Contact rich motion synthesis for 3D characters
The focus will be on developing controllers for physics-based characters involved in contact rich physical interactions, such as climbing, wrestling, getting up, and manipulating objects. The research will include data captured from real world sensors (motion capture) to learn controllers that produce natural human motions. The learned controllers will then be used to generate motion for 3D characters in a real-time environment. Another objective will be to determine novel feature mappings and exploration strategies that accelerate the learning process.
Knowledge of video games, physics simulation, and/or 3D graphics is beneficial. The project will build on concepts in machine learning, specifically reinforcement learning. Experience with motion capture technology is also helpful (cameras, IMUs, optical markers).
Candidates should have strong programming skills, particularly in C++ or Python.
Desired program of studies
Masters with thesis, Doctorate
Sensors, Networks and Connectivity
Accepted candidates will be funded through a stipend provided by me. Funding is competitive with other Quebec universities.
Keywords: Character animation, reinforcement learning, contact, motion synthesis
Starting : Autumn 2022