Projets novateurs réalisés

Etude la séparation de phase photo-induite d’un système hybride radicalaire/cationique

Le développement de revêtements de faible brillance polymérisés aux UV est une tendance dans le domaine des revêtements des couvre-planchers. La brillance d’un revêtement dépend de la diffraction de la lumière, et il est essentiel d’avoir un certain degré de rugosité de surface pour obtenir un aspect mat. La technologie étudiée ici pour réduire la brillance des vernis polymérisés aux UV est basée sur “l’auto-rugosité” induite par la polymérisation de systèmes hybrides radicalaire/cationique Le mélange de composés époxy et acrylate peut se séparer en phases/domaines lorsque la polymérisation est initiée par photo-induction. Cette séparation de phase des systèmes hybrides entraîne une hétérogénéité de phase microstructurale et une séparation microphasique des différents composants. La morphologie de surface formée par la séparation de phases résulte de la compétition entre la séparation de phases due à la thermodynamique et les changements physiques associés à la conversion des monomères. Ces changements physiques comprennent l’augmentation de la viscosité ainsi que la gélification et la vitrification, qui limitent toutes deux la diffusion des phases incompatibles.

Post-quantum cryptography for signing software on avionic systems

Avionics on modern aircraft are becoming complex computer systems supported by increasingly sophisticated software. These systems often require software upgrades that are performed during aircraft maintenance on the ground. In order to ensure the integrity of software upgrades and database updates, digital signature of the corresponding data files has been used as means to prevent cyber attacks that could replace the upgraded software with versions including malicious software or bad data. However, the potential advent of quantum computers present a threat to this code signing mechanism, as an attacker having access to a suitable quantum computer could fake these signatures.
Indeed, quantum computers introduce a new type of algorithm using the properties of quantum mechanics. It is believed that with enough quantum bits (qubit), a quantum computer using this algorithm would be able to crack most of the current public key signing and encryption algorithms used today.

Optimal recharging scheduling for urban electric buses

Implementation of alternative fuel technologies for public transportation modes stems from increasing environmental concerns, technological improvements, and increasing demand for autonomous transportation, which is well exhibited by recent trends to replace diesel buses with battery electric buses (e-bus). Yet, scheduling and operational planning of electric vehicle (EV)-based transit modes is challenging due to additional considerations for energy consumption models and driving modes. In this project, a holistic scheduling and optimization modelling tool is developed for battery and fuel-cell operated e-bus transit. For a given timetable and charging constraints at the depot or on-route location, this study identifies the best charging strategy that minimizes the infrastructure investment and operating cost.

Développement d’approches par apprentissage actif et en budget contraint à partir de données de cardiologie

La sténose de la valve aortique est la deuxième maladie cardio-vasculaire la plus fréquente et la maladie cardiaque valvulaire la plus fréquente. Sa prévalence au sein de la population ne cesse de croître. Non traitée, cette maladie est mortelle et seule une chirurgie permet d’empêcher sa progression. Il y a un besoin urgent de progrès majeurs dans le dépistage, le diagnostic, la stratification des risques et le traitement de cette maladie. Le principal défi dans ce contexte est l’absence de biomarqueurs et de stratégies pour identifier le moment optimal pour l’intervention dans le traitement. Dans le cadre de ce projet, des jeux de données comportant des résultats de cardiologie seront utilisés pour élaborer des modèles prédictifs. Le travail des étudiantes sera d’identifier les données les plus représentatives des sous-maladies de la sténose de la valve aortique afin de les caractériser par des cardiologues, de façon à créer des modèles robustes.

Développement d’un système de détection automatique des erreurs de parallaxe dans les modèles stéréoscopiques

Le projet proposé vise à développer un système de détection des erreurs de parallaxe présentes sur les modèles stéréoscopiques par intelligence artificielle, et plus particulièrement par l’utilisation des réseaux de neurones convolutifs. Ces derniers sont reconnus pour leur performance en vision par ordinateur, notamment dans la détection automatisée d’éléments fournis à travers des exemples d’entrainement. Le système de détection proposé sera composé d’une architecture principale pour répondre aux problématiques rencontrées par la compagnie dans le processus création des modèles stéréoscopiques : la présence d’une erreur de parallaxe. Les algorithmes développés seront intégrés dans un système opérationnel qui permettra d’automatiser une bonne partie de la phase d’analyse et ainsi de réduire le temps et les coûts de cette opération pour la compagnie.

Design and fabrication of hybridized nanophotocatalyst for waste water treatment using natural sunlight.

Sustainable method of treating waste water is one of the major issues across the world. Different techniques have been developed by various scientific communities but each possess some limitations. So, keeping in the mind of these major limitations such as high cost , external power source requirement, huge setups , non recyclable etc, design of novel solar light driven nanomaterials are proposed. Since, the use of solar energy is only ~5% in the water industry which needs to be further extend more by using solar driven nanomaterial for water treatment. Thus, our proposed research direct towards to harvest full solar spectrum (UV-Vis-NIR). Hence, the idea for developing novel hybrid nanophotocatalyst for getting the desired properties in a single composite WS2/hexa-WO3/g-C3N4 are proposed. Here, WS2 acts as light harvester, g-C3N4 provides the high conductive surface, wide space tunnelling structure that allows fast mobility of charge carriers will be provided by hexa-WO3.

Expected Outcome:
1. High efficiency in the waste water treatment.
2. Effective utilisation of solar light.
3. Promoting sustainable energy.

