiques mises en jeu. En début de brochure figurent les grilles d’analyse par année, et en fin de brochure des index par thèmes mathématiques, thèmes informatiques et compétences informatiques. Tous les sujets contenant des programmes en Scratch ont été traduits au format Scratch 3 pour faciliter leur usage en classe. Les programmes correspondants sont disponibles sur le studio Scratch de la C3i. Vous êtes libres de les remixer pour les utiliser en classe.

Due to the impact of the global COVID-19, supply chain (SC) risk management under the ripple effect is becoming an increasingly hot topic in both practice and research. In our former research, a robust dynamic bayesian network (DBN) approach has been developed for disruption risk assessment, whereas there still exists a gap between the proposed simulated annealing (SA) algorithm and commercial solver in terms of solution quality. To improve the computational efficiency for solving the robust DBN optimisation model, a tabu search heuristic is proposed for the first time in this paper. We design a novel problem-specific neighborhood move to keep the search in feasible solution space. The computational experiments, conducted on randomly generated instances, indicate that the average gap between our approach and commercial solver is within 0.07 %, which validates the performance of the proposed method.

Given permutations π, σ1 and σ2 , the permutation π (viewed as a string) is said to be a shuffle of σ1 and σ2 , in symbols π ∈ σ1 σ2 , if π can be formed by interleaving the letters of two strings p1 and p2 that are order-isomorphic to σ1 and σ2 , respectively. Given a permutation π ∈ S2n and a bijective mapping f : Sn → Sn , the f -Unshuffle-Permutation problem is to decide whether there exists a permutation σ ∈ Sn such that π ∈ σ f (σ). We consider here this problem for the following bijective mappings: inversion, reverse, complementation, and all their possible compositions. In particular, we present combinatorial results about the permutations accepted by this problem. As main results, we obtain that this problem is NP-complete when f is the reverse, the complementation, or the composition of the reverse with the complementation.

Joint-Embedding Self Supervised Learning (JE-SSL) has seen a rapid development, with the emergence of many method variations but only few principled guidelines that would help practitioners to successfully deploy them. The main reason for that pitfall comes from JE-SSL's core principle of not employing any input reconstruction therefore lacking visual cues of unsuccessful training. Adding non informative loss values to that, it becomes difficult to deploy SSL on a new dataset for which no labels can help to judge the quality of the learned representation. In this study, we develop a simple unsupervised criterion that is indicative of the quality of the learned JE-SSL representations: their effective rank. Albeit simple and computationally friendly, this method ---coined RankMe--- allows one to assess the performance of JE-SSL representations, even on different downstream datasets, without requiring any labels. A further benefit of RankMe is that it does not have any training or hyper-parameters to tune. Through thorough empirical experiments involving hundreds of training episodes, we demonstrate how RankMe can be used for hyperparameter selection with nearly no reduction in final performance compared to the current selection method that involve a dataset's labels. We hope that RankMe will facilitate the deployment of JE-SSL towards domains that do not have the opportunity to rely on labels for representations' quality assessment.

Recent progress has been made towards learning invariant or equivariant representations with self-supervised learning. While invariant methods are evaluated on large scale datasets, equivariant ones are evaluated in smaller, more controlled, settings. We aim at bridging the gap between the two in order to learn more diverse representations that are suitable for a wide range of tasks. We start by introducing a dataset called 3DIEBench, consisting of renderings from 3D models over 55 classes and more than 2.5 million images where we have full control on the transformations applied to the objects. We further introduce a predictor architecture based on hypernetworks to learn equivariant representations with no possible collapse to invariance. We introduce SIE (Split Invariant-Equivariant) which combines the hypernetwork-based predictor with representations split in two parts, one invariant, the other equivariant, to learn richer representations. We demonstrate significant performance gains over existing methods on equivariance related tasks from both a qualitative and quantitative point of view. We further analyze our introduced predictor and show how it steers the learned latent space. We hope that both our introduced dataset and approach will enable learning richer representations without supervision in more complex scenarios.

