Canada Spends Logo

Explore our Budget 2026 analysis

Grant: Alliance Grants

University of Toronto

$56,000.00

Identifying and Tackling Task Bottlenecks for Sample-Efficient Learning

FY 2024-25

Summary
The Alliance project will advance the state of the art in human-centric robotic manipulation, particularly for lower-volume manufacturing applications where current systems have significant limitations. Existing collaborative robots struggle with complex manipulation tasks; learning-based methods for manipulation require vast training datasets as well as substantial design and implementation expertise. To address these issues, the project will focus on identifying and mitigating task bottlenecks-critical regions of the task space that are highly sensitive to errors. We will develop techniques to concentrate learning efforts in these challenging areas, improving the sample efficiency of both imitation learning and reinforcement learning. The key research topics include developing metrics for contact-rich tasks to guide learning, identifying task bottlenecks through self-supervised learning, and designing strategies for optimal data collection through "next best trajectory" learning. This targeted approach will enable robotic manipulation to scale to more complex tasks, reducing failures and enhancing adaptability. The project's industrial partner, Magna International Inc., will provide essential resources and expertise, facilitating real-world testing and demonstrating the scalability of the proposed methods. The outcomes will support the broader adoption of advanced robotics in manufacturing in Canada.
Program
Alliance Grants
Location
Ontario, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data

As published

Application id
602121-2024
Name nom
Kelly, Jonathan JS
Department département
Aerospace Studies, Institute for
Organization id
31
Institution établissement
University of Toronto
Province en
Ontario
Province fr
Ontario
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2024
Competition year année de concours
2024
Award amount
56000
Program id
ALLRP
Program name en
Alliance Grants
Program name fr
Subventions Alliance
Group en
RESEARCH PARTNERSHIPS
Group fr
PARTENARIATS DE RECHERCHE
Committee code
96
Committee name en
RPP Internal Decision Cttee
Committee name fr
Cté de décision interne - PPR
Area of application code
500
Area of application group en
Manufacturing processes and products
Area of application group fr
Procédés et produits industriels
Area of application en
Manufacturing processes and products
Area of application fr
Procédés et produits industriels
Research subject code
2600
Research subject group en
Robotics
Research subject group fr
Robotique
Research subject en
Robotics
Research subject fr
Robotique
Field of research list names en
Not available
Field of research list names fr
Non disponible
Socioeconomic objective list names en
Not available
Socioeconomic objective list names fr
Non disponible
Application title
Identifying and Tackling Task Bottlenecks for Sample-Efficient Learning
Keywords
grasping; imitation learning; metric learning; motion planning; reinforcement learning; riemannian geometry; robotic manipulation
Application summary
The Alliance project will advance the state of the art in human-centric robotic manipulation, particularly for lower-volume manufacturing applications where current systems have significant limitations. Existing collaborative robots struggle with complex manipulation tasks; learning-based methods for manipulation require vast training datasets as well as substantial design and implementation expertise. To address these issues, the project will focus on identifying and mitigating task bottlenecks-critical regions of the task space that are highly sensitive to errors. We will develop techniques to concentrate learning efforts in these challenging areas, improving the sample efficiency of both imitation learning and reinforcement learning. The key research topics include developing metrics for contact-rich tasks to guide learning, identifying task bottlenecks through self-supervised learning, and designing strategies for optimal data collection through "next best trajectory" learning. This targeted approach will enable robotic manipulation to scale to more complex tasks, reducing failures and enhancing adaptability. The project's industrial partner, Magna International Inc., will provide essential resources and expertise, facilitating real-world testing and demonstrating the scalability of the proposed methods. The outcomes will support the broader adoption of advanced robotics in manufacturing in Canada.