Marie Charbonneau

Control Methods for Dynamic Walking and Safe Interaction

Principal Supervisor
Prof. Dr. Giorgio Metta 
Fondazione Istituto Italiano di Tecnologia

Collaboration partners:

  • Telerobot Labs
  • University of Hertfordshire

Competence Area: Embodiment


Biography

Marie Charbonneau received a bachelor's degree in Mechanical Engineering from the University of Sherbrooke (Canada) in 2010, leading her to begin her career as a mechanical design and simulation consultant. In 2011, she was introduced to robotics through mentoring for the FIRST Robotics Challenge. This experience enouraged her to undertake master studies in Robotics Engineering through the Erasmus Mundus EMARO program, carried out as a joint degree between the University of Genoa, Warsaw University of Technology and ETH Z├╝rich. She has been working as an early stage researcher at the Istituto Italiano di Tecnologia in Genoa since 2015.


Objective

This project focuses on the development of a controller for a humanoid robot which supports whole body movement control, while allowing for safe operation within a human environment. The objective is twofold: to achieve stable reactive dynamic walking and to ensure compliance to physical interactions between humans and robot. Movement planning and control, as well as the management of kinematic and dynamic constraints, are essential points of the project, which shall lead to robots which can be used in human spaces.

As a result from this project, a framework will be developed for movement planning and whole-body control, targeted to humanoid robots. The robot shall be able to stand, walk and keep balance while unknown forces may be applied to it anytime, for instance through physical interaction with people.The developed framework shall include a novel technique for avoiding joint limits, a balancing controller, and a movement planner for dynamic walking. The resulting controller will allow the management of multiple kinematic and dynamic constraints, in a hierarchy of tasks respecting the whole body dynamics of the robot and overall goals of the controller (e.g. reaching certain configurations). At the lowest level, a torque controller will perform the determined policies with millisecond precision. The iCub platform will be used for experiments and validation of the approach, making use of the pressure sensors on its artificial skin, the joint position and force-torque sensors, as well as low-level joint torque controllers implemented on the iCub.