We investigate questions related to the notion of recognizability of sequences of morphisms, a generalization of Mossé's Theorem. We consider the most general class of morphisms including ones with erasable letters. The main result states that a sequence of morphisms with bounded alphabet size is eventually recognizable for aperiodic points, improving and simplifying a result of Berthé et al. (2019). This also provides a new simple proof for the recognizability of a single morphism on its shift space. The main ingredient of the proof are elementary morphisms.

Abstract The focus of this work is sign spotting –given a video of an isolated sign, our task is to identify whether and where it has been signed in a continuous, co-articulated sign language video. To achieve this sign spotting task, we train a model using multiple types of available supervision by: (1) watching existing footage which is sparsely labelled using mouthing cues; (2) reading associated subtitles (readily available translations of the signed content) which provide additional weak-supervision ; (3) looking up words (for which no co-articulated labelled examples are available) in visual sign language dictionaries to enable novel sign spotting. These three tasks are integrated into a unified learning framework using the principles of Noise Contrastive Estimation and Multiple Instance Learning. We validate the effectiveness of our approach on low-shot sign spotting benchmarks. In addition, we contribute a machine-readable British Sign Language (BSL) dictionary dataset of isolated signs, BslDict , to facilitate study of this task. The dataset, models and code are available at our project page.

The goal of this work is to detect and recognize sequences of letters signed using fingerspelling in British Sign Language (BSL). Previous fingerspelling recognition methods have not focused on BSL, which has a very different signing alphabet (e.g., two-handed instead of one-handed) to American Sign Language (ASL). They also use manual annotations for training. In contrast to previous methods, our method only uses weak annotations from subtitles for training. We localize potential instances of fingerspelling using a simple feature similarity method, then automatically annotate these instances by querying subtitle words and searching for corresponding mouthing cues from the signer. We propose a Transformer architecture adapted to this task, with a multiple-hypothesis CTC loss function to learn from alternative annotation possibilities. We employ a multi-stage training approach, where we make use of an initial version of our trained model to extend and enhance our training data before re-training again to achieve better performance. Through extensive evaluations, we verify our method for automatic annotation and our model architecture. Moreover, we provide a human expert annotated test set of 5K video clips for evaluating BSL fingerspelling recognition methods to support sign language research.

Recently, sign language researchers have turned to sign language interpreted TV broadcasts, comprising (i) a video of continuous signing and (ii) subtitles corresponding to the audio content, as a readily available and large-scale source of training data. One key challenge in the usability of such data is the lack of sign annotations. Previous work exploiting such weakly-aligned data only found sparse correspondences between keywords in the subtitle and individual signs. In this work, we propose a simple, scalable framework to vastly increase the density of automatic annotations. Our contributions are the following: (1) we significantly improve previous annotation methods by making use of synonyms and subtitle-signing alignment; (2) we show the value of pseudo-labelling from a sign recognition model as a way of sign spotting; (3) we propose a novel approach for increasing our annotations of known and unknown classes based on in-domain exemplars; (4) on the BOBSL BSL sign language corpus, we increase the number of confident automatic annotations from 670K to 5M. We make these annotations publicly available to support the sign language research community.

Systems that can efficiently search collections of sign language videos have been highlighted as a useful application of sign language technology. However, the problem of searching videos beyond individual keywords has received limited attention in the literature. To address this gap, in this work we introduce the task of sign language retrieval with free-form textual queries: given a written query (e.g., a sentence) and a large collection of sign language videos, the objective is to find the signing video in the collection that best matches the written query. We propose to tackle this task by learning cross-modal embeddings on the recently introduced large-scale How2Sign dataset of American Sign Language (ASL). We identify that a key bottleneck in the performance of the system is the quality of the sign video embedding which suffers from a scarcity of labeled training data. We, therefore, propose SPOT-ALIGN, a framework for interleaving iterative rounds of sign spotting and feature alignment to expand the scope and scale of available training data. We validate the effectiveness of SPOT-ALIGN for learning a robust sign video embedding through improvements in both sign recognition and the proposed video retrieval task.

[...]

