2025Activity reportProject-TeamCRAFT
RNSR: 202524734F- Research center Inria Centre at Université Grenoble Alpes
- In partnership with:Université de Grenoble Alpes, Institut polytechnique de Grenoble
- Team name: Computational design and fabRicAtion of FuncTional artefacts
- In collaboration with:Laboratoire Jean Kuntzmann (LJK)
Creation of the Project-Team: 2025 September 01
Each year, Inria research teams publish an Activity Report presenting their work and results over the reporting period. These reports follow a common structure, with some optional sections depending on the specific team. They typically begin by outlining the overall objectives and research programme, including the main research themes, goals, and methodological approaches. They also describe the application domains targeted by the team, highlighting the scientific or societal contexts in which their work is situated.
The reports then present the highlights of the year, covering major scientific achievements, software developments, or teaching contributions. When relevant, they include sections on software, platforms, and open data, detailing the tools developed and how they are shared. A substantial part is dedicated to new results, where scientific contributions are described in detail, often with subsections specifying participants and associated keywords.
Finally, the Activity Report addresses funding, contracts, partnerships, and collaborations at various levels, from industrial agreements to international cooperations. It also covers dissemination and teaching activities, such as participation in scientific events, outreach, and supervision. The document concludes with a presentation of scientific production, including major publications and those produced during the year.
Keywords
Computer Science and Digital Science
- A5.1.5. Body-based interfaces
- A5.1.8. 3D User Interfaces
- A5.5. Computer graphics
- A5.5.1. Geometrical modeling
- A5.6. Virtual reality, augmented reality
- A6.1. Methods in mathematical modeling
- A6.1.1. Continuous Modeling (PDE, ODE)
- A6.1.4. Multiscale modeling
- A6.1.5. Multiphysics modeling
- A6.2.6. Optimization
- A6.3.1. Inverse problems
- A6.5. Mathematical modeling for physical sciences
- A6.5.1. Solid mechanics
- A9.2. Machine learning
- A9.12.4. 3D and spatio-temporal reconstruction
Other Research Topics and Application Domains
- B5. Industry of the future
- B5.2. Design and manufacturing
- B5.5. Materials
- B5.7. 3D printing
- B9.2. Art
- B9.5. Sciences
- B9.5.1. Computer science
- B9.5.2. Mathematics
- B9.5.3. Physics
- B9.5.5. Mechanics
1 Team members, visitors, external collaborators
Research Scientists
- Melina Skouras [Team leader, INRIA, Researcher, from Sep 2025]
- Guillaume Coiffier [INRIA, Researcher, from Nov 2025]
- Xavier Tellier [INRIA, Starting Research Position, from Dec 2025]
Faculty Members
- Georges-Pierre Bonneau [UGA, Professor, from Sep 2025, HDR]
- Marco Freire [GRENOBLE INP, Associate Professor, from Sep 2025]
- Stefanie Hahmann [GRENOBLE INP, Professor, from Sep 2025, HDR]
PhD Students
- Ofir Mirkin [CNRS, from Oct 2025]
- Chandradeep Pokhariya [INRIA, from Oct 2025]
Administrative Assistants
- Diane Courtiol [INRIA]
- Nathalie Gillot [INRIA]
- Laura Leone [Randstad, from Aug 2025]
Visiting Scientist
- Hyunyoung Kim [UNIV BIRMINGHAM, from Nov 2025]
2 Overall objectives
Recent progress in technology, be it related to software, hardware or platforms’ accessibility, has created fantastic opportunities regarding what is nowadays possible to fabricate and who can exploit these new tools. 3D printing and laser cutting, for example, allows experts as well as casual users to create custom structures with a total control on their geometries. This enables the engineering of novel materials with complex mesostructures as well as the design of large-scale assemblies with arbitrary external shapes. Applications are virtually limitless and span fields as diverse as engineering, architecture, robotics, medicine and personalized fabrication. In practice, however, designing an object that is truly functional is highly challenging. Not only does the user need to be able to predict the behavior of the object, which is often unintuitive, but they also have to account for different, and sometimes conflicting, requirements for the object related for examples to ergonomics or aesthetics, in addition to its functionality. Some of these objectives solely depend on the shape of the object and can be evaluated from sketches or images (the style or the appearance of the object for example), others (the object’s stability, its rigidity, its robustness) require a full 3D representation as well as a mechanical model of the object and are evaluated using numerical simulation. Lastly, some aspects like the ergonomics of the object require a prototype to be fabricated so that the user can manipulate the object in real conditions. Therefore, in practice, designing a functional object implies iterating over exploratory, modeling, simulation and fabrication phases and requires the use of dedicated tools that are typically targeted to expert users.
CRAFT addresses the efficient design of objects whose function relates to their geometry and mechanics. We envision a workflow based on a system that will allow the user to sketch (in 2D or 3D) the shape of the object that they want to fabricate, let them easily specify its desired mechanical behavior as well as the materials to be used and automatically compute instructions usable by the manufacturing machine. The system will also allow the user to easily edit their input and will adjust the output accordingly.
