Development of an end-to-end differentiable robot motion control architecture h/f

A propos d'Inria

Inria est l'institut national de recherche dédié aux sciences et technologies du numérique. Il emploie 2600 personnes. Ses 215 équipes-projets agiles, en général communes avec des partenaires académiques, impliquent plus de 3900 scientifiques pour relever les défis du numérique, souvent à l'interface d'autres disciplines. L'institut fait appel à de nombreux talents dans plus d'une quarantaine de métiers différents. 900 personnels d'appui à la recherche et à l'innovation contribuent à faire émerger et grandir des projets scientifiques ou entrepreneuriaux qui impactent le monde. Inria travaille avec de nombreuses entreprises et a accompagné la création de plus de 200 start-up. L'institut s'eorce ainsi de répondre aux enjeux de la transformation numérique de la science, de la société et de l'économie. Development of an end-to-end differentiable robot motion control architecture
Le descriptif de l'offre ci-dessous est en Anglais
Type de contrat : CDD

Contrat renouvelable : Oui

Niveau de diplôme exigé : Bac +5 ou équivalent

Fonction : Ingénieur scientifique contractuel

Niveau d'expérience souhaité : De 3 à 5 ans

Mission confiée

This assignment is part of the OSS4EAI (open source software for embodied AI) project. OSS4EAI explores the design of end-to-end differentiable learning architectures. IT revisits fundamental algorithms in robotics, particularly for physical simulation, and extends them to automate the transfer to real robots where data is sparse. This approach will BE implemented in maintained open-source software to accelerate the learning of robotic behaviors and propose new forms of data and model sharing.

Principales activités

Your goal will BE to develop a motion learning pipeline for manipulation and locomotion using the differentiable simulator developed in OSS4EAI, comparing the resulting pipeline with standard reinforcement learning. Here is a more detailed breakdown of the tasks involved :
- Learn how to use the existing on wheeled-biped robots
- Interface Simple as a forward-dynamics simulator in the Upkie software, and evaluate its performance
- Implement a rolling-without-slipping constraint in C++, in close collaboration with simulator developers
- Document the performance comparison on the project website
- Implement a policy learning pipeline using Simple as a differentiable simulator
- The algorithmic layout will BE specified with researchers actively working in the project
- Compare the performance of learned policies on the real robot

Based on the results of this first phase, the project can BE extended to a second year where we will extend the pipeline to a broader range of robot tasks (e.g. humanoid locomotion).

Compétences

Candidates should hold either a PhD degree in robotics or an MSc degree with more than five years of experience in the robotics industry.

Required skills :
- Experience in running live code on real robots
- Knowledge in robotics modeling : kinematics, dynamics, Jacobian matrices
- Programming skills in C++ and Python
- Spoken and written technical English

Other skills that will BE appreciated :
- Experience in model predictive control or reinforcement learning
- Programming skills in PyTorch

Although the project requires a strong background in robotics and software engineering, prior knowledge of reinforcement learning is not a pre-requisite as things can BE learnt on the GO with existing software and support from the team.

Avantages
- Subsidized meals
- Partial reimbursement of public transport costs
- Leave : 7 weeks of annual leave + 10 extra days off due to RTT (statutory reduction in working hours) + possibility of exceptional leave (sick children, moving home, etc.)
- Possibility of teleworking (after 6 months of employment) and flexible organization of working hours
- Professional equipment available (videoconferencing, loan of computer equipment, etc.)
- Social, cultural and sports events and activities
- Access to vocational training
- Social security coverage