Innovative Projects Realized

Graph-Driven Strategic Intelligence: Innovations in Forecasting, Delisting, and Marketing Optimization – Part 1

THIS IS A GENERIC TEXT PUT IN PLACE AS THERE WAS NO PROJECT OVERVIEW

Federation of heterogeneous data sources for the linked data back-end in the Gold Fish mobile application

The overall goal is to create the backend of a mobile personal organizer that suggests professional events (conferences, colloquia, workshops, exhibitions) and contacts to establish while attending events, to members. Currently, the target audience are professionals in the biomedical domain. Given the need to feed data and meta-data from heterogeneous sources into the application, the industrial partner chose to implement the backend using semantic data management technologies. The main challenge at this stage is therefore to create the database (triplestore) that federates the chosen sources (social network profiles, event descriptions, domain terminologies, standards, etc.). Tools designed and co-designed by members of the academic team will be used and improved during this stage, in particular, methods for matching schemas and ontologies (domain models) originating in independent sources as well as for recognizing the alternative representations of the same entity (e.g. event) in independently created datasets. Thus, the academic team brings its joint expertise in the semantic technologies and its tools to the project that thoroughly benefits the industrial partner.

AI-Driven AR Avatars for Situated Learning

In this project, we explore using augmented reality avatars for supporting learning in low-traffic makerspaces. While makerspaces primarily provide community members with access to tools, they also serve as places to learn where community members build expertise with the equipment and fabrication processes. This learning process can either be guided formally using the makerspace as a working classroom or informally by learning through watching and socializing with peers. These learning opportunities are present in high-traffic makerspaces or at high-use times; however many makerspaces do not have consistently high traffic, or relevant expertise may be unavailable at appropriate times. Artificial intelligence offers opportunities to provide individualized learning support, however, lacks the physicality for learning spaces like makerspaces. Thus, we want to investigate ways that AI-driven AR avatars can support learning experiences in low-traffic makerspaces when human instructors or peer learning is unavailable. Our work can help democratize access to learning experiences in physical spaces such as makerspaces by increasing access to learning resources. This democratization can help to enable lifelong learning in key areas such as trades or fabrication technologies and help to create a flexible workforce that is resilient to a changing future.

Vanadium-Induced Protein Aggregation

Human health is impacted every day by metal contact, ranging from direct contact with metal surfaces to exposure to corrosion by-products, nanoparticles, or ions. The interaction of human proteins with these materials remains poorly understood, but metal-protein contact has been implicated as the cause of many adverse physiological reactions, such as allergies, contact dermatitis, and cancer. It has been suggested that metal exposure can result in structural changes and the aggregation of proteins, which in turn can influence corrosion reactions in protein-rich environments. Protein aggregation is harmful because it is toxic to cells, and thought to be the cause of many neurodegenerative diseases. Vanadium is an alloying element used in the most common titanium biomaterial, and exposure to this metal is suggested to be especially harmful. In this project I will travel to Colorado State University to work in Dr. Debbie Crans’ research group where they study vanadium chemistry and toxicity, and I will investigate the interactions between vanadium from corroding biomedical implants and human proteins. This will provide a new system for the Crans group to study, and will benefit my home group, Dr. Yolanda Hedberg’s group at Western University, where we study biomedical implant corrosion.

The Impacts of Driver Monitoring Systems on Mitigating Drowsiness Due to Conditional Automation

Driving automation is becoming increasingly available with the advancement in sensors and computational power. The next generation, i.e., conditional automation, allows drivers to engage in other activities like sleeping. If the system cannot operate in certain conditions due to limitations, the driver is required to takeover vehicle control. However, sleepy drivers might not be fit for taking over. Driver monitoring systems (DMS) can use driver physiological and behavioural data (e.g., heart rate, eye-tracking), vehicle kinematics (e.g., lane position), subjective measures, or their combination to detect unsafe driver states. DMS can be used to inform the vehicle and initiate interventions (e.g., warning systems, adaptive interfaces) to alert the driver. There is a vital need to understand how conditional automation can lead to drowsiness and whether and how the vehicle can intervene using a DMS to prepare the driver for a takeover request (TOR).

The objective of this study is to understand how drowsy drivers interact with a TOR in conditionally automated vehicles. The second goal of the study is to evaluate DMS-based drowsiness interventions to prepare the drowsy driver for an upcoming TOR. For this purpose, a driving simulator study will be conducted at the Driver-Vehicle Interaction Lab, Ulm University.

Understanding Drone Pilot Needs to Develop a VR Training System

Using drones to inspect power lines, which can involve travelling in isolated areas, makes this process faster and more secure. Yet, training drone pilots for this task is time-consuming and expensive. In this project, we propose to partner with Connect Atlantic Utility Services Corporation to develop a Virtual Reality training system for drone pilots. This system will utilize the industry partner’s training scenarios and geographical data to create a realistic experience. Moreover, we will design a system that captures the drone pilots’ training needs by talking with current drone pilots. In return, Connect will acquire a custom-made VR training system for drone pilots that will help enhance training efficacy, reduce operational costs, and expedite the deployment of skilled drone pilots for powerline inspections.