To progress towards this long term objective, CRAFT strive to address the three following key areas:
-
User-interfaces The current design workflow involves too many different tools, and many of these tools are very difficult to use by non-experts. Indeed existing tools typically require the user to fully master all their parameters to be really effective and to design things in a certain order using pre-existing components, which might impede their creativity. Sketching, on the contrary, is a very intuitive way to specify geometries and behaviors. Unfortunately, the use of sketches is currently limited to the exploratory phase of the design pipeline.
The CRAFT team investigates the use of 2D and 3D sketches in subsequent stages of the pipeline, including whose where physics play an important role, i.e. we want to explore the use of sketches for modeling, designing and prototyping the object.
-
Materials The mechanical behavior of an object largely depends on its constitutive materials and their local arrangement or internal structure (e.g. mechanical properties of individual threads and shape of the knits in knitted textiles, size and density of the holes in porous materials, thickness and composition of layers in laminates). The mechanical properties of these materials limit the design space of what can be physically fabricated. Notably, the contraints related to materials are taken into account relatively late in the design cycle, as geometry and mechanics are typically treated sequentially. For example, if one wants to design a chair using a wooden plate, the chair will need to be split into several pieces to account for the limited flexibility of the wood. This means that the designer will need to go back to the first stage and specify the shape of a new chair consisting in an assembly of planar surfaces. On the contrary, if we manage to increase the compliance of the wood (e.g. by introducing cuts following kerfing techniques 21), it might be possible to get a chair whose shape better approximates the initial sketch, saving us one extra iteration through the design cycle.
The CRAFT team explores the design of such metamaterials (i.e. materials whose mechanical properties are principally dictated by the materials' internal geometric structures) and the numerical characterization of their effective geometric and mechanical behavior at macro scale. We focus on nonlinear metamaterials subject to large deformations, as this deformation regime enables their use for very interesting applications, related, e.g., to shape morphing.
-
Algorithms Efficient algorithms are key when designing mechanical structures. In many applications, there are already used to predict the mechanical behavior of the objects (forward simulation), but effective algorithms for inverse modeling, i.e. for automatically adjusting design parameters (rest shape of the object, distribution of base materials to be used,...) from high-level specifications involving both the geometric and mechanical behavior of the structure are still largely lacking. This might be explained by the difficulty to satisfy all the (conflicting) requirements for these tools: ideally, they should be robust, accurate and fast. In practice, the absence of such solvers translates into the need of iterating between modeling and simulation stages to obtain a design with desired properties.
The CRAFT team investigates multiple strategies to address these challenges and considers both inverse modeling algorithms that provide results to the user that are obtained fully automatically, and algorithms for adjusting existing designs based on edits made by the user, so that they can control the results.
3 Research program
Our research program is structured around three axes, each of which relates to one of the three areas discussed in Section 2. We detail these axes below.
3.1 Sketch-based systems that account for the physics
Participants: Stefanie Hahmann, George-Pierre Bonneau, Mélina Skouras, Guillaume Coiffier.
The design of a functional object, i.e. a piece of furniture, an architectural structure or a garment, typically starts by an exploration phase during which the designer sketches 2D representations of the object with variations in style. 3D versions corresponding to the drawings are then created using complex modeling software packages such as Blender or Maya. In the last few years, several systems have been proposed to facilitate this 2D to 3D conversion. However, since the problem of constructing a 3D shape from a 2D curve is fundamentally ill-posed (an infinite number of 3D points project to the same 2D point of a given plane), hypotheses on the properties of the 2D curves or 3D shape must be made. For example, True2Form 26 leverage design and perceptual principles to infer the depth of the drawn curves when lifted from 2D to 3D (e.g. sketched curves correspond to lines of principle curvatures of the 3D surface). The resulting 3D curves are then smoothly interpolated. The work by Fondevilla et al. 11, 12 and Jung et al. 14 are other examples of such tools where the underlying surfaces are supposed to be piecewise developable. Finally, in the recent years, machine-learning-based approaches have also been investigated for the task of converting 2D or 3D sketches to full 3D models 20, 10, 16.
While these approaches work well to infer the geometry of objects whose shapes are independent on the physics, these methods cannot be directly applied to reconstruct deformable objects that largely deforms under the effect of external forces, a garment that has stretched due to the contact of the fabric with the body, the panels of a tent subject to tensile forces, etc. This is also true for generative design methods, that currently cannot guarantee strict satisfaction of the physical constraints, all the more so that many designs that get created in the exploratory phase are very different from existing ones. On the other hand, prior work exists to invert the physics of given input objects 24, 23, 17. However, these approaches aim at matching idealized 3D target models given as input and do not support imprecise or partial data such as the one that arise from a sketch.
This research axis addresses the following question: is it possible to combine the strengths of these two lines of work and do better, in term of matching quality and/or performances, than simply running one tool after the other? In particular, we want to investigate the use of the physics of the object as a regularizer for the problem. Rather than interpolating curves in 3D and rely on priors related to the geometry (e.g. the 3D surfaces are developable or they should be as smooth as possible), we will assume that we know the material the object is made of and rely on priors related to the physics, i.e. satisfaction of physical laws (e.g. balance of forces).
In this research axis, we also want to explore the use of 3D curves sketched in virtual or augmented reality as input of the inverse problem. Another research question we want to address relates to the specifications of the mechanical behavior of the object and not only its geometry, e.g. to specify the magnitude and location of external loads.