Towards Developing an Artificial Intelligence-based System for Detection of Cyber Attacks in Modern Industrial Control Systems

Modern Industrial Control Systems (ICS) are increasingly getting connected to the Internet to facilitate operations. To ensure safety on the internet, the ICS communications are being encrypted. This poses a challenge for the traditional Intrusion Detection Systems that used to rely on visible messages and control data communication for detecting the presence of known attacks or anomalies in system behavior. In this work, we aim to develop an AI-based system for intrusion detection in a modern ICS by modeling the encrypted network communications across an ICS. This is particularly challenging due to the following reasons: (1) a modern ICS is a multi-vendor system, consisting of devices with proprietary hardware and software; (2) collecting and analyzing the traffic for time-sensitive applications becomes difficult due to the geographically distributed nature of modern ICS networks; and, (3) the ICS components use proprietary communication protocols and encryption algorithms. Designing an IDS that scales across all these factors is a complex data modeling problem. The expected outcome of this project is an AI-based framework for analysis of encrypted ICS communications for early intrusion detection.

Caractérisation fonctionnelle et valorisation des peptides de défense des arbres de la famille des Salicacées

L’agriculture moderne recherche des solutions pour diminuer l’utilisation de pesticides chimiques. Dans ce contexte, les acteurs de la recherche et de l’innovation ont un rôle à jouer pour développer des pesticides non-chimiques : les biopesticides. Le projet s’inscrit dans ce contexte, en caractérisant une famille de peptides antimicrobiens (appelés RISPs – Rust Induced Secreted Protein) des arbres de la famille des Salicacées (peuplier et saule) et en évaluant la valorisation de ces peptides comme biopesticides. Les peptides RISPs présentent une activité antimicrobienne ciblée envers les champignons de type Pucciniales, principaux agents des maladies des rouilles des cultures, et une activité élicitrice du système immunitaire de la plante. RISP est le premier exemple de peptide de défense multifonctionnel chez les plantes. La caractérisation de la famille RISP représente à la fois une opportunité de recherche scientifique fondamentale – celle d’effectuer un apport conceptuel majeur en biologie – ainsi qu’une opportunité de démarche appliquée – celle de développer des biopesticides à partir de certains membres de la famille.

Using Artificial Intelligence to Characterize the Dynamics of Seismic Risk

The Pacific Northwest has the potential to experience large earthquakes (magnitude ~9) along the Cascadia Subduction Zone (CSZ) fault, which is located approximately 120 km off the city of Vancouver. The last CSZ earthquake occurred in 1700, and it is expected to happen every 500 years. An occurrence of this event could lead to large economic and social consequences, which raises an importance of seismic risk assessment in this region. However, seismic risk is subject to change over time due to variations in the built environment (e.g., aging infrastructure) and population shifts. The Covid-19 pandemic is an ongoing issue in British Columbia (and the world) that has changed population distribution dramatically across the region. Under such circumstances, our understanding of seismic risk and its consequences might be substantially different from what is anticipated per the existing risk models In Canada. This research aims to utilize Artificial Intelligence (AI) methods to characterize the dynamics of risk associated with changes in population distribution due to pandemics, which will provide a dynamic setup to continuously evaluate seismic risk. This framework will also serve as an adaptive decision-making tool to evaluate policy measures and their implications on community resilience.

Low cost Microchip for Point of Care high Sensitive Assays

We have developed microfluidic polymer chips, and their associated reader for rapid differential diagnosis of disease states2. Each polymer chip is 2?2 cm2, and needs only 10 ?l of whole blood sample of the patient for diagnoses. The turnaround time from introducing the blood sample to get the results is 15-30 min. The chip contains all the reagents, and thus it is disposable and can be used in doctors’ office/emergency rooms of the hospitals without any need for central laboratory equipment. The device is also fairly sensitive and we wish to extend the limit of detection for measuring disease biomarkers in blood down to femtogram/ml.

Applications of Wearable Data and AI to Augment the i-Share Platform

i-Share is a project being conducted by the Université de Bordeaux. It is one of the largest mental health studies in the world, with the goal of collecting self-report information on mental health (e.g., stress, sleep, physical exercise, depression) from over 20,000 students from French universities.
Currently, students have access to personal devices, such as smartphones and smartwatches, that continuously collect health data in real-time (e.g., physical activity, sleep, heart rate). These objective data can be integrated with i-Share’s previously existing information and allow researchers to better understand and improve mental health of students.
The goal of this project is to explore the integration of sensors embedded in smart technologies with i-Share. More specifically, we will explore the integration of data from Apple Health (Apple’s health data repository) into i-Share’s capabilities. We will also develop Artificial Intelligence algorithms in order to generate insight into these datasets.

Ensemble Application of Symbolic and Subsymbolic AI for Sentiment Analysis

Deep learning has unlocked new paths towards the emulation of the peculiarly-human capability of learning from examples. While this kind of bottom-up learning works well for tasks such as image classification or object detection, it is not as effective when it comes to natural language processing. Communication is much more than learning a sequence of letters and words: it requires a basic understanding of the world and social norms, cultural awareness, commonsense knowledge, etc.; all things that we mostly learn in a top-down manner. In this project we will integrate top-down and bottom-up learning via an ensemble of symbolic and subsymbolic AI tools, which we will apply to the interesting problem of polarity detection from text. In particular, we will integrate logical reasoning within deep learning architectures to build a commonsense knowledge base for sentiment analysis.