Combination of multi-horizon models for demand forecasting

This project aims to develop a retail demand forecasting model that can both handle the long-term and short-term forecasting, and adjust its parameters as more data come in. General long-term prediction models are relatively precise because the context often remains same over time, but can not quickly adapt to unforeseen events, like the global pandemics. It is then necessary to develop model with multi-horizon perspectives. With the understanding and results achieved by this project, accurate and real-time improvement solutions could be proposed and implemented.

A Serious Approach to Sickness Prevention in Motion Base Simulators

Training personnel to operate machinery in the construction workplace requires a major devotion of time, resources, and safeguards. Recent methods involving virtual reality and motion base simulators have drastically enhanced the training process, but some users of these new methods report sickness and discomfort. This research aims to remedy this issue with a theoretical and data-driven approach to (1) identifying what makes a simulator nauseogenic, and (2) reducing or avoiding these patterns of motion.

Ice coverage prediction for the St-Lawrence River

This project aims at creating a model for forecasting ice formation in the St. Lawrence Seaway between the Welland Canal and Quebec City. This will improve drastically the planning of all maritime operations during the winter transition period, before the freeze-up.

Tightly coupled visual-inertial-LI DAR SLAM for real time application

Since Amazon robotics expanded the use of drones to package deliveries to customers, drone applications have been expanded to many industries along with its ability to perform various tasks autonomously. The fundamental technology of drones’ autonomy comes from perceiving its surrounding, creating its own map based on onboard sensors and estimate its location within the map.

Impact of Bioavailable Peptides Generated from Collagen Hydrolysates on Bone Remodeling

Genacol Canada Corporation Inc. manufactures a collagen hydrolysate (CH) product, sourced from bovine (Genacol® Original Formula), that has shown promising and positive results in three clinical trials for decreasing joint pain, as well as having a positive effect on articular cartilage. Following digestion, and first pass metabolism, CHs release peptides that have been shown to be the health promoting component of the supplement. New research trends have suggested that joint pain is a multi-tissue disease, and not only attributed to articular degradation.

Real-time Transcription and Knowledge Extraction of Video Consults

Dialogue is Canada's leading telemedicine provider, founded to improve well-being by using technology to deliver excellent care. Dialogue is a pioneer of virtual health care dedicated exclusively to organizations that want to improve the health and well-being of their members and families.

Advancing Autonomous Thermalling of Unmanned Aerial Gliders

As Unmanned Aerial Vehicles (UAVs) become more ubiquitous, a special class of UAVs known as Unmanned Aerial Gliders (UAGs) promises to offer more efficient flight by using atmospheric energy to remain afloat. In order to facilitate the usage of UAGs in various applications, researchers have developed algorithms which allow for autonomous flight of UAGs. The developed algorithms, however, still lag in performance as compared to piloted UAGs, and require an extensive amount of calibration upfront, making them difficult to implement on gliders of various sizes and properties.

Data Science in Pilot Performance Assessment

Automatically assessing a pilot performance during a flight training session is a capability that can enhance the flight instructor during his duty. From data gathered during a flight maneuver, we are looking for a way to automatically assess pilot performance to augment instructor performance and provide objectivity during flight training assessment.

Development of an AI-controlled closed-loop neuromodulation system form chronic conditions

The treatment of chronic conditions accounted for 58% of the annual healthcare spend in Canada in 2012, primarily through the use of pharmaceuticals. However, these are generally best suited to treat acute diseases, as with chronic use, side effects can accumulate over time while therapeutic effects diminish. Neuromodulation of the Peripheral Nervous System (PNS) represents a promising and adaptable treatment alternative to pharmaceuticals in many cases.

Efficient Deep Learning Methods that Only Require Few Labeled Data

In this postdoc, we plan to focus on computer vision tasks where existing deep learning methods require lots of labeled samples to work well. Acquiring labeled samples is time-consuming and often impractical. Thus, we investigate three different classes of methods to alleviate the label scarcity problem: active learning, weakly-supervised learning, and few-shot learning. In active learning, the goal is to label the most important samples to maximize the performance of the model while reducing labeling costs. In weakly supervised learning, the goal is to train models using weak labels.