3.2 Design and modeling of nonlinear metamaterials
Participants: Mélina Skouras, George-Pierre Bonneau, Stefanie Hahmann, Marco Freire.
Metamaterials are materials with non-ordinary macroscale properties that arise from the material's internal geometric structure. While some materials found in nature, such as bone structure, can be seen as metamaterials, most of these materials are human-engineered. A class of metamaterials that is particularly interesting for design is that of auxetic metamaterials, materials that orthogonally stretch (resp. compress) when uniaxially stretched (resp. compressed). Indeed, this behavior allows the material to significantly change its volume (or area, when considering surface materials), which can then be exploited to create curvature, following Gauss's Theorema Egregium 13.
In this research axis we explore the design and the modeling of metamaterial that exhibit both particular geometric properties (e.g. auxeticity) and mechanical properties (e.g. high and isotropic stiffness) as these materials are particularly useful for inverse design applications involving geometric and mechanical objectives. For example, for the case of the design of a deployable architectural structure, they would allow to leverage approaches based on conformal geometry and to obtain structures that are structurally sound. While most prior work focused on the design and modeling of metamaterials subject to small deformations 15, we want to explore the design of metamaterials subject to large transformations, typically associated to a nonlinear behavior. More specifically, we will focus on:
- planar metamaterials subject to large deformations
- curved metamaterials subject to small deformations
This research axis includes the discovery and parametrization of novel metamaterial families, the characterization of their mechanical properties, the numerical exploration of the spaces of geometric and mechanical properties and the numerical optimization of the geometry of these parametric metamaterials in order to improve their performances or to obtain metamaterials with target properties.
To characterize the geometric and mechanical behavior of the metamaterials we follow homogenization-based methods that consider periodic structures made of the repetition of a representative unit-cell with periodic boundary conditions and derive their macroscale behavior by a perturbative approach 22.
One of our goal is to find a representation for the homogenized materials that is as general as possible (for example, based on the interpolation of elasticity tensors) so that it can be reused for multiple types of metamaterials, or other structured materials like fabrics or foams. Indeed, such a representation will be very useful when developing inverse design algorithms as they will allow to abstract the fine-scale structure of the materials away and ease their generalization.
We will evaluate our models both numerically and experimentally. Numerical experiments will consist in comparing the behavior of one or repeated representative unit-cells when using the original mechanical model and a finely discretized mesh to that obtained when using the macro-scale model, enforcing periodicity. Regarding the real experiments, we will fabricate samples consisting in repeated cells and compare their actual behavior to that predicted in simulation using our own equipment or that of our collaborators.
3.3 Algorithms for solving inverse problems by leveraging geometric and mechanical properties
Participants: Mélina Skouras, George-Pierre Bonneau, Stefanie Hahmann, Guillaume Coiffier.
This research axis focuses on the development of effective algorithms for solving inverse design problems whose objective function is defined in terms of high-level specifications provided by the user and involves geometric and mechanical properties of the structure (e.g. desired target shape and stiffness for the object to design), and where constraints correspond to the physics of the system (e.g. force balance within the structure). When the displacement undergone by the structure from its reference state is small, the physical constraints can usually be linearized, and the problem to solve fits the frame of traditional topology and shape optimization, which has been actively investigated by the applied mathematics community 9. This is not the regime we are most interested in, which involves large deformations and leads to nonlinear constraints. Besides, one of our aims is to bring the physics into the exploratory stage of the design pipeline, in which specifications are often incomplete and imprecise. Our problems are then generally highly ill-posed and lack the regularity properties requested by typical shape optimization schemes. In this setting, our primary goal is not to analyze the convergence and sensibility properties of our algorithms, but rather to propose effective and practical algorithms that leverage the geometry and the mechanics of the structure to design.
There exists multiple methods for solving constrained minimizing problems such as Augmented Lagrangian methods, sequential quadratic programming or interior point methods 18. Ipopt 25, for example, is a very effective open-source implementation of the latter type, that we largely use in our work. In this research axis, we do not intend to contribute to the development of new such general-purpose minimizers and consider ourselves as users of constrained minimization solvers developed by other groups. Instead, we focus on the outer loop in which the minimizer is called and on strategies to obtain meaningful results, by relying on the geometry and the mechanics of the structure to design. Indeed, the minimizers mentioned above are all descent-based and only converge to local minima. Thus they require to be called with a relevant initial guess to provide an exploitable result. Besides, the original problem may have a prohibitively high number of variables and be computationally intractable.
In particular, we plan to explore multi-scale methods that solve the problem at multiple resolutions. The macro-scale representations of the metamaterials of our second research axis is particularly useful in this regard as they can be used as intermediate reduced models. Indeed, rather than optimizing for the internal structure of the material directly, we can work with homogenized parameters and then map them back to the original variables. While this type of approach has been used in the past, including by the members of the CRAFT team19, 27, 21, many questions remain open. For example, how to properly connect multiple representative elements of the considered metamaterial? How to properly grade these materials so that they best match the full-scale simulations (i.e. what is the maximal rate of local variation that is permitted without breaking the scale separation assumption at the core of the homogenization approach)? How to deal with metamaterials that exhibit finite curvature at rest? All these questions relate to geometric and mechanical properties of the structures, that need to be taken into account.