Comparative Analysis of the Electrified Haber Bosch and the Electrochemical Ammonia Synthesis Approaches for Green Ammonia Production

Currently, green ammonia is produced using a process that combines green hydrogen and nitrogen. There’s a newer method called Direct Ammonia Synthesis (DAS) that can be better because it can produce green ammonia directly, without going through the extra steps of producing green hydrogen and then combining it with nitrogen. This could save a lot of money and energy. We’re studying both methods and will conduct a comparative analysis. The partner company, FuelPositive, has a system using the current method, and we want to compare it with the new DAS approach. We’ll look at how much ammonia each method makes using the same amount of power, water, and nitrogen. Additionally, we’ll use a neural network model to compare the DAS method with FuelPositive’s approach.

Seasonal Domestic Harvest Labour Access in British Columbia

British Columbia’s agriculture sector faces substantial labour issues including difficulties in recruitment, retention, and workforce engagement of domestic seasonal workers, which significantly impacts its productivity and sustainability. This project addresses the critical need for a comprehensive understanding of the domestic seasonal labour market: who the workers are, what attract them, how we can increase their participation in the short term and long term. The study will focus on collect in-depth interview data from growers and domestic workers in apple, cherry and grape industries in the Okanagan region. The results can provide recommendations to improve the sector’s competitiveness, increase appeal and strategies for recruitment, retention and relationship management of domestic labour.

Non-wood fibrillated cellulose composites

The aim of the proposed research is to develop renewable textiles from agricultural residues. The project will use an alternative process, proposed by Earth Protex, for converting primarily wheat straw into a wood pulp like material while minimizing the losses during the process. The pulp will undergo further chemical treatment under mild conditions before being sheared into smaller fibrils, referred to as fibrillated cellulose. The fibrillated cellulose will then be reassembled into filaments which can be spun into yarns to produce renewable and sustainable textile alternatives to cotton and polyester, which both come with significant environmental challenges. Wheat straw is an ample residue in the Canadian Prairies and the development and eventual commercialization of the proposed technologies can bring economical, social and environmental benefits to local communities, as well as environmental benefits across Canada and the globe by contributing to a transition towards sustainable textiles and clothing.

Exploring factors that influence avian diversity and community composition in Detroit, MI. and potential synergies with the health and wellbeing of neighbourhood residents.

A collaboration between health geographers from Michigan State University and ornithologists from Carleton University, this proposed research project will investigate the factors influencing avian diversity and community composition in Detroit, MI. and the potential relationship between human health and access to biodiversity and nature. The results of this research project will lead to several high-quality manuscripts that will inform sustainable urban land-use policies that benefit people and wildlife. Being transdisciplinary, this research project will address complex socio-ecological issues facing cities globally, but is uniquely positioned to answer questions pertinent to post-industrial cities like Detroit. Of interest is the potential role of vacant lands in promoting bird diversity in urban spaces, and exploring the social implications of these spaces for neighbourhood residents in light of their conservation potential. A continuation of an already strong and long-standing collaboration, this proposed research will strengthen knowledge of urban bird ecology, and the potential synergies between urban conservation and the enhancement of human well-being in cities.

Modèle de prévision dynamique du comportement des renouvellements de polices d’assurance

Le département de Prévision et Analytique de TD Assurance est actuellement en charge de prédire les ventes et les revenues des différents produits d’assurance sur une base annuelle et mensuelle. Prédire les revenues, et ceci de manière précise, est crucial pour l’entreprise et son bon fonctionnement. Les modèles jusqu’alors utilisés ont étés construits il y a un peu plus de dix ans et ont été implémentés avec le logiciel Excel. Ces modèles sont complexes et requièrent trop de manipulations manuelles, engendrant non seulement des erreurs potentielles mais également une perte de temps. Le but du présent projet est de développer de nouveaux modèles, tout aussi performant, sinon plus, de les automatiser et de les dynamiser, notamment grâce au logiciel SAS. Nous nous concentrons sur les modèles de renouvellements et d’annulation de polices d’assurance afin de généraliser le phénomène à l’ensemble des produits d’assurance. Ces deux modèles sont dynamiques car directement liés l’un à l’autre. L’utilisation de séries chronologiques est ici requise pour construire les dits modèles.

Dynamic Trust Modeling in Federated Learning Through Balancing Utility and Privacy

The surge in data-intensive machine learning (ML) applications necessitates effective incentives for data owners (DOs) to contribute data and train ML models collaboratively. The decision to participate in collaboration depends on the balance between utility gains and privacy loss. This project focuses on federated learning (FL), where DOs participate in collaborative learning without sharing raw data. While FL preserves privacy to some extent, vulnerabilities exist in preserving data privacy through shared models. Existing literature proposes privacy guarantees but lacks a clear method to measure the utility gain of each participant. The project aims to study participants’ contributions to collaborative learning and determine compensation based on different privacy levels. The project then investigates the utility gain of each participant. In practice, FL often faces data heterogeneity, resulting in different privacy needs. This can prompt DOs to exit at different times, managing privacy budgets according to their gained utility. Exits can deplete the data pool, impoverishing model quality and potentially triggering a cascade effect. Conversely, rapid disengagement by some participants may inspire others to contribute more data, aiding joint training for a personalized model. The project explores these dynamic utility-privacy trade-offs, analyzing participants’ evolving trust in the FL procedure.