Design-by-Morphing: A Morphing-Centered Design Approach

• Design-by-Morphing is a design approach that encodes feasible transformation pathways using morphing parameters, facilitating optimized aerodynamic, structural, and kinematic performance.
• It leverages techniques like baseline interpolation, physics-aware inverse design, and differentiable simulations to manage complex shape morphing and material variations.
• The framework integrates geometric, material, and control variables to produce robust solutions in applications ranging from airfoil optimization to morphing vehicles.

Searching arXiv for recent and foundational papers on "Design-by-Morphing" and closely related shape-morphing design frameworks. {"query":"all:\"design-by-morphing\" OR ti:\"Design-by-Morphing\"","max_results":10,"sort_by":"relevance"} {"query":"Design-by-Morphing arXiv","max_results":10,"sort_by":"relevance"} Design-by-Morphing denotes a family of design procedures in which morphing is itself the design medium: instead of prescribing a final geometry directly, one parameterizes a morphing process, a morphable mechanism, or a morphing-capable design space, and then optimizes or analytically synthesizes that representation toward aerodynamic, structural, kinematic, or multifunctional objectives. In the arXiv literature, the term has been used for linear combinations of baseline shapes in fluid and airfoil optimization, for inverse design of sheets, kirigami, and composites that morph into target shapes, and for co-design of topology, actuation, and control in morphing vehicles (Sheikh et al., 2022, Wang et al., 2023, Bergonti et al., 2023). The literature suggests that the unifying idea is not a single algorithm but a shift from direct geometry editing to a morphing-centered parameterization of feasible transformations.

1. Core meanings and formulations

The literature uses the same label for several technically distinct formulations. In one formulation, morphing defines a search space over existing geometries. In another, morphing is the target behavior of a structure whose geometry, material distribution, or internal energy landscape is synthesized so that a stimulus or load drives the desired transformation. A third usage couples morphology and control, treating morphing joints, actuator choices, and flight strategy as a joint optimization problem.

Interpretation in the literature	Representative formulation	Representative papers
Baseline-shape morphing	$$P(x)=\frac{1}{\sum_{m=1}^N w_m}\sum_{n=1}^N w_n S_n(x)$$; $$S(\mathbf{x};\mathbf{w})=S_0(\mathbf{x})+\sum_{p=1}^n w_p\Delta S_p(\mathbf{x})$$; $A(x^*)=(a_n/\gamma)\sum_{j=1}^5 \omega_j A_j(x^*)$	(Sheikh et al., 2022, Sheikh et al., 2022, Farooq et al., 27 Sep 2025)
Physics-aware inverse design of morphing structures	$$\Phi(\mathbf{x})=F_{\rm phys}+\lambda_{\rm shape}\sum g_i^2+\lambda_{\rm geom}\sum c_j^2$$; graded elastica; porous tapered elastica	(Wang et al., 2023, Kansara et al., 2023, Zhang et al., 2022)
Energy-landscape programming for bistability	spring potential $$V_s=\tfrac12 k_s(\|\mathbf p_{1s}-\mathbf p_{2s}\|_2-\ell_s)^2$$ with first- and second-order stability constraints	(Bharaj et al., 2018)
Morphology-control co-design	$$J_1(x_h)=\frac1{n_s}\sum E_i,\;J_2(x_h)=\frac1{n_s}\sum T_i$$ with nested trajectory optimization	(Bergonti et al., 2023)

A plausible implication is that Design-by-Morphing is best understood as a methodological umbrella rather than a single parameterization. What remains invariant is the decision to encode design intent in morphing variables that already reflect attainable deformations, rather than in purely geometric coordinates.