By working with metamaterials, we can convert discrete parameters to continuous fields of continuous variables with discontinuities located at certain points. Some of the resulting problems share similarities with those investigated in the computer graphics community in the context of surface parametrization or meshing, which also involve fields with potential singularities. However, a notable difference here is that we also need to consider the mechanics of the structure, e.g. to locate the singularities in areas free of high stresses, while parametrization methods typically only consider the geometry of the object.
We will evaluate our models and algorithms by direct comparison between simulations based on intermediate representations and full-scale simulations, as well as comparisons with results of experiments carried out in the fablab of our partners at Laboratoire Navier (ENPC).
4 Application domains
The applications of our research are very diverse and relevant, amongst others, to the fields of architecture, engineering, garment manufacturing, robotics, medicine, and art. In the short term, given our current network of collaborators, we will focus more specifically on applications related to architecture and civil engineering, garment manufacturing and art.
4.1 Architecture
Architectural structures need to satisfy multiple objectives: they need to be visually pleasing (aesthetics objective, related to the geometry of the structure), structurally sound (mechanical objective) and have a shape/space adapted to the intended use for the structure (functional objective), making them an ideal case study for our work. These structures can be made of rigid blocks, i.e. in the case of masonry-based structures, or made of flexible components, as in the case of tensile structures or gridshells. We are particularly interested in the latter category as they involve thin materials (beams, fabrics) that undergo large deformation. We will benefit from our strong collaboration with our partners from Laboratoire Navier at ENPC (Arthur Lebée, Olivier Baverel and Romain Mesnil) to work on this topic.
4.2 Garment design
Garments are typically made of flat panels that are sewn together to form a 3D shape. Designing the shapes of the panels that correspond to a desired 3D draped garment is a highly unintuitive task. Indeed, this implies finding the right layout for the panels as well as their geometry, which in turn depends on the mechanical properties of the fabric used. This interplay between topology, geometry and materials makes the design of bespoke garments particularly challenging, even more so if additional criteria such as reduction of waste are also taken into account. Members of the CRAFT team Mélina Skouras a nd Stefanie Hahmann have experience with efficient simulation and intuitive modeling of garments, e.g. using sketches. They have been collaborating on this topic with Adrien Bousseau (GraphDeco team at Inria Sophia-Antipolis) and intend to continue working on this line of work.
4.3 Art
Art can take many forms. In the context of this project, we are particularly interested in tangible expressions of art, e.g. through the design of sculptures or decorative items and plan to collaborate with artists to propose novel design tools to help the creation of artistic artefacts. For example, we have recently worked with the artist Pauline de Chalendar on the 3D surface reconstruction of freeform surfaces from 3D sketches and are now considering ways to actually fabricate real structures that are self-supported while keeping the unique artistic style arising from the use of 3D strokes.
5 Social and environmental responsibility
5.1 Footprint of research activities
With 5 permanent researchers and 3 non-permanent researchers, CRAFT is a relatively small team. We mostly work on our laptops and use a more powerful computer, which is more resource intensive, on a very limited basis, for very specific tasks. We all work from home on a regular basis and most of the team members bike to work or come to the office by public transportation. While we did not quantified the footprint of our research, we expect it to be limited.
5.2 Impact of research results
Our software is meant to be intuitive to use, and accessible to a wide range of users, experts but also fabrication hobbyists, students and researchers (in design, engineering, architecture, ...), artists. The output of our research will have a direct societal impact as it will change the way people design and fabricate objects, by allowing them to fully leverage the power of numerical technology.
Regarding environmental impact, our goal is to provide tools that ease the design of artefacts by reducing the number of iterations between the steps of the design cycle, and the number of intermediate fabricated prototypes required to converge towards the final design. Therefore, we expect our tools to contribute to save computing ressources and material ressources, even though this has not been finely quantified at this stage.
6 Highlights of the year
6.1 Creation of the CRAFT team
The CRAFT team was officially created in September 2025. It welcomed two newly recruited researchers Marco Freire and Guillaume Coiffier in addition to the previous members from ANIMA, Mélina Skouras and Stefanie Hahmann , and from the MAVERICK team, Georges-Pierre Bonneau .
6.2 PhD defense
Siyuan He, co-advised by Mélina Skouras and Arthur Lebée (ENPC), successfully defended his PhD in December 2025, in front of a multidisciplinary jury composed of researchers in computer graphics, physics and mechanics. Siyuan's work focused on the design of isotropic inflatable metamaterials and their use for 3D surface prototyping, combining discrete differential geometry, numerical simulation and optimisation, and fabrication.
6.3 Presentation at ACM SIGGRAPH Asia
RibbonSculpt, the work by Stefanie Hahmann and Georges-Pierre Bonneau , with their PhD student Anandhu Sureshkumar and collaborators Amal Dev Parakkat (LTCI/Télécom Paris) and Marie-Paule Cani (LIX/Ecole Polytechnique, Académie des Sciences), was presented at ACM SIGGRAPH Asia 2025, the most prestigious conference in computer graphics, along with ACM SIGGRAPH. RibbonSculpt is an interactive sketching system for creating freeform surfaces from 3D strokes in immersive environments (see Section 8.1).
