Phd Position F - M Doctoral Position In Computer Animation H/F - 95

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.PhD Position F/M Doctoral position in computer animation
Le descriptif de l'offre ci-dessous est en Anglais
Type de contrat : CDD

Niveau de diplôme exigé : Bac +5 ou équivalent

Autre diplôme apprécié : BS or MS

Fonction : Doctorant

Niveau d'expérience souhaité : Jeune diplômé

A propos du centre ou de la direction fonctionnelle

The Inria Saclay-Île-de-France Research Centre was established in 2008. It has developed as part of the Saclay site in partnership with Paris-Saclay University and with the Institut Polytechnique de Paris .

The centre has , 32 of which operate jointly with Paris-Saclay University and the Institut Polytechnique de Paris; Its activities occupy over 600 people, scientists and research and innovation support staff, including 44 different nationalities.

Contexte et atouts du poste

This PhD work will be part of the MediTwin project (Dassault Systems/Inria), to offer new tools for the simulation of inhomogenous media. The PhD thesis will happen at Inria Saclay and within the Laboratoire d'Informatique de l'Ecole polytechnique (LIX), in the Turing building in Palaiseau, France.

Interested candidates should contact Jiong Chen (****@****.**) and Mathieu Desbrun (****@****.**) before Nov 1st, 2025 for full consideration. The expected starting date is Jan 6, 2026.

Mission confiée

The objective of this graduate work is to develop the theory and algorithm(s) for mesh-free, boundary-driven solvers for nonlinear PDEs arising in a wide range of applications in Computer Graphics and, more broadly, Computational Physics. The proposed method is supposed to feature an explicit and interpretable representation, allowing for straight-forward reduction or refinement of the solution space through online, adaptive resampling. In addition, it should support fine-grained parallelism to leverage modern GPU computing for large-scale problems.

Context. Despite decades of advancements in meshing algorithms, discretizing the inside of complex domain with high-quality elements remains a laborious process. Consequently, recent research has increasingly explored boundary-only [Ni et al., 2024; Sugimoto et al., 2022] and meshless representations [Chen et al., 2024; Schreck et al., 2019] for simulation. These approaches offer significant flexibility by avoiding volumetric tessellation, making them particularly well-suited for handling complex geometries, even enabling eficient adaptivity and refinement of simulation results. However, they also face substantial challenges: they often lead to dense linear systems that are costly to solve, and remain incapable of handling nonlinear problems. These limitations remain active topics of research, and to date, such methods have not consistently outperformed conventional numerical schemes (e.g., Finite Element Method), which is exactly the gap this dissertation attempts to address.

Linearization. Solving nonlinear problems eficiently often hinges on identifying an effective linear proxy, ensuring rapid and stable convergence toward the nonlinear solution. This could be achieved analytically by prescribing a suficiently good, if not optimal, approximate linear operator with known boundary representation (like Laplace's operator). The solution is then refined by iteratively solving a certain order of deformation equations in a boundary-only manner [Liao, 1997], progressively transitioning the so-lution from the linear to the nonlinear regime. More generally, a nonlinear equation can be linearized and solved numerically at each Newton step, where recent advance-ments in scalable solvers for linear boundary operators [Chen et al., 2024, 2025a] and their nonlinear, Gaussian-Process-based extensions [Chen et al., 2025b] can be applied, completely avoiding the painstaking meshing of the domain. These different strategies are intended to accommodate diverse objectives, such as computational cost constraints, flexibility in problem parameterization, and accuracy demands, and can be further im-plemented within either deterministic or stochastic frameworks (e.g., [Sawhney et al., 2023]) depending on specific applications.

R-Adaption. Unlike linear problems, nonlinear ones cannot generally be solved via a surface-only approach in nature: the solution typically cannot be determined from boundary conditions alone, which necessitates the discretization of the entire domain and introduces additional costs associated with these extra degrees of freedom. To improve the eficiency of this type of computations, we can leverage r-adaptive methods (or moving mesh methods [Huang and Russell, 2010]) by concentrating degrees of freedom where they are most needed. However, mesh-based r-adaptive methods are often inefficient and unstable: they require solving meshing equations simultaneously with the physical equations and ensuring the injectivity of the elements as well; otherwise, the entire simulation could be jeopardized. In sharp contrast, implementing r-adaption in a mesh-free context is significantly simpler and inherently more robust. We can flexibly add, remove, split, or merge point samples on demands, guided by the local statistics of the evolving solution. From a computational aspect, if the nonlinear operator (or its inverse) exhibits a similar screening effect to linear ones under certain transformations, we can expect this resampling process to cause only local updates to the entire system. This means it would introduce minimal overhead to current solver configurations, opening up possibilities for massively large-scale computations.

Principales activités

The candidate will perform research, write papers and help at times with reporting duties.

Compétences

The candidate must have extensive experience in computer graphics, and in particular, with computer animation and simulation, as proven by publications in these areas.

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 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

Rémunération

2300 euros gross per month