Doctorant PhD Position

INRIA; Villers-lès-Nancy; Sur place; EUR 40 000 - 60 000; Plein temps

Résumé du poste

Un institut de recherche à Villers-lès-Nancy recherche un candidat pour un doctorat explorant l'optimisation de la diversité de qualité pour le contrôle des robots. Le projet implique le développement de méthodes d'apprentissage pour des comportements adaptatifs en utilisant des retours humains. Le candidat travaillera avec des chercheurs sur des plateformes robotiques collaboratives. Ce poste offre une rémunération de 2300 € bruts par mois, avec des avantages tels que 7 semaines de congés annuels et télétravail possible.

Prestations

Restauration subventionnée

Transports publics remboursés partiellement

7 semaines de congés annuels

Qualifications

• Bonne expérience en robotique et en contrôle.
• Compétences en simulation et apprentissage par renforcement.
• Intérêt pour les robots de manipulation et à pattes.

Responsabilités

• Revoir l'état de l'art en optimisation de diversité de qualité.
• Développer des cadres d'apprentissage pour acquérir des politiques diverses.
• Intégrer les retours multimodaux pour la collaboration humain-robot.

Connaissances

Robotique

Contrôle

Apprentissage machine

Python

C++

ROS

Planification de mouvement

Formation

Master en robotique ou domaine connexe

Outils

Isaac

MuJoCo

Description du poste

Contexte et atouts du poste

Context and funding:

This position is funded by the PEPR AS3 project. Within this framework, the HUCEBOT team is developing multimodal strategies for online control and adaptation of dynamic legged robot platforms. This PhD project explores quality diversity optimization as an alternative to traditional reinforcement learning for robot control. While reinforcement learning typically converges to a single solution for a specific task, quality diversity methods learn repertoires of policies that can solve different tasks or approach the same task in multiple ways. The research will investigate structured approaches to policy search, combining behavioral exploration, diversity-driven optimization, and meta-learning for rapid adaptation. A particular focus will be on integrating multimodal human feedback to enable intuitive adaptation in collaborative scenarios involving physical interaction, such as assistive manipulation or wearable exoskeletons. The goal is to develop more flexible and adaptable robot behaviors suited for real-world human-robot collaboration.

About the team:

The candidate will join the Human Centered Robotics team (HUCEBOT) in the Inria Center of the University of Lorraine in Nancy, France. The team HUCEBOT develops control, learning, and interaction skills of human-centered robots, such as humanoid, mobile manipulators and exoskeletons. The team develops learning and control algorithms for teleoperated / supervised / autonomous robots, involved in complex manipulation tasks in man-made environments. It also develops AI-based control for wearable exoskeletons designed to assist humans at work, drones and quadrupeds to explore complex environments. The team has excellent robotics facilities, including several humanoid robots (Talos, iCub, G1), manipulators, drones, passive and active exoskeletons, wearable sensors, force plates etc. Its laboratory has a 3D printing facility and a mechatronic workshop for prototyping and maintenance, and a motion capture room with Qualisys and Xsens sensors. The team consists of many research scientists, postdocs, PhD and has the support of 1 software and 1 mechatronics engineer. The team is international - English and French speaking. French is not required, although free French classes are available in the institute for non-French speakers.

About the laboratory and Nancy:

The Inria Center of the University of Lorraine, is co-located with the Loria laboratory, in the Science and Technology Campus of the University of Lorraine (Nancy, France), next to the Botanical Gardens, at 20 minutes by public transportation or bike from the Nancy train station and City Center. Several student residences and facilities are at walking distance. Nancy is a University town, with a high quality of life and a vibrant student, Erasmus and expat community. Life is Nancy is very affordable compared to Paris, it is easy to find a student residence or apartment. Team members can also access to SUAPS, the University’s sports facilities.

About Nancy in France:

Nancy is the capital of the Grand Est region in France. It is well connected by train to Paris (90 min), Strasbourg (90 min), Luxembourg and Germany. There are direct trains from Nancy to the Paris airport CDG and the Luxembourg airport LUX. The region around Nancy is ideal for outdoor activities: there are many country trails, long bike trails, forests, mountains, lakes, ski in winter too.

Mission confiée

This PhD project explores quality diversity optimization as an alternative to traditional reinforcement learning for robot control. While reinforcement learning typically converges to a single solution for a specific task, quality diversity methods learn repertoires of policies that can solve different tasks or approach the same task in multiple ways. The research will investigate structured approaches to policy search, combining behavioral exploration, diversity-driven optimization, and meta-learning for rapid adaptation. A particular focus will be on integrating multimodal human feedback to enable intuitive adaptation in collaborative scenarios involving physical interaction, such as assistive manipulation or wearable exoskeletons. The goal is to develop more flexible and adaptable robot behaviors suited for real-world human-robot collaboration.

The candidate will collaborate with Guillaume Bellegarda (researcher), Enrico Mingo Hoffman (researcher), Jean-Baptiste Mouret (researcher), Serena Ivaldi (researcher), and Mohamed Chetouani (Professor, Sorbonne Université).

Principales activités

• Review state-of-the-art in quality diversity optimization and meta-learning for robot control.
• Develop learning-based frameworks for acquiring diverse policy repertoires.
• Investigate online adaptation and policy selection strategies for dynamic task variations.
• Integrate multimodal human feedback for adaptive human-robot collaboration.
• Deploy and validate controllers in hardware experiments on collaborative robotic platforms.
• Analyze results, write papers and dissertation.

Compétences

• Technical skills:
• Background in robotics, control, machine learning.
• Excellent skills and/or experience with simulation frameworks (i.e. Isaac, MuJoCo), reinforcement learning, and motion planning
• Excellent skills in Python, C++, ROS
• Interest and preferably experience in manipulation and legged robots
• Soft skills:
• Excellent communication skills at work, and ability to report progress
• Not afraid of challenging projects.
• Rigour and intellectual honesty
• Curiosity and desire to learn
• Practical mindset and ability to develop robust and reliable solutions
• Autonomy and organizational skills
• Love working in a multi-cultural environment
• Team player

Avantages

• Restauration subventionnée
• Transports publics remboursés partiellement
• Congés: 7 semaines de congés annuels + 10 jours de RTT (base temps plein) + possibilité d'autorisations d'absence exceptionnelle (ex : enfants malades, déménagement)
• Possibilité de télétravail (après 6 mois d'ancienneté) et aménagement du temps de travail
• Équipements professionnels à disposition (visioconférence, prêts de matériels informatiques, etc.)
• Prestations sociales, culturelles et sportives (Association de gestion des œuvres sociales d'Inria)
• Accès à la formation professionnelle
• Sécurité sociale

Rémunération

2300 € gross/month