7 Latest software developments, platforms, open data
7.1 Latest software developments
7.1.1 ribbonSculpt
-
Keywords:
3D reconstruction, 3D interaction
-
Functional Description:
RibbonSculpt takes as input a set of 3D strokes (OBJ format) and outputs a sculpted surface mesh.
It is a prototype implementation of the following paper:
RibbonSculpt: Voronoi Ball based 3D Sculpting from Sparse VR Ribbons Siggraph Asia 2025 Authors:
Anandhu Sureshkumar, LTCI-Telecom Paris, Institut Polytechnique de Paris, France Amal Dev Parakkat, LTCI-Telecom Paris, Institut Polytechnique de Paris, France Georges-Pierre Bonneau, Univ. Grenoble Alpes, CNRS, INRIA, Grenoble INP, LJK, France Stefanie Hahmann, Univ. Grenoble Alpes, CNRS, INRIA, Grenoble INP, LJK, France Marie-Paule Cani, LIX-Ecole Polytechnique/CNRS, Institut Polytechnique de Paris, France
- URL:
-
Contact:
Georges-Pierre Bonneau
-
Partners:
Telecom Paris, Ecole Polytechnique
8 New results
8.1 3D Sketch-based modeling: Surfacing hand-drawn 3D strokes
Participants: Georges-Pierre Bonneau, Stefanie Hahmann.
This work is done with our PhD student Anandhu Sureshkumar in collaboration with Amal Dev Parakkat (LTCI/Télécom Paris) and Marie-Paule Cani (LIX/Ecole Polytechnique). It is about surfacing sparse and rough 3D sketches (curve strokes) into a manifold surface best fitting the intended shape. While 3D sketches, see Figure 1(left), already provide the user with a good perception of the intended shape, they must be surfaced before any reuse in a downstream application. This remains a challenge when sketches are disconnected, disordered, sparse and noisy sets of strokes originating typically from hand-drawn sketches in Virtual Reality (VR). In 2025, we contributed to two novel 3D sketching and surfacing systems: VRSurf and RibbonSculpt.
Top row (left to right): A sparse set of unoriented 3D strokes, sampled version of the strokes used as input to our method, and VRSurf result, a closed manifold surface that interpolates through the samples. Second row: Different stages of VRSurf's balloon inflation process, starting from an initial user-given seed point.
VRSurf is a surface generation (or "surfacing") method able to solve the challenging problem of inferring closed surface geometries from any sparse set of unconstrained 3D strokes, thus offering flexibility and freedom to users drafting 3D shapes. Inspired by a balloon inflation metaphor 4 the method compensates for the lack of topological information and converts sparse and unoriented 3D stroke drawings into closed manifold surfaces. Seeded in the sparse scaffold formed by the strokes, a smooth, closed surface is inflated to progressively interpolate the input strokes, sampled into lists of points, see Figure 1.
We pose the problem as an interpolation problem over sampled points on the strokes. This is achieved thanks to an adaptive mesh, initialized as a maxi- mal sphere inside the set of strokes, which then inflates as a balloon and iteratively interpolates through the sample points. We cast the deformation method as the resolution of a biharmonic Laplacian with linear constraints augmented by an intermediate step of vir- tual surface expansion while continuously remeshing the surface to ensure robustness and maintain an organic appearance. If needed, extra balloons are automatically seeded and inflated to robustly pro- cess folded surfaces.
RibbonSculpt is the first method for interactive freeform shape design in VR through progressive sketching of sparse, oriented ribbons. It allows the real-time creation and progressive refinement of a closed surface of any topological genus. The resulting surface is watertight, manifold, and suitable for additive manufacturing.
RibbonSculpt is an interactive sketching system for surfacing 3D strokes in immersive environments 3 aimed at improving the VR-based modeling experience. Rather than reconstructing a surface from a completed VR sketch, our approach supports real-time, incremental surface generation by continuously updating a volumetric proxy derived from user-sketched ribbons. A mesh extracted at each step is smoothed using Laplacian-based energy minimization, yielding a surface that interpolates the ribbons and provides immediate visual feedback for iterative refinement.
Key to the method is a novel algorithm for computing at interactive frame rates a volumetric proxy, which, although transparent to the user, defines an abstract volume enclosed by the ribbon strokes and serves as a reference for surface generation. This proxy shape is computed and continuously updated using a new Voronoi ball filtering criterion, and is smoothed prior to display through a constrained surface fairing process based on local curvature.
8.2 MatAIRials: Isotropic Inflatable Metamaterials for Freeform Surface Design
Participants: Mélina Skouras, Siyuan He.
Our approach starts by precomputing the mapping between the parameters of the sealing patterns and the contraction ratios of the inflated metamaterials (a). Then, given an input 3D triangular mesh (b), we flatten it to the plane using conformal parametrization (c) and use the local scaling factor (color plot in Figure (c)) of the surface to trace the layout of the sealing patterns of the triangles of a superimposed regular grid (d). The resulting graded patterns can be used to fabricate a prototype whose curved inflated shape approximates the input mesh through metric frustration (e).