2. Morphing as a search space over existing shapes

In aerodynamic and hydrodynamic optimization, Design-by-Morphing is used as a low-dimensional yet extrapolative shape basis. In "Airfoil Optimization using Design-by-Morphing" (Sheikh et al., 2022), the design space is constructed from $N$ pre-existing, topologically equivalent airfoils $S_1,\dots,S_N$, each represented as a collocation vector of $y$-coordinates sampled at $F+1$ uniformly spaced chordwise locations. A new airfoil is generated by normalized linear combination of the baselines, with weights $$S(\mathbf{x};\mathbf{w})=S_0(\mathbf{x})+\sum_{p=1}^n w_p\Delta S_p(\mathbf{x})$$0. When all weights are nonnegative the construction is interpolatory; allowing negative weights performs extrapolation beyond the convex hull of the baselines. Because extrapolation can generate self-intersections, the method adds collision removal, reinjection of points, and smoothing. The bi-objective optimization targets $$S(\mathbf{x};\mathbf{w})=S_0(\mathbf{x})+\sum_{p=1}^n w_p\Delta S_p(\mathbf{x})$$1 and $$S(\mathbf{x};\mathbf{w})=S_0(\mathbf{x})+\sum_{p=1}^n w_p\Delta S_p(\mathbf{x})$$2, and an NSGA-II search over a 25-dimensional weight vector produced about 80 non-dominated airfoils spanning $$S(\mathbf{x};\mathbf{w})=S_0(\mathbf{x})+\sum_{p=1}^n w_p\Delta S_p(\mathbf{x})$$3 and $$S(\mathbf{x};\mathbf{w})=S_0(\mathbf{x})+\sum_{p=1}^n w_p\Delta S_p(\mathbf{x})$$4. Using 25 baselines drawn from the UIUC database, the method reconstructed 1,618 of 1,620 airfoils at MAE $$S(\mathbf{x};\mathbf{w})=S_0(\mathbf{x})+\sum_{p=1}^n w_p\Delta S_p(\mathbf{x})$$5, and 98% at MAE $$S(\mathbf{x};\mathbf{w})=S_0(\mathbf{x})+\sum_{p=1}^n w_p\Delta S_p(\mathbf{x})$$6 (Sheikh et al., 2022).

A closely related formulation appears in "Optimization of the Shape of a Hydrokinetic Turbine's Draft Tube and Hub Assembly Using Design-by-Morphing with Bayesian Optimization" (Sheikh et al., 2022). There, a reference geometry $$S(\mathbf{x};\mathbf{w})=S_0(\mathbf{x})+\sum_{p=1}^n w_p\Delta S_p(\mathbf{x})$$7 and morphing modes $$S(\mathbf{x};\mathbf{w})=S_0(\mathbf{x})+\sum_{p=1}^n w_p\Delta S_p(\mathbf{x})$$8 define the candidate surface $$S(\mathbf{x};\mathbf{w})=S_0(\mathbf{x})+\sum_{p=1}^n w_p\Delta S_p(\mathbf{x})$$9. The resulting shape is evaluated by CFD, and the expensive black-box optimization is handled by Mixed-Variable Multi-Objective Bayesian Optimization with Gaussian-process surrogates and a HedgeMO acquisition portfolio. For the hydrokinetic turbine case, 50 warm-up shapes and 75 epochs of batch evaluation yielded 425 total CFD runs; the best design was reported at $A(x^*)=(a_n/\gamma)\sum_{j=1}^5 \omega_j A_j(x^*)$0, approximately 2.7% above the Sharp-heel baseline's 0.9370 (Sheikh et al., 2022).

The same logic extends from static geometry to prescribed kinematics. In "Optimized Fish Locomotion using Design-by-Morphing and Bayesian Optimization" (Farooq et al., 27 Sep 2025), the swimmer envelope is written as a normalized linear superposition of five baseline amplitude profiles, including anguilliform, carangiform, horizontal, and two unconventional "weird" shapes. The weights satisfy $A(x^*)=(a_n/\gamma)\sum_{j=1}^5 \omega_j A_j(x^*)$1, with a hyperspherical-angle parameterization removing one redundant degree of freedom. Bayesian optimization over the shape coefficients together with undulation frequency $A(x^*)=(a_n/\gamma)\sum_{j=1}^5 \omega_j A_j(x^*)$2 and non-dimensional wavelength $A(x^*)=(a_n/\gamma)\sum_{j=1}^5 \omega_j A_j(x^*)$3 produced an optimal propulsive efficiency of 82.4%, whereas the second- and third-best profiles achieved 51.8% and 42.8%, respectively (Farooq et al., 27 Sep 2025).