Overparametrization is a key factor in the absence of convexity to explain global convergence of gradient descent (GD) for neural networks. Beside the well studied lazy regime, infinite width (mean field) analysis has been developed for shallow networks, using on convex optimization technics. To bridge the gap between the lazy and mean field regimes, we study Residual Networks (ResNets) in which the residual block has linear parametrization while still being nonlinear. Such ResNets admit both infinite depth and width limits, encoding residual blocks in a Reproducing Kernel Hilbert Space (RKHS). In this limit, we prove a local Polyak-Lojasiewicz inequality. Thus, every critical point is a global minimizer and a local convergence result of GD holds, retrieving the lazy regime. In contrast with other mean-field studies, it applies to both parametric and non-parametric cases under an expressivity condition on the residuals. Our analysis leads to a practical and quantified recipe: starting from a universal RKHS, Random Fourier Features are applied to obtain a finite dimensional parameterization satisfying with high-probability our expressivity condition.

Construction based on 3D concrete printing has become an attractive technology, infusing digital technology to the sector. Towards this digitalization, we propose methods based on computer vision for the monitoring of constructions built using contour crafting (a 3D concrete printing method). Our methods can be divided into two types: damage and quality monitoring.For damage monitoring, we work on crack segmentation. Given that cracks are difficult to annotate accurately, an accurate segmentation is difficult using supervised machine learning. With a method inspired by weakly supervised classification, we improved the accuracy of the predicted cracks' width in synthetic data: the precision increases up to 26% with respect to training with raw inaccurate annotations; this method applies to real data straightforwardly. With a method based on transfer learning, we show that models trained solely on our synthetic images are able to segment cracks in real images. In both cases, the predicted crack width accuracy improves when measured with unsupervised metrics.For quality monitoring, we work on inline anomaly detection in 3D printed pieces. First, we segment the interlayer lines in images acquired during printing (F-score = 91%). We analyze these lines by measuring geometrical properties at pixel level; we define as anomalies the regions with values outside the ranges of user-admissible values. We also classify the located layers based on their texture: either as good or as one of 3 defective cases (macro-averaged F-score = 94%).

[...]

Software Defined Networks have opened the door to statistical and AI-based techniques to improve efficiency of networking. Especially to ensure a certain Quality of Service (QoS) for specific applications by routing packets with awareness on content nature (VoIP, video, files, etc.) and its needs (latency, bandwidth, etc.) to use efficiently resources of a network. Monitoring and predicting various Key Performance Indicators (KPIs) at any level may handle such problems while preserving network bandwidth. The question addressed in this work is the design of efficient, low-cost adaptive algorithms for KPI estimation, monitoring and prediction. We focus on end-to-end latency prediction, for which we illustrate our approaches and results on data obtained from a public generator provided after the recent international challenge on GNN [12]. In this paper, we improve our previously proposed low-cost estimators [6] by adding the adaptive dimension, and show that the performances are minimally modified while gaining the ability to track varying networks.

Software Defined Networks have opened the door to statistical and AI-based techniques to improve efficiency of networking. Especially to ensure a certain Quality of Service (QoS) for specific applications by routing packets with awareness on content nature (VoIP, video, files, etc.) and its needs (latency, bandwidth, etc.) to use efficiently resources of a network. Predicting various Key Performance Indicators (KPIs) at any level may handle such problems while preserving network bandwidth. The question addressed in this work is the design of efficient and low-cost algorithms for KPI prediction, implementable at the local level. We focus on end-to-end latency prediction, for which we illustrate our approaches and results on a public dataset from the recent international challenge on GNN [1]. We propose several low complexity, locally implementable approaches, achieving significantly lower wall time both for training and inference, with marginally worse prediction accuracy compared to state-of-the-art global GNN solutions.

[...]

Deep learning belongs to the broader family of machine learning methods and currently provides state-of-the-art performance in a variety of fields, including medical applications. Deep learning architectures can be categorized into different groups depending on their components. However, most of them share similar modules and mathematical formulations. In this chapter, the basic concepts of deep learning will be presented to provide a better understanding of these powerful and broadly used algorithms. The analysis is structured around the main components of deep learning architectures, focusing on convolutional neural networks and autoencoders.

