[robotics-worldwide] [jobs] Fully funded PhD position in Robotics / Machine learning / evolutionary computation @ Inria - Nancy, France

Jean-Baptiste Mouret mouret at isir.upmc.fr
Mon May 11 05:55:59 PDT 2015

**Please forward this email to anyone who might be interested** 

A fully funded Ph.D. position is available in our group at Inria, in France. The selected PhD student will work in the ERC project ResiBots (funded by the European Research Commission), which funds 2 PhD student, an engineer, several post-docs, and all the hardware needed for the experiments with the robots. To submit your application: http://t.co/d4q3bJOnL8

The overall goal of the project is to develop novel learning algorithms that make it possible for autonomous robots to quickly discover compensatory behaviors when they are damaged (e.g. a broken motor) or when they encounter an unforeseen situation.
Demonstration: https://www.youtube.com/watch?v=p6oegqDrCDs
See also: http://pages.isir.upmc.fr/~mouret

Post-doctoral positions are also available (please e-mail jean-baptiste.mouret at inria.fr for details).

The position will start October 2015, but we would like to select the candidate before July, if possible sooner. We will close the call as soon as we will have found a suitable candidate.

About the position
The objective of this PhD is to develop and evaluate novel algorithms for damage recovery in robotics. The algorithms will be based on a combination of off-line evolutionary algorithms [1] and on-line learning based on Bayesian optimization [2] with Gaussian processes [3]. A central idea is that no diagnosis of the damage will be performed: the recovery process will be based on a trial-and-error process. We already obtained promising results with a hexapod robot (see https://www.youtube.com/watch?v=p6oegqDrCDs and [1]). 

The first objective is to generalize this approach to two other robotics platforms : (1) a mobile manipulator, and (2) the iCub robot (humanoid). The second objective is to make the robots able to take the environment into account during the trial-and-error learning process. For instance, a damaged, walking robot needs to avoid the obstacles while learning a compensatory behavior. As many experiments as possible will be performed on real robots, but the team will benefit from the help of an engineer who will be dedicated to the project.
[1] Cully, Antoine, Jeff Clune, and Jean-Baptiste Mouret. "Robots that can adapt like natural animals." arXiv preprint arXiv:1407.3501 (2014). http://arxiv.org/abs/1407.3501

Duration: 3 years
Monthly salary : 1,580 € net (medical insurance included) for the first two years, 1,661 € net the third year.

Skills & profile
Prospective candidates should hold a master degree in computer science, machine learning, artificial intelligence, or robotics. A good knowledge of C++ will be useful, as well as a willingness/enthousiasm to work with real robots.

How to apply
Submit your application as soon as possible on: http://t.co/d4q3bJOnL8

About Inria
Established in 1967, Inria is a public research body fully dedicated to computational sciences. The position will be at the Nancy-Grand Est center, near Nancy (East of France, 1.5 hours of Paris by fast train, 1.5 hour of Luxembourg by train). See: http://www.inria.fr/en/centre/nancy

We are creating a new research team about robotics and long-term autonomy and this position is a very good occasion to join the team: https://team.inria.fr/larsen/

Best regards,
— Jean-Baptiste Mouret

More information about the robotics-worldwide mailing list