These formulations share a specific technical advantage: a small set of morphing weights can induce global, correlated shape changes, including extrapolative ones, without imposing spline orders or local control-point topologies. The literature also makes clear that this advantage is balanced by dependence on baseline selection and the possibility of invalid geometries under extrapolation (Sheikh et al., 2022).

3. Physics-aware inverse design of morphing structures

A second major line of work uses Design-by-Morphing for inverse design of structures that must physically morph into a prescribed target shape. In "Physics-aware differentiable design of magnetically actuated kirigami for shape morphing" (Wang et al., 2023), the design variables are the planar coordinates of kirigami nodes and a magnetization parameter $A(x^*)=(a_n/\gamma)\sum_{j=1}^5 \omega_j A_j(x^*)$4 for each panel. The deployed state is parameterized by a single degree of freedom $A(x^*)=(a_n/\gamma)\sum_{j=1}^5 \omega_j A_j(x^*)$5, with nodal positions given analytically by $A(x^*)=(a_n/\gamma)\sum_{j=1}^5 \omega_j A_j(x^*)$6 through a composition of Euclidean transforms. The total energy is

$A(x^*)=(a_n/\gamma)\sum_{j=1}^5 \omega_j A_j(x^*)$7

with hinge bending energy and magnetic potential under a uniform field $A(x^*)=(a_n/\gamma)\sum_{j=1}^5 \omega_j A_j(x^*)$8. Shape matching, rigid-deployability, and equilibrium are combined in a constrained optimization solved by SQP in MATLAB. The reported acceleration is substantial: the one-dimensional solve for equilibrium angle takes $A(x^*)=(a_n/\gamma)\sum_{j=1}^5 \omega_j A_j(x^*)$9 versus $$\Phi(\mathbf{x})=F_{\rm phys}+\lambda_{\rm shape}\sum g_i^2+\lambda_{\rm geom}\sum c_j^2$$0model in full 3D FEA, typical SQP convergence is $$\Phi(\mathbf{x})=F_{\rm phys}+\lambda_{\rm shape}\sum g_i^2+\lambda_{\rm geom}\sum c_j^2$$1 iterations, and total design time is $$\Phi(\mathbf{x})=F_{\rm phys}+\lambda_{\rm shape}\sum g_i^2+\lambda_{\rm geom}\sum c_j^2$$2. Reported results include a 3×3 kirigami designed in 4 min with shape-error $$\Phi(\mathbf{x})=F_{\rm phys}+\lambda_{\rm shape}\sum g_i^2+\lambda_{\rm geom}\sum c_j^2$$3 for circular deployment at $$\Phi(\mathbf{x})=F_{\rm phys}+\lambda_{\rm shape}\sum g_i^2+\lambda_{\rm geom}\sum c_j^2$$4, arbitrary target shapes with $$\Phi(\mathbf{x})=F_{\rm phys}+\lambda_{\rm shape}\sum g_i^2+\lambda_{\rm geom}\sum c_j^2$$5 boundary error, and a two-way contractible kirigami with three stable states at zero and $$\Phi(\mathbf{x})=F_{\rm phys}+\lambda_{\rm shape}\sum g_i^2+\lambda_{\rm geom}\sum c_j^2$$6 (Wang et al., 2023).