This work, done with our former PhD student Siyuan He and our former intern Meng-Jan Wu , in collaboration with Arthur Lebée , focuses on the design of inflatable pads, such as those used as mattresses or protective equipment. These are structures made of two planar membranes sealed according to periodic patterns, typically parallel lines or dots. In this work, we propose to treat these inflatables as metamaterials 1.
By considering novel sealing patterns with 6-fold symmetry, we are able to generate a family of inflatable materials whose macroscale contraction is isotropic and can be modulated by controlling the parameters of the seals. We leverage this property of our inflatable materials family to propose a simple and effective algorithm based on conformal mapping that allows us to design the layout of inflatable structures that can be fabricated flat and whose inflated shapes approximate those of given target freeform surfaces.
8.3 Programming developable surfaces using multilayer inflatables
Participants: Mélina Skouras, Ofir Mirkin.
Given a developable ruled surface (a), our method outputs two welding layouts (b) that can be used to seal three stacked membranes forming an inflatable bilayer consisting of staggered tubes with varying radii that best approximate the input surface when inflated (c). These layouts can then be used to fabricate an actual prototype made of TPU-coated nylon (d) whose deployed shape closely matches simulations.
This work is done with our former intern Ofir Mirkin (now PhD student in the team), in collaboration with our collaborators from the PMMH lab: José Bico and Etienne Reyssat and their former PhD student Nathan Vani . It presents a novel type of inflatable structure, made of three stacked membranes forming inflatable bilayers, consisting of staggered cylindrical or conical tubes. Our structures are fabricated flat using a quasi-inextensible material and bend out-of-plane when the two networks of tubular chambers are inflated to the same pressure. We program the local curvature of the structure by locally adjusting the width of the welding lines used to seal each pair of membranes 2.
We leverage this new architected inflatable structure in an inverse design algorithm that allows us to generate inflatable ribbons of a desired target shape. Our method takes as input a developable surface, represented as a 2D polyline or a 3D quad strip, and outputs the layouts of the three membranes to be welded together. Various prototypes made of thermosealable Thermoplastic PolyUrethane (TPU) or TPU-coated nylon are fabricated to demonstrate the capabilities of our simulation approach. Inflatable bilayers present as a promising platform for shape-morphing and soft robotics applications.
9 Partnerships and cooperations
9.1 International research visitors
9.1.1 Visits of international scientists
Hyunyoung Kim
-
Status
researcher
-
Institution of origin:
University of Birmingham
-
Country:
England
-
Dates:
November 2025
-
Context of the visit:
collaboration
-
Mobility program/type of mobility:
research stay (International Exchanges award from the Royal Society)
9.1.2 Visits to international teams
Research stays abroad
Stefanie Hahmann and Georges-Pierre Bonneau
-
Visited institution:
FU Berlin. Mathematical Geometry Processing group (Prof. Konrad Polthier)
-
Country:
Germany
-
Dates:
July 2025
-
Mobility program/type of mobility:
research stay
9.2 National initiatives
9.2.1 ANR MatAIRialS
Participants: Mélina Skouras, Siyuan He, Chandradeep Pokhariya, Xavier Tellier.
MatAIRialS (Architectured Inflatable mateRials for designing functional Shells) aims at investigating the design of large scale inflatable shells able to carry loads, made of superimposed membranes that are sealed according to quasi-periodic patterns and that locally behave as inflatable architected materials. Our idea is to pave the patterns and adjust their parameters so as to modulate the geometric and mechanical properties of the inflatable structure and in turn its deployed shape and stiffness. We intend to especially focus on the mechanical characterization of our architected materials, both numerically and experimentally, and to demonstrate the practical interest of our tools for real applications by fabricating prototypes of diverse sizes, including architectural-sized demonstrators.
This project is coordinated by Mélina Skouras and has as partners Laboratoire Navier (ENPC) and Laboratoire PMMH (ESPCI-CNRS-Université PSL).
10 Dissemination
10.1 Promoting scientific activities
10.1.1 Scientific events: organisation
General chair, scientific chair
- Georges-Pierre Bonneau : Conference co-chair of SMI 2025
- Stefanie Hahmann co-chaired the international conference on Geometric Modeling and Processing (GMP 2025).
- Mélina Skouras was the courses chair for ACM SIGGRAPH ASIA 2025
10.1.2 Scientific events: selection
Member of the conference program committees
- Marco Freire was a member of the program committee of the Journées de l'AFIG 2025.
- Stefanie Hahmann : Symposium on Solid and Physical Modeling (SPM’25)
- Stefanie Hahmann : Shape Modeling International (SMI’25)
- Stefanie Hahmann : Internat. conf. on Computer Graphics and Applications (GRAPP 2025)
- Georges-Pierre Bonneau : Solid and Physical Modeling (SPM’25)
- Georges-Pierre Bonneau : EnvirVis 2025
- Mélina Skouras : Eurographics 2025
- Mélina Skouras : Symposium on Geometry Processing (SGP) 2025
Reviewer
- CRAFT faculties are regular reviewers of most of the major journals and conferences of the domain.