Image deblurring is an essential image restoration problem arising in several fields from astronomy to medical science. It amounts to restoring an image from a degraded, blurry and noisy, version of it. Bayesian image deblurring seeks for the posterior distribution of the image (and blur kernel, when unknown) given an observation model and some prior knowledge on the unknowns. Closed form for the posterior distribution can rarely be computed analytically, so Bayesian approximation tools are deployed to derive an estimation of it. This thesis brings novel contributions in that topic, by introducing novel Bayesian methods for tackling two important scenarios of image deblurring problems. First, we are interested in the so-called blur identification problem, of estimating spatially varying blur kernels given a clean image and its degraded version. We construct a probabilistic state-space model accounting for the smoothness among the neighboring blur kernels, and propose a novel algorithm based on bootstrap PF (BPF) to sample weighted particles, and thus construct the sought posterior distribution. Numerical experiments on various spatially-variant blur maps and images illustrate the benefits and good stability of our approach. Second, we focus on the blind image deblurring problem of jointly estimating the image and blur kernel given the blurry noisy image. We adopt the variational Bayesian approach, to build an appropriate approximation of the posterior distribution. We introduce majorization steps to maintain closed form updates even for non conjugate priors and non-Gaussian noise. This yields the variational Bayesian algorithm (VBA). We then propose to unfold VBA over neural network layers, following the recently introduced deep unrolling paradigm. This yields the unfolded VBA, benefiting from reduced parameter tuning, fast computations on GPU architecture, and improved quantitative restoration results. The superiority of unfolded VBA over state-of-the-art blind deblurring techniques is illustrated on three datasets involving color/grayscale natural images and various blur shapes

3D reconstruction is a long standing problem in computer vision. Yet, state-of-the-art methods still struggle when the images used have large illumination changes, many occlusions or limited textures. Deep Learning holds promises of improving 3D reconstruction in such setups, but classical methods still produce the best results. In this thesis we analyse the specificity of deep learning applied to multiview 3D reconstruction and introduce new deep learning based methods.The first contribution of this thesis is an analysis of the possible supervision for training Deep Learning models for sparse image matching. We introduce a two-step algorithm that first computes low resolution matches using deep learning and then matches classical local features inside the matches regions. We analyze several levels of supervision and show that our new epipolar supervision leads to the best results.The second contribution is also a study of supervision for Deep Learning but applied to another scenario: calibrated 3D reconstruction in the wild. We show that existing unsupervised methods do not work on such data and we introduce a new training technique that solves this issue. We then exhaustively compare unsupervised approach and supervised approaches with different network architectures and training data.Finally, our third contribution is about data representation. Neural implicit representation were recently used for image rendering. We adapt this representation to the multiview reconstruction problem and we introduce a new method that, similar to classical 3D reconstruction techniques, optimizes photo-consistency between projections of multiple images. Our approach outperforms state-of-the-art by a large margin.

Over the past few years, numerous computational models have been developed to solve Optimal Transport (OT) in a stochastic setting, where distributions are represented by samples and where the goal is to find the closest map to the ground truth OT map, unknown in practical settings. So far, no quantitative criterion has yet been put forward to tune the parameters of these models and select maps that best approximate the ground truth. To perform this task, we propose to leverage the Brenier formulation of OT. Theoretically, we show that this formulation guarantees that, up to sharp a distortion parameter depending on the smoothness/strong convexity and a statistical deviation term, the selected map achieves the lowest quadratic error to the ground truth. This criterion, estimated via convex optimization, enables parameter tuning and model selection among entropic regularization of OT, input convex neural networks and smooth and strongly convex nearest-Brenier (SSNB) models. We also use this criterion to question the use of OT in Domain-Adaptation (DA). In a standard DA experiment, it enables us to identify the potential that is closest to the true OT map between the source and the target. Yet, we observe that this selected potential is far from being the one that performs best for the downstream transfer classification task.