In "Inverse design and additive manufacturing of shape-morphing structures based on functionally graded composites" (Kansara et al., 2023), the inverse problem is posed through the graded elastica equation, with local bending stiffness $$\Phi(\mathbf{x})=F_{\rm phys}+\lambda_{\rm shape}\sum g_i^2+\lambda_{\rm geom}\sum c_j^2$$7 and modulus determined by the Voigt rule of mixtures $$\Phi(\mathbf{x})=F_{\rm phys}+\lambda_{\rm shape}\sum g_i^2+\lambda_{\rm geom}\sum c_j^2$$8. Given a target axisymmetric profile $$\Phi(\mathbf{x})=F_{\rm phys}+\lambda_{\rm shape}\sum g_i^2+\lambda_{\rm geom}\sum c_j^2$$9, the framework analytically computes the strip width and modulus profile needed to realize the target curvature, avoiding an iterative inverse solver. The finite-element pipeline includes a voxelated-solid model and a reduced shell model, and additive manufacture is performed on a Stratasys Objet260 Connex using FLX9070 and FLX9095 photopolymers. For the hemisphere case, the reported RMS deviation is $$V_s=\tfrac12 k_s(\|\mathbf p_{1s}-\mathbf p_{2s}\|_2-\ell_s)^2$$0 of the span (Kansara et al., 2023).

"Shape-morphing structures based on perforated kirigami" (Zhang et al., 2022) replaces variable thickness by variable porosity in a sheet of uniform thickness. The local bending stiffness is written as $$V_s=\tfrac12 k_s(\|\mathbf p_{1s}-\mathbf p_{2s}\|_2-\ell_s)^2$$1, and the porosity function is computed from a porous tapered elastica model so that a prescribed axisymmetric target profile is attained upon buckling. The paper reports finite-element and experimental validation, and examines the load-bearing capacity of morphed half-ellipsoids via indentation tests, with a geometric rigidity metric $$V_s=\tfrac12 k_s(\|\mathbf p_{1s}-\mathbf p_{2s}\|_2-\ell_s)^2$$2 (Zhang et al., 2022).

Taken together, these papers replace purely kinematic matching by inverse design formulations in which geometry, material variation, and stimulus enter the governing mechanics explicitly. This suggests that, in this branch of Design-by-Morphing, feasible morphing is treated as a constrained mechanics problem rather than a post hoc validation step.

4. Unit-cell mechanisms, tessellation, and stability

The review "Shape Morphing Metamaterials" (Dudek et al., 14 Jan 2025) provides a general classification that helps organize many morphing-design papers. It identifies two main unit-cell mechanisms: structural anisotropy and internal rotations. Structural anisotropy uses direction-dependent stiffness to induce bending, twisting, or Gaussian-curvature changes, while internal-rotation cells rely on nearly rigid struts or panels connected by slender hinges, often modeled with a torsional spring $$V_s=\tfrac12 k_s(\|\mathbf p_{1s}-\mathbf p_{2s}\|_2-\ell_s)^2$$3. The same review distinguishes two tessellation strategies: kinematic compatibility, in which adjacent cells must deform without gaps or overlaps, and geometric frustration, in which closed loops cannot satisfy all local preferences simultaneously and the frustration energy can drive instability or buckling (Dudek et al., 14 Jan 2025).

This classification maps directly onto several Design-by-Morphing implementations. Functionally graded composites and perforated kirigami belong to the structural-anisotropy class because local bending stiffness is programmed through modulus or porosity fields (Kansara et al., 2023, Zhang et al., 2022). Magnetically actuated kirigami and many origami-like frameworks belong to the internal-rotation class because rigid panels are connected by hinges and deployment is dominated by rotational kinematics (Wang et al., 2023). The review further organizes the design process into four stages—target and mechanism mapping, parameter selection and optimization, numerical simulation and validation, and prototyping and experimental testing—which aligns closely with the workflows reported across these papers (Dudek et al., 14 Jan 2025).

A more explicit energy-landscape version appears in "Metamorphs: Bistable Planar Structures" (Bharaj et al., 2018). There, two desired planar forms are sampled into rigid-bar linkages, and internal springs are added with design variables $$V_s=\tfrac12 k_s(\|\mathbf p_{1s}-\mathbf p_{2s}\|_2-\ell_s)^2$$4. The objective is to make both target configurations strict local minima of the total spring potential by enforcing first-order force balance and a positive-definite reduced Hessian. The nonlinear program minimizes a force-residual plus energy regularizer subject to eigenvalue constraints, and the solution strategy uses iterative spring addition guided by modal analysis, then an Augmented Lagrangian Method with BFGS. The result is a computational design method for planar structures that can morph into two different bistable forms and hold them under external force application (Bharaj et al., 2018).