10.1.3 Journal
Member of the editorial boards
- Georges-Pierre Bonneau : Computer Aided Design (CAD), Elsevier
- Georges-Pierre Bonneau : Computers & Graphics, Elsevier
- Stefanie Hahmann : Computer Aided Design (CAD), Elsevier
- Stefanie Hahmann : Computers & Graphics, Elsevier
- Mélina Skouras : associate editor for Computer Graphics Forum (CGF), Wiley
Reviewer - reviewing activities
- Mélina Skouras : ACM Transactions on Graphics, ASME Journal of Mechanical Design
10.1.4 Leadership within the scientific community
- Georges-Pierre Bonneau is member of the Scientific Council of the GdR IG-RV since 2023.
- Georges-Pierre Bonneau is co-chair of the "Prix de thèse" of the GDR IGRV (CNRS)
- Stefanie Hahmann is a member of the SMI (Shape Modeling International Association) steering committee.
- Stefanie Hahmann is an elected member of the European Association for Computer Graphics - chapitre français (EGFR) and serves as secretary in the steering committee.
10.1.5 Scientific expertise
- Stefanie Hahmann is a member of the SMI award committee since 2025.
- Stefanie Hahmann was Chair of the SMA Fellow Award committee, Solid Modeling Association.
- Stefanie Hahmann was a Jury member in the SMA Bézier Award committee, Solid Modling Association.
- Mélina Skouras was a member of the awards committee of the ACM SIGGRAPH / Eurographics Symposium on Computer Animation (SCA) 2025
10.1.6 Research administration
- Georges-Pierre Bonneau is a member of the Habilitation committee of the École Doctorale MSTII of Univ. Grenoble Alpes since 2005.
- Georges-Pierre Bonneau is head of the department Geometrie-Image of Laboratoire Jean Kuntzmann since 2023.
- Georges-Pierre Bonneau is a member of the scientific council of the Labex IRMIA++ Strasburg since 2020.
- Mélina Skouras is a member of the board (directoire) of the MSTIC pole at Université Grenoble Alpes since December 2023
10.2 Teaching - Supervision - Juries - Educational and pedagogical outreach
Georges-Pierre Bonneau and Stefanie Hahmann are both full Professor of Computer Science. Marco Freire is Associate Professor in Computer Science. He was recruited in September 2025 at Ensimag/Grenoble INP. They all teach general computer science and applied mathematics topics at basic and intermediate levels, and advanced courses in geometric modeling, computer graphics, and visualization at the master levels.
Georges-Pierre Bonneau is head of the Computer Science department of Polytech' Grenoble, Grenoble INP.
- Master: Marco Freire , Object Oriented Programming, 18h ETD, M1, Ensimag/Grenoble INP.
- Master: Marco Freire , Software Analysis, Design and Validation, 40h ETD, M1, Ensimag/Grenoble INP.
- Master: Marco Freire , Software Engineering Project, 10.5h, M1, Ensimag/Grenoble INP.
- Master : Stefanie Hahmann : Geometric Modeling, 46h, M1, Ensimag/Grenoble-INP.
- Master : Stefanie Hahmann : Surface Modeling, 37.5h, M2, Ensimag/Grenoble-INP.
- Master : Stefanie Hahmann : M2-projects, 10h, M2, Ensimag/Grenoble-INP.
- Bachelor: Stefanie Hahmann : Numerical methods, 87h, L3, Ensimag/Grenoble-INP.
- Bachelor: Stefanie Hahmann : Analysis, 31h, L1, UGA.
- Master: Georges-Pierre Bonneau : Image Synthesis, 23h, M1, Polytech-Grenoble, France
- Master: Georges-Pierre Bonneau : Data Visualization, 40h, M2, Polytech-Grenoble, France
- Master: Georges-Pierre Bonneau : Digital Geometry, 23h, M1, UGA
- Master: Georges-Pierre Bonneau : Information Visualization, 22h, Mastere, ENSIMAG, France.
- Master: Georges-Pierre Bonneau : Computer Graphics II, 18h, M2MoSiG, UGA, France.
- Mastère: Georges-Pierre Bonneau : Scientific Visualization, M2, ENSI France.
- Master: Mélina Skouras , Surface Modeling, 13.5h, M2, Ensimag/Grenoble-INP.
- Master: Mélina Skouras , Computer Graphics II, 18h, M2 MoSIG, UGA.
- Master: Mélina Skouras , Numerical mechanics, 9h, M2, Ensimag/Grenoble-INP.
10.2.1 Supervision
- Stefanie Hahmann co-supervises the PhD of Anandhu Sureshkumar with Georges-Pierre Bonneau , Amal Dev Parakkat (Télécom Paris), Marie-Paule Cani (Ecole Polytechnique).
- Mélina Skouras co-supervises the PhD of Chandradeep Pokhariya with Georges-Pierre Bonneau and Arthur Lebée (ENPC).
- Mélina Skouras co-supervises the PhD of Ofir Mirkin with Emmanuel Siéfert (LIPhy).
- Mélina Skouras co-supervised the PhD of Siyuan He (who graduated in December 2025) with Arthur Lebée (ENPC).
10.2.2 Juries
- Stefanie Hahmann served as a reviewer and member of the jury for Cyril Douthe's HDR defense, Labo. Navier, Ecole des Ponts, ParisTech 2025.