In the context of constructing an embedded system to help visually impaired people to interpret text, in this paper, an efficient High-level synthesis (HLS) Hardware/Software (HW/SW) design for text extraction using the Gamma Correction Method (GCM) is proposed. Indeed, the GCM is a common method used to extract text from a complex color image and video. The purpose of this work is to study the complexity of the GCM method on Xilinx ZCU102 FPGA board and to propose a HW implementation as Intellectual Property (IP) block of the critical blocks in this method using HLS flow with taking account the quality of the text extraction. This IP is integrated and connected to the ARM Cortex-A53 as coprocessor in HW/SW codesign context. The experimental results show that the HLS HW/SW implementation of the GCM method on ZCU102 FPGA board allows a reduction in processing time by about 89% compared to the SW implementation. This result is given for the same potency and strength of SW implementation for the text extraction

Cloud computing is being embraced more and more by telecommunication operators for on-demand access to computing resources. Knowing that 5G Core reference architecture is envisioned to be cloud-native and service-oriented, we propose, in this paper, offloading to the cloud, some of 5G delay-tolerant Network Functions and in particular the Network Data Analytics Function (NWDAF). The dynamic selection of cloud resources to serve off-loaded 5G-NWDAF, while incurring minimum cost and maximizing utilization of served next generation Node-Bs (gNBs) requires agility and automation. This paper introduces a framework to automate the selection process that satisfies resource demands while meeting two objectives, namely, cost minimization and utilization maximization. We first formulate the mapping of gNBs to 5G-NWDAF problem as an Integer Linear Program (ILP). Then, we propose an algorithm to solve it based on branch-cut-and-price technique combining all of branch-and-price, branch-and-cut and branch-and-bound. Results using pricing data from a public cloud provider (Google Cloud Platform), show that our proposal achieves important savings in cloud computing costs and reduction in execution time compared to other state-of-the-art frameworks.

[...]

[...]

Virtualization is getting unprecedented attention from Mobile Network Operators (MNOs) as it provides agility in deployment, especially when coupled with the Cloud that offers inherent elasticity and load-balancing of resources. MNOs have to ensure operational excellence by meeting several objectives. In this context, we propose in this paper, a framework for optimizing the mapping of next Generation Node-Bs (gNBs) to Software-Defined 5G Core (5GC) delay tolerant Network Functions (NFs). These NFs are considered to be deployed as a Virtual Machine (VM) pool, or containers, in order to minimize cloud computing cost, processing power and at the same time maximize network load. First, we formulate this problem as an integer linear program, while taking into account multiple constraints including Virtual Central Processing Unit (vCPU) capacity, central processing load limits and integrality of mapping relations between gNBs and 5GC NFs. Then, we propose an algorithm to solve large problem instances based on Branch, Cut and Price (BCP) combining all of “Branch and Price”, “Branch and Cut” and “Branch and Bound” frameworks. We present several schemes reflecting different optimization goals that the MNO can foster: virtualization cost, power minimization, network load or all. Simulation results demonstrate the good performance of our proposed algorithm to solve the gNBs-VM pool mapping for all evaluated schemes, while also emphasizing the advantages of a particular one (EWoS-333 for Equal Weight optimization Scheme) that can decrease virtualization cost by almost one order of magnitude compared to a static selection scheme, while considering the other two objectives.

With the emergence of virtualization and software automation for mobile networks, network slicing is enabling operators to dynamically provision network resources tuned to suit heterogeneous service requirements. This article investigates the architectures of the fifth generation (5G) of mobile networks experimental prototypes with a focus on network slicing. We present some existing 5G prototypes and identify their gaps. We then propose an architecture and a design of a 5G micro-service-based prototype. This prototype has the ability to auto-con-figure radio resources for network slices using machine-learning-powered decisions based on real-time acquired performance metrics. Finally, we discuss some use cases on top of this prototype and their related results before concluding.

