Post-doc: humanoid locomotion/ machine learning / optimal control theory

Impressive results have been achieved in humanoid robotics thanks to recent advances in optimal control and to the increase of computational power. In particular, we can now optimize whole-body motions in real time. Nevertheless, the current methods essentially assume that the model of the robot is fully known.
The objective of this post-doc is to relax this constraint so that we can generate whole-body motions with uncertain models. A particularly challenging case of uncertain models will be mechanical damage (e.g. a broken arm) and long-term changes of the body (e.g., a backpack). To do so, we will rely on data-efficient learning methods (e.g. Bayesian optimization) so that the robot can improve its behavior in a few trials.
The main test platform is the iCub humanoid robot (an advanced, child-like robot ; see http://www.icub.org). Experiments will take advantage of our motion capture facilities and will rely our team’s expertise in both machine learning (Jean-Baptiste Mouret) and humanoid control (Serena Ivaldi & Karim Bouyarmane).
The selected post-doc will work in interaction with ERC project ResiBots (funded by the European Research Commission, http://www.resibots.eu), the EU FP7 project CODYCO (http://www.codyco.eu), and the recently accepted H2020 ANDY. The overall goal of the Resibots 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. The CODYCO project is focused on whole-body control of humanoid robots and adaptation in presence of multiple contacts; the ANDY project is tackling the challenge of human-robot collaboration in tasks that involve dynamic and kinematic constraints of both parties.
Demonstrations: [ResiBots] https://www.youtube.com/watch?v=T-c17RKh3uE , [CoDyCo] https://www.youtube.com/watch?v=9XRI4BeXN78
See also: http://members.loria.fr/JBMouret and http://www.loria.fr/~sivaldi/
ABOUT THE APPLICANT
A successful post-doc applicant should have a strong robotics and/or machine learning background with a track record of top-tier research publications, including relevant conferences (e.g., RSS, ICRA, IROS or ICML, IJCAI, AAAI) and journals (e.g., AURO, TRo, IJRR, JMLR, MLJ, Neural Computation). A Ph.D. in robotics or another field clearly related to robotics and/or machine learning is a must.
Expertise in working with real robot systems is a big plus. A robotics engineer is dedicated to the project, but good C++ programming skills will be useful. Strong knowledge of optimal control and/or machine learning are required, but we will definitely consider applicants who have expertise in only one of these two fields.
ABOUT THE POSITION
The position is for 18 months (possibly 24). The ``raw salary'' (French "brut") will start at 2621 euros (2115 euros ``net''), depending on experience and qualifications. French medical insurance is included.
Speaking French is not required.
ABOUT INRIA NANCY-GRAND-EST
Inria is a French national research institution that focuses on computer science and applied mathematics. Inria employs about 600 permanent researchers, 1000 engineers/technicians, 600 Ph.D. students, and 200 post-docs. Inria Nancy - Grand-Est centre is one of the eight research centre; it is situated near Nancy, a 1-million inhabitants (with a lot of students) in the East of France (1.5 hours from Paris by fast train).
Inria can provide help for housing and visa applications (scientific resident card and help for husband/wife).
HOW TO APPLY
Please apply online here:
http://tinyurl.com/icub-inria
[Applications that are not submitted on the submission website will be ignored].
Applicants will be asked to provide:
- a research statement / cover letter;
- a detailed CV;
- your top 3 publications;
- 3 references.
Please ensure to include a link to your research web-site as well as your date of availability. Deadline for applications: 31st of October.
Security and defense procedure : In the interests of protecting its scientific and technological assets, Inria is a restricted-access establishment. Consequently, it follows special regulations for welcoming any person who whishes to work with the institute. The final acceptance of each candidate thus depends on applying this security and defense procedure.
MORE INFORMATION
- http://www.resibots.eu
- http://members.loria.fr/JBMouret
- http://www.codyco.eu
- http://www.loria.fr/~sivaldi/