A different stability mechanism is used in "Morpho-plastic cellular metamaterials" (Charpentier et al., 2024). There, irreversible plastic deformation at curved hinge regions produces residual opening angles $$V_s=\tfrac12 k_s(\|\mathbf p_{1s}-\mathbf p_{2s}\|_2-\ell_s)^2$$5 after unloading, so the structure memorizes curvature without embedded actuators, fluids, or swelling. The inverse-mapping workflow discretizes a target curve or target Gauss-curvature field, uses analytical mappings such as $$V_s=\tfrac12 k_s(\|\mathbf p_{1s}-\mathbf p_{2s}\|_2-\ell_s)^2$$6, and then fabricates a flat plate or chain that is programmed by stretching and releasing. Reported examples include spirals, hearts, self-intersecting curves, domes, saddles, and multistable plates, with load-to-self-weight ratio approximately 20 under point loading at the apex and residual joint tilt angles up to $$V_s=\tfrac12 k_s(\|\mathbf p_{1s}-\mathbf p_{2s}\|_2-\ell_s)^2$$7 per cell (Charpentier et al., 2024).

5. Optimization, differentiability, and learned surrogates

A persistent difficulty in Design-by-Morphing is that high-fidelity evaluation is expensive. Several papers therefore focus on surrogate, differentiable, or operator-learning replacements for brute-force simulation.

The differentiable kirigami framework already illustrates one route: exact analytic Jacobians of the kinematic map and closed-form energies enable back-propagation through equilibrium and make minute-scale inverse design practical, with equilibrium-angle solves below $$V_s=\tfrac12 k_s(\|\mathbf p_{1s}-\mathbf p_{2s}\|_2-\ell_s)^2$$8 and total design times of 2–8 min per target (Wang et al., 2023). A data-driven alternative is "SimuLearn: Fast and Accurate Simulator to Support Morphing Materials Design and Workflows" (Yang et al., 2020), which combines Abaqus-generated training data with graph-based Interaction Networks. The reported rollout time is approximately 0.61 s on an 8-core Intel i9, compared with approximately 10 min per trial for plain FEA, yielding more than $$V_s=\tfrac12 k_s(\|\mathbf p_{1s}-\mathbf p_{2s}\|_2-\ell_s)^2$$9 speedup, with mean vertex-coordinate error approximately 2.9 mm and 97% fidelity compared to FEA (Yang et al., 2020).

"Shape-morphing programming of soft materials on complex geometries via neural operator" (Chen et al., 16 Jan 2026) extends surrogate modeling to irregular computational domains. Its Spectral and Spatial Neural Operator fuses Laplacian eigenfunction encoding with graph convolutions, is trained with AdamW and OneCycleLR, and is combined with a genetic algorithm for voxel-level inverse design. Reported forward-prediction errors include L2 values of 1.19% for a dart, 0.55% for a human shape, 0.35% for a stingray, 0.91% for a dome, 1.13% for a butterfly, and 0.12% for a thin-walled structure. The paper also reports discretisation invariance and super-resolution design: in the 290-voxel dart case, zero-shot prediction achieved L2 $$J_1(x_h)=\frac1{n_s}\sum E_i,\;J_2(x_h)=\frac1{n_s}\sum T_i$$0, falling to L2 $$J_1(x_h)=\frac1{n_s}\sum E_i,\;J_2(x_h)=\frac1{n_s}\sum T_i$$1 after fine-tuning on 5k samples (Chen et al., 16 Jan 2026).

A further development is "PhysMorph-GS: Differentiable Shape Morphing via Joint Optimization of Physics and Rendering Objectives" (Song et al., 21 Nov 2025), which closes a "rendering gap" by coupling differentiable MPM with 3D Gaussian splatting. A deformation-aware upsampling bridge maps