Research project title
Applying Machine Learning to Accelerate Digital Twin Construction
Education level
Master or doctorate
Director/co-director
Director: Bentley Oakes
End of display
February 3, 2027
Areas of expertise
Modelling and simulation studies
Virtual reality and related simulations
Primary sphere of excellence in research
Modeling and Artificial Intelligence
Secondary sphere(s) of excellence in research
Industry of the Future and Digital Society
Unit(s) and department(s)
Department of Computer Engineering and Software Engineering
Conditions
- A friendly yet professional attitude
- Ability to perform high-quality research and publish worthwhile contributions
- Strong self-motivation skills and the ability to ‘push’ projects
- Attendance to and engaging presentations at conferences
- Completion of degree requirements with high course marks
- Note that we conduct our research and discussions in English. At Polytechnique, some graduate courses may have to be taken in French
- Experience with LLMs and/or reinforcement learning is a plus.
Please see my students page for details on applying.
Detailed description
Digital Twins (DTs) are digital replicas of complex systems, such as machinery, factories, or even entire cities. Through advanced sensing, modelling, simulation, and precise control, their uses include scenario prediction, warnings of upcoming danger, and optimization for improved performance or lowered resource consumption. DTs are expected to reshape business and society by 2035 with a global market value exceeding C$64 billion within the next few years. Thus, investing in constructing DTs is important for Canada's industry, government, and citizens to understand and control their complex systems, especially in the manufacturing and infrastructure domains.
Despite the promising economic, societal, and environmental impact of DTs, current research has not yet addressed major DT construction barriers. Primarily, most users have disjointed and inconsistent domain knowledge (DK) such as system requirements, simulators, and models, with insufficient quality to construct a robust DT. Also, existing DT methodologies and tools are not suited for those without modelling experience, and they provide little guidance or recommendations. The user takes more time and expends more cognitive effort on this ad-hoc process than is necessary. This research program aims to address these barriers and significantly reduce the DT creation time and effort, such that users can assemble an initial DT for their system and begin leveraging services such as visualization and verification within a working week.
In particular, this research project focuses on the definition and application of machine learning/artificial intelligence techniques (LLMS, reinforcement learning, recommendations, etc.) to collect DK and use it to better develop DTs. The research question is: 'given the user's DK, how can this be used to better engineer DTs?'
Financing possibility
Funding available
Bentley Oakes
Assistant Professor