Slicing is an emergent technology for 5G. It decomposes a single Radio Access Network (RAN) into multiple virtual networks "slices" to meet demands in term of throughput, mobility, latency, reliability, etc. Slicing needs real-time reconfigurations to keep current with demands' dynamics. This results in an increased cost of Operation Expenditures (OPEX). We approached this challenge as an optimization problem of infrastructure's resources. We virtualized and pooled Baseband Units (BBUs) resources on cloud. Dynamic allocation and interconnection with Remote Radio Heads (RRHs) are made possible by leveraging the advents of Network Function Virtualization (NFV) and Software-defined Networking (SDN). We implemented Distributed Base Station (DBS) using open software platform along to a public service orchestration tool for clouds. Our contribution is integrating service selection and deployment with real-time monitoring that allowed auto-control of resources by looping resources' lifecycle. In our experiments, we deployed several slices and we tested two scenarios. First scenario addressed slices' auto-scaling (Infrastructure Scale- Out/Scale-In) when free resources are available in the pool. Second scenario simulated slices' breathing (orchestration of resources) when the pool of resources is exhausted. In first scenario, results show that leveraging cloud elasticity by auto-scaling resources saves costs by providing exactly "what-is-needed" "when-it-is- needed" in term of cloud computing. In second scenario, results show that slices' breathing maximizes the usability by employing our "inhale-and-exhale" heuristic. It is about reusing resources from under-loaded slices in favor of overloaded ones with seamless effect on the user-experience.

[...]

[...]

[...]

[...]

[...]

[...]

[...]

[...]

In this thesis, we focus we aim to represent the surfaces of 3D shapes using deformable elements. Many works from recent years have inspired this thesis. The most important is AtlasNet, a neural network architecture that learns how to deform a set of square patches to reconstruct the surface of any given shape in a “papier maché” like approach. We chose to start this thesis from AtlasNet, building upon its design to extend the philosophy of this approach to new tasks such as element discovery and joint shape collection parameterization. Our work was motivated by a range of industrial 3D-related challenges such as generating 3D representation, texturing, and many more. This thesis presents two main technical contributions. The first one is to represent a collection of objects by a set of joint deformable elements that we call elementary structures. We designed these elementary structures to learn and represent consistent and meaningful parts of the objects in the collection. They are similar to primitives. We propose two models of elementary structures : (i) patch deformation and (ii) point leaning. The first learns parametric surfaces that enable us to generate a continuous surface representation of the collection of shapes. The second learns a parametric set of points not constrained by topology, allowing us to generate more complex and interpretable elementary structures. An adjustment module learns to position and deform the elementary structures to reconstruct any shape of the collection. Using the elementary structures and the adjustment modules, we can generate an accurate reconstruction that preserves correspondences between shapes. The shape of those elements can evolve during training from a simple initialization to more complex shapes that describe the semantic part of the collection. This reconstruction strategy enables us to understand the underlying parts that best describe the collection’s objects. We see it as an analysis-by-synthesis strategy. The second technical contribution of this thesis builds upon elementary structures to learn a parameterization of a single shape or a collection’s shapes. We combine the best features of the point learning and patch deformation modules in order to build a method that generates continuous surfaces not constrained by topology. This representation is conceptually similar to an atlas, i.e., a set of compatible homeomorphisms between a 3D shape and a 2D domain. We created a pipeline whose components mimic the defining elements of atlases. Namely, we learn a 2D domain homeomorphic to the surface, a parameterization which maps it to the surface and a chart which maps on it the surface. We show that we can merge the continuous surface aspect of the patch deformation modules and the flexibility of the point learning modules by representing the 2D domain as a probability density function. Our pipeline can represent several shapes using a single joint 2D domain while producing a joint atlas representation of a collection’s shapes. Our method could allow using existing parameterization techniques such as texture mapping, displacement, roughness, ambient occlusion, height maps and many more on a collection of shapes. Both methods presented in this thesis improve the reconstruction quality as measured on benchmarks with shapes of our collection and improved the coherence of the reconstruction.

Line and plane recognition is one of the essential tasks in the field of computer vision. While methods for digital plane recognition based on digital geometry can detect the global optimal solution, they require enormous computation time. In this paper, we propose a digital line/plane recognition method using multiresolutional representation as mixed integer linear programming problems which require much less computation time than the conventional methods. Our method reduces both the outliers of points and the search space of the line/plane parameters at low resolutions. Experimental results show that the computation time was reduced significantly by our method.

