Automatic evaluation for industrial training simulation using reinforcement learning

The proposed research project will improve CM Labs Simulations’ construction training simulators by automatically evaluating trainees and providing useful feedback about their progress frequently throughout the training cycle. Machine learning techniques will be applied to a large dataset of training simulator recordings in order to discern good skill progression of heavy machine operation for various training scenarios and various levels of expertise. Inverse reinforcement learning and world model algorithms will be leveraged to learn reward functions, thus providing a mechanism to score students on their progress and easily identify what actions lead to improving their performance as heavy vehicle and construction equipment operators. The resulting framework will allow for automatic evaluation of trainees even when expert instructors are unavailable.

Faculty Supervisor:

Samira E Kahou;Sheldon Andrews

Student:

Partner:

CMLabs

Discipline:

Computer science

Sector:

Information and cultural industries; Professional, scientific and technical services

University:

École de technologie supérieure

Program: