Pause
Read
CEA vacancy search engine

Adaptive Latent World Models efficient Inference on autonomous robots


Thesis topic details

General information

Organisation

The French Alternative Energies and Atomic Energy Commission (CEA) is a key player in research, development and innovation in four main areas :
• defence and security,
• nuclear energy (fission and fusion),
• technological research for industry,
• fundamental research in the physical sciences and life sciences.

Drawing on its widely acknowledged expertise, and thanks to its 16000 technicians, engineers, researchers and staff, the CEA actively participates in collaborative projects with a large number of academic and industrial partners.

The CEA is established in ten centers spread throughout France
  

Reference

SL-DRT-27-0440  

Direction

DRT

Thesis topic details

Category

Technological challenges

Thesis topics

Adaptive Latent World Models efficient Inference on autonomous robots

Contract

Thèse

Job description

Autonomous robotics requires AI models capable of capturing, understanding, and representing the physical and spatial properties of the environment to effectively predict how the environment will evolve under the influence of actions (causality). These models, known as World Models, can thus simulate a large number of possible action sequences and future states to make optimal decisions. In this context, latent space World Models of the JEPA (Joint-Embedding Predictive Architecture) type have recently stood out by learning to model a representation of the world in a compact latent space and predicting its evolution directly within that same latent space. These models have significantly matured this year, with notable advancements in the learning phase and online adaptation, but their deployability on resource-constrained targets remains largely unexplored in the state of the art. The main objective of the thesis is therefore to optimize the inference of adaptive JEPA models using model compression techniques (quantization, pruning) to meet the constraints of embedded systems in autonomous robotics (real-time constraints, low latency, memory and computational compactness, low power consumption).

University / doctoral school

Sciences et Technologies de l’Information et de la Communication (STIC)

Paris-Saclay

Thesis topic location

Site

Saclay

Requester

Position start date

01/09/2027

Person to be contacted by the applicant

BEN CHEHIDA Karim karim.benchehida@cea.fr

CEA

DRT/DSCIN/DSCIN/LIAE

CEA, Campus Paris-Saclay, Nano-INNOV, Bât. 862-PC172, F-91191, Gif-sur-Yvette Cedex

0169088645

Tutor / Responsible thesis director

 

 

 

 

 

En savoir plus