1. Control laws for autonomous cars in uncertain road conditions
We are working on smart vehicles controllers that adapt their behaviors according to road conditions (snow, ice, etc.). 1/10th scale test cars will be used for experimental validations:
2. Control architectures for including learning in feedback laws
Detailed description here
Field of application
Many robotic systems could benefit from low-level controller that learn and improve based on past experience.
Advancement of knowledge
We are exploring new approches to include learning in low-level motion controllers of robotic systems.
Sub-project no1 : Parallel control architectures based on “reflexes”
On approach that we are exploring is concentrating the adaptation in “reflexes” that override a baseline controller in specific situations.
Sub-project no2 : Physics-based featured
Another approach we are exploring is using features based on dimensionless number, in order for the learned policy to be generalizable to many situations and platforms.
3. Lightweight portable robots for assembly tasks
We are working on developing lightweight robotic arms that a user can wear and that can act as a co-worker in many situations. This project is conducted in collaboration with Exonetik, a Sherbrooke-based start-up.
The main technological development focus on 1) lightweight actuators and transmissions and 2) controllers that are robust to the undesirable motions of the human user.
4. Motion-assistance for floor-based patient transfer systems
In collaboration with Arjo (Magog), we are developing an active system to reduce the necessary efforts to move patients during transfers.
5. Line inspection robots
In collaboration with Hydro-Québec, we are working on the controls and motion planing of their linedrone system:
Challenges include, high-bandwidth localization of the line position with respect to the drone, robust motion controller that can resist wind gusts, planing trajectories in uncertain environnements.