A digital annulus is defined as a set of grid points lying between two circles sharing an identical center and separated by a given width. This paper deals with the problem of fitting a digital annulus to a given set of points in a 2D bounded grid. More precisely, we tackle the problem of finding a digital annulus that contains the largest number of inliers. As the current best algorithm for exact optimal fitting has a computational complexity in O(N 3 logN) where N is the number of grid points, we present an approximation method featuring linear time complexity and bounded error in annulus width, by extending the approximation method previously proposed for digital hyperplane fitting. Experiments show some results and runtime in practice.

The 5G networks are expected to meet the ever increasing number of demands, connected devices and data traffic volume, which pushes network operators to their limits in terms of flexibility, elasticity and profitability. SDN, NFV, Network Slicing, ML and Cloud computing are considered as the key technologies to tackle these problems. However, the design and the maturity of these technologies raise new problems in terms of traffic engineering and routing optimization, network infrastructure sharing and distributed SDN control management. In this context, we address, in this thesis, the following problems: 1) how to reactively and proactively define the flow rules and where to place them in order to avoid network congestion, 2) how to create an isolated and efficient E2E network slice by taking into account NFV and SDN to optimize the traffic routing and avoid congestion in a specific network slice, 3) how many SDN controllers needed, where to place them and their corresponding data plane domains in order to improve the performances, and 4) how to leverage intelligence via ML in SDN in order to solve complex problems arising in routing optimization, Network Slicing and controller placement, while guaranteeing the requested QoS.Our first contribution presents an SDN-based rules placement approach that aims to dynamically: i) predict traffic congestion by using mainly NN, then ii) learn optimal paths and reroute traffic to improve network link utilization by deploying a DQN agent. As a second contribution of this thesis, we design and implement an SDN based architecture for E2E network slicing, which proactively and dynamically adapts radio slices to the transport network slices. In the third and last contribution, we present a new method that dynamically computes the optimal number of controllers, determines their optimal locations, and at the same time partitions the set of data plane switches into clusters and assigns them to these controllers

We present a method for fitting a digital line/plane from a given set of 2D/3D grid points. In the framework of discrete geometry, a digital line or plane is defined as a set of grid points located between two parallel lines or planes separated by a small distance. Our purpose is to identify such a pair of Euclidean lines/planes that represents a given set of points. This problem is formulated as a mixed integer/linear programming problem, with the objective to maximize the number of points between the two lines/planes.

[...]

Thinning is an image operation whose goal is to reduce objet points in a "topology-preserving" way. In the process, therefore, the denition of reductive points plays an important role. Most of conventional definitions are based on the concept of point connectivities and they lack topological tools, especially for three-dimensional digital images. In this paper, we introduce a new notion of thinning based on polyhedral complexes which give a topology in the discrete space and the recursive definition of reductive points. We also clarify their topological characteristics which provide an effective thinning algorithm.

In this paper, we aim to develop an algorithm for the navigation of an autonomousrobot using optical flow observed through a vision system mounted on a mobile robot. Optical flow is used for generating a temporal-model of the robot motion. The temporalmodel enables us to detect a planar area on which robot moves by matching with optical flow, and estimate a ego-motion of the mobile robot. If we obtain the planar area and a ego-motion, we can navigate the mobile robot autonomously. We show experiments for the detection of the planar area and estimation of the ego-motion of the robot using the image sequence observed through the mobile robot.

Any watershed, when defined on a stack on a normal pseudomanifold of dimension d, is a pure (d − 1)-subcomplex that satisfies a drop-of-water principle. In this paper, we introduce Morse stacks, a class of functions that are equivalent to discrete Morse functions. We show that the watershed of a Morse stack on a normal pseudomanifold is uniquely defined, and can be obtained with a linear-time algorithm relying on a sequence of collapses. Last, but not the least, we prove that such a watershed is the cut of the unique minimum spanning forest, rooted in the minima of the Morse stack, of the facet graph of the pseudomanifold.