[robotics-worldwide] [jobs] Internship in Machine learning for optimization of human-robot movements in INRIA, France
The LARSEN team in INRIA Nancy (France) is offering a research internship position for master students.
Internship at master level
Title: Machine learning for optimization of human-robot movements with energetic/ergonomics criteria
Duration: 6 months, starting ASAP
The goal of this internship is to combine state-of-the-art machine learning techniques to whole-body controllers of simulated humanoid robots and human-like characters, to optimize the movement with respect to ergonomics and energetic criteria.
The internship is realized in the context of the European Project AnDy , focused on human-robot collaboration. One of our objectives is to provide the robot with the capabilities to optimize its actions so as to take into account the ergonomics comfort.
For doing so, a human operator will be equipped with inertial and force sensors (Xsens MVN Link suit and force plates) to record some demonstrations of pick & place whole-body motions. The motions will be replayed in a simulation and control framework, using a whole-body QP controller, where the human is approximated by a rigid body model using a URDF description, as we did in . The whole-body controller will be parameterized, as well as the trajectories for the pick & place task, computed from the human demonstrations.
The objective is to apply state-of-the-art machine learning algorithms to optimize some parameters of the controller and/or of the motion trajectories, with respect to energetic and ergonomics criteria. A possible solution will be to take on a reinforcement learning approach, where a measure of performance of the whole movement is computed and used to optimize the parameters, as we did in  for humanoid motions. Particular attention will be devoted to try data-efficient algorithms that enable a fast optimization in few trials, developed in the context of the Resibots project . The student will develop software modules interfacing our machine library Limbo  with our control framework based on OpenSOT .
The next step will be to test his/her optimization module online with the iCub robot, helping a real human in the pick & place task by optimizing the location of the object to be picked from the robot (handover).
Active collaboration with the researchers in the AnDy and the Resibots projects will be required. Demonstrations with the iCub robot handling an object to a human wearing the XSens suit are expected at the end of the internship.
The intern will be supervised by Serena Ivaldi in close interaction with Jean-Baptiste Mouret and Pauline Maurice.
The ideal candidate is a master student in Computer Science / Machine Learning / Robotics, with good C++ programming skills and passion for machine learning and real robots applications.
INRIA is a public institute for research in computer science in France. The team is international, French is not necessary. Our lab is near the center of Nancy, a nice University town at about 1h30 by train from Paris.
The internship salary is fixed by law at about 500-550 euro/month. Interns benefit of partially reimbursed transport fees, health insurance and lunch at 2 euro at our restaurant.