- Stefanie Hahmann served as the president of the jury for Abdelmagid's Ph.D. defense, ENS Architecture Paris-Malaquais, Univ. Gustav Eiffel, 2025.
- Comité de Sélection MCF 27: Stefanie Hahmann served as Vice-President of the recruiting committee (COS) for an Assistant Professor (Maître de Conférences) position in Computer Science at Ensimag/Grenoble INP.
- Comité de Sélection PR 27: Stefanie Hahmann served as member of the recruiting committee (COS) for a Full Professor position in Computer Science at INSA de Lyon.
- Georges-Pierre Bonneau served as the president of the jury for Briac Toussaint's PhD defense, UGA 2025.
- Georges-Pierre Bonneau served as reviewer and member of the jury for Mohammed Kissi's PhD defense, Univ. Paris Sciences et Lettres (PSL) 2025.
- Georges-Pierre Bonneau served as the president of the jury for Aymen Merrouche's PhD defense, UGA 2025.
- Mélina Skouras served as an examiner in the jury for Nicolas Montagne's PhD defense, Ecole des Ponts, ParisTech, 2025.
- Comité de Sélection MCF 27: Mélina Skouras served as a member of the recruiting committee (COS) for an Assistant Professor (Maître de Conférences) position in Computer Science at Aix Marseille Université.
- Mélina Skouras served as a member of the admission jury for the recruitement of Inria research officers (CRCN), 2025.
10.2.3 Educational and pedagogical outreach
Guillaume Coiffier and Marco Freire participated in this year's laboratory visit for the students of the École Normale Supérieure de Rennes, which took place in Grenoble (LIG, LJK, Inria). They presented the computer graphics and computational fabrication research fields, and the activities of the CRAFT research team.
11 Scientific production
11.1 Major publications
- 1 articleMatAIRials: Isotropic Inflatable Metamaterials for Freeform Surface Design.Computer Graphics Forum445August 2025HALDOIback to text
- 2 inproceedingsProgramming developable surfaces using multilayer inflatables.Proceedings of the 10th ACM Symposium on Computational FabricationSCF'25 - 10th ACM Symposium on Computational Fabrication23Cambridge (MA), United StatesNovember 2025, 1 - 13HALDOIback to text
- 3 inproceedingsRibbonSculpt: Voronoi Ball based 3D Sculpting from Sparse VR Ribbons.Proceedings of the SIGGRAPH Asia 2025 Conference PapersHong-Kong, ChinaACM2025, 1-11HALDOIback to text
- 4 articleVRSurf: Surface Creation from Sparse, Unoriented 3D Strokes.Computer Graphics Forum442May 2025HALDOIback to text
11.2 Publications of the year
International journals
International peer-reviewed conferences
11.3 Cited publications
- 9 inbookConception optimale de structures.Springer2006back to text
- 10 articleRapid 3D Model Generation with Intuitive 3D Input.2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)2024DOIback to text
- 11 articlePatterns from Photograph: Reverse-Engineering Developable Products.Computers and Graphics66August 2017, 4-13HALDOIback to text
- 12 articleFashion Transfer: Dressing 3D Characters from Stylized Fashion Sketches.Computer Graphics Forum4062021, 466-483HALDOIback to text
- 13 bookDisquisitiones generales circa superficies curvas.Typis Dieterichianis, Gottingae1828back to text
- 14 articleSketching Folds: Developable Surfaces from Non-Planar Silhouettes.ACM Transactions on Graphics345October 2015, 155:1--155:12HALDOIback to text
- 15 article3D metamaterials.Nature Reviews Physics2019DOIback to text
- 16 article3D VR Sketch Guided 3D Shape Prototyping and Exploration.2023 IEEE/CVF International Conference on Computer Vision (ICCV)2023DOIback to text
- 17 articleInverse Elastic Shell Design with Contact and Friction.ACM Transactions on Graphics376November 2018, 1-16HALDOIback to text
- 18 inbookNumerical Optimization.Springer2006back to text
- 19 articleElastic Textures for Additive Fabrication.ACM Trans. Graph.3442015DOIback to text
- 20 articleStrokes2Surface: Recovering Curve Networks From 4D Architectural Design Sketches.Computer Graphics Forum4322024, e15054URL: https://onlinelibrary.wiley.com/doi/abs/10.1111/cgf.15054DOIback to text
- 21 articleComputational Design of Laser-Cut Bending-Active Structures.Computer-Aided Design151103335Best Paper Award at Solid and Physical Modeling 2022October 2022, 1-12HALDOIback to textback to text
- 22 bookHomogenization of Thick and Heterogeneous Plates.Wiley2015DOIback to text
- 23 articleComputational design of actuated deformable characters.ACM Trans. Graph.3242013DOIback to text
- 24 articleDesigning inflatable structures.ACM Trans. Graphics3342014DOIback to text
- 25 articleIPOPT: An interior point algorithm for large-scale nonlinear optimization.2002back to text
- 26 articleTrue2Form: 3D Curve Networks from 2D Sketches via Selective Regularization.ACM Transactions on Graphics3342014HALDOIback to text
- 27 articleTwo-Scale Topology Optimization with Microstructures.ACM Trans. Graph.3642017DOIback to text