Implementing advanced customer labeling techniques in e-commerce recommender systems

The intern will work on a research project aimed at enhancing the shopping experience for customers on Air Liquide Canada’s website by developing a personalized recommendation system. The intern will explore and apply advanced ML techniques to leverage new sources of information such as textual descriptions and user characteristics, which will help better understand customer behavior and preferences. By implementing these improvements, AL expects to have a more engaging shopping experience for its customers.

A Hybrid Learning and Distributed Optimization Method for Multi-zone Building Control

Commercial buildings like shopping centres or offices are heavy energy users. Thanks to a better control of their heating, ventilation, and air conditioning (HVAC) system used to regulate indoor temperature, buildings can be used to assist the electric power grid by, for example, by shifting their power consumption during off-peak times or by participating to demand response programs. In this work, we consider large-scale commercial buildings organized into multiple temperature zones serviced by an HVAC rooftop unit.

Predicting the exploitability of vulnerabilities

Everyday, hundreds of new vulnerabilities are discovered and disclosed to the users of the systems they affect. The sheer volume of vulnerabilities makes it difficult, if not impossible, for system administrators to rapidly address every vulnerability. Furthermore, research shows that only 5% of vulnerabilities are eventually exploited. This situation brings about a need to prioritize some vulnerabilities over others, with the vulnerabilities most likely to be exploited treated as priorities.

Intelligent e-cargo bike-oriented deliveries

From a perspective of reducing greenhouse gases, Distribution Bike City Inc. is currently designing a hybrid bike for parcel deliveries. Indeed, delivering parcels by hybrid bike is an excellent way to reduce greenhouse gas emissions compared to cars, whether they are combustion, hybrid, or fully electric. Deliveries by hybrid bike allow the decongestion of roads and are often an option as fast as the car in dense urban areas. Moreover, the hybrid bike is faster than the standard bike due to the additional power provided by the battery.

Intelligent Load Balancing for Satellite Networks (ICARUS)

ICARUS aims to design and simulate adaptive packet forwarding techniques for a multi-beam beam hopping high throughput satellites (HTS) which can offer Terabits per second throughput. The target online learning-based solution will support variable quality of service requirements, accurate time management and fault-tolerant operations, that enables autonomous configuration of HTS. The project will train Highly Qualified Personnel (HQP) who will bridge the gap between industry and academia.

Realistic Synthetic X-ray Radiography Image Data: A New Paradigm for the Advancement of Automated Defect Recognition (ADR) in Digital X-ray Radiography Applications.

Non-destructive Testing (NDT) of materials is conventionally carried out by humans, to assess the reliability and fitness-for-use of safety-critical components. Common methods include digital X-ray radiography (to be considered in this project), Thermography, Ultrasonography, etc. Formerly, there was hardly any automated assistance for detecting flaws in digital X-ray images. However, in recent years, computational advancements offer the possibility for computer algorithms to be trained to assist in such task.

Data Drift Detection and Monitoring

When a previously trained machine learning model is put into production, the production phase begins where said model makes predictions on the inputs provided to it. When the distribution of production data changes over time, we talk about data drift. Then the model is likely to become less efficient, or even obsolete. The project consists of building an intelligent system capable of alerting in the event of a data drift that would have a significant impact on the system.

Development of Module for Fault Detection and Diagnosis for Wind-Do Microgrid PLC

In this project Wind-Do and ETS will work together to optimize the Wind-Do technology by integrating new functionality to their developed microgrid to detect and to localize in real-time the faults and anomalies. This new development will help technicians to get a fast response on the behavior of the system and turn the system ON rapidly without scanning all installation equipment as well as without revising the programmable logical controller program, which is used for monitoring the Wind-Do microgrid’s equipment.

Harmonization of the engineering design and evolutionary computation

The research project aims at integrating artificial intelligence (AI) into a user interface (UI) in the context of engineering design. Based on this context, the AI-aided UI of the CAD software will actively return hints and advice on missing parts of the product and/or what will be the next steps of the design. Such a tool will reduce human-related errors, like unwanted redundancy or a forgotten component, problems that are generally identified later in the design process.

Advanced analytical and control methods for safe and intuitive motion learning of physical interactions with humanoid service robots

Service robots are robots made to work alongside and be of aid to humans in every day environments. These robots must be safe, reliable, and easy to interact without endangering humans nor the environment. The purpose of this partnership aims to develop and implement advanced control methods to enhance the safety and functionality of a class of human-like service robots, called humanoid robots. We first propose a method that enhances a robot’s ability to ensure that it is, in fact, able to detect interaction forces by analyzing the positioning of the robot limbs.