Dans un contexte où les problématiques et les pratiques de recherche académique bousculent les frontières disciplinaires et où les institutions scolaires prônent l’interdisciplinarité et multiplient ...les dispositifs interdisciplinaires, ce numéro se propose de présenter l’actualité de la recherche en didactique des sciences et des technologies sur la question des rencontres entre les disciplines. Après une première partie développant les évolutions de la recherche académique et de l’institution scolaire, l’article d’introduction montre comment la question de l’interdisciplinarité, la manière dont on envisage les liens entre les disciplines sont liées aux finalités attribuées à l’enseignement des sciences, avec deux orientations contrastées mais non contradictoires potentielles, celle d’une orientation de formation épistémologique et celle d’une orientation de formation citoyenne. Les contributions de la partie thématique du dossier se positionnent différemment sur ces deux orientations quant aux enjeux attribués à l’enseignement des sciences, en lien avec les enjeux de l’interdisciplinarité. Ils articulent des réflexions épistémologiques et des analyses de dispositifs interdisciplinaires en classe et en formation des enseignants. In a context where academic research issues and practices are shaking up the boundaries between disciplines and where educational institutions promote interdisciplinarity and multiply interdisciplinary devices, this issue aims to present current research in science and technology teaching concerning the confluence of disciplines. After a first part developing the evolutions of academic research and of the scholar institution, the introductory article shows how the issue of interdisciplinarity, the way in which the links between disciplines are seen, are linked to the aims assigned to science education, with two contrasting but not conflicting potential orientations for training : an epistemological and a citizen orientation. The contributions of the thematic section of the review take different positions on these two orientations regarding the aims assigned to science teaching and in relation to the aims of interdisciplinarity. They articulate epistemological reflections and analyses of interdisciplinary devices in classroom and teacher training.
This paper presents two novel methods to generate haptic feedback for a human operating a large, hydraulic vehiclemanipulator. Such manipulators can be found in a wide range of applications (e.g. ...farming works, roadside maintenance, forestry applications). In general, they are controlled solely by a human operator, as the automation of such systems is too challenging due to their unstructured working environment. The work with such manipulators is often physically and psychologically demanding. In the presence of obstacles or other restrictions in the working area, haptic feedback can help to relieve the operator and increase his awareness. If the vehicle platform additionally moves during the operating process, state-of-the-art methods cannot be directly applied to generate haptic feedback. Therefore, we propose two shared-control concepts, based on potential field methods supporting the operator during his task. The first one is a modified trajectory guidance approach. The second one is a directly generated feedback approach, which does not compute a desired joystick position, but directly the haptic feedback. Both proposed methods do not need the reference trajectory of the manipulator, only the detected obstacles as a set of points. The reference trajectory is specified by the operator and is unknown to the automation. The two concepts are verified in a simple simulation environment. Finally, they are validated on a test bench, with a force-feedback joystick.
This paper presents the experimental validation and comparison of shared-control concepts for large vehicle-manipulators with a hydraulic robotic arm. The operation of such systems is difficult and a ...supporting automation can ease the psychological and physiological load on the human operator. The validation is performed with a real-time simulation of the vehicle-manipulator, with a graphical user interface and with the help of a force feedback joystick. The evaluation is done based on the deviation from the ground truth of the manipulator reference. Additionally, personal questionnaires are used to access the acceptance of the assistance system. The main advantage of this concept is that the automation does not need the reference of the manipulator, which facilitates a later real-life implementation easier. Therefore, this paper has three contributions: the validation of a novel shared-control concept for a vehicle-manipulator. Additionally, a practical tuning procedure is presented, in which the test persons adjust the controller parameters according to their preferences which leads to better results. Finally, the shared-control concept is extended with a supporting force feedback on the haptic user interface. These concepts are validated and compared with eachother on a test-bench in which the advantages of an active force feedback on the joystick are shown.
Self-supervised learning (SSL) is at the origin of unprecedented improvements in many different domains including computer vision and natural language processing. Speech processing drastically ...benefitted from SSL as most of the current domain-related tasks are now being approached with pre-trained models. This work introduces LeBenchmark 2.0 an open-source framework for assessing and building SSL-equipped French speech technologies. It includes documented, large-scale and heterogeneous corpora with up to 14,000 hours of heterogeneous speech, ten pre-trained SSL wav2vec 2.0 models containing from 26 million to one billion learnable parameters shared with the community, and an evaluation protocol made of six downstream tasks to complement existing benchmarks. LeBenchmark 2.0 also presents unique perspectives on pre-trained SSL models for speech with the investigation of frozen versus fine-tuned downstream models, task-agnostic versus task-specific pre-trained models as well as a discussion on the carbon footprint of large-scale model training. Overall, the newly introduced models trained on 14,000 hours of French speech outperform multilingual and previous LeBenchmark SSL models across the benchmark but also required up to four times more energy for pre-training.
Self-Supervised Learning (SSL) using huge unlabeled data has been successfully explored for image and natural language processing. Recent works also investigated SSL from speech. They were notably ...successful to improve performance on downstream tasks such as automatic speech recognition (ASR). While these works suggest it is possible to reduce dependence on labeled data for building efficient speech systems, their evaluation was mostly made on ASR and using multiple and heterogeneous experimental settings (most of them for English). This questions the objective comparison of SSL approaches and the evaluation of their impact on building speech systems. In this paper, we propose LeBenchmark: a reproducible framework for assessing SSL from speech. It not only includes ASR (high and low resource) tasks but also spoken language understanding, speech translation and emotion recognition. We also focus on speech technologies in a language different than English: French. SSL models of different sizes are trained from carefully sourced and documented datasets. Experiments show that SSL is beneficial for most but not all tasks which confirms the need for exhaustive and reliable benchmarks to evaluate its real impact. LeBenchmark is shared with the scientific community for reproducible research in SSL from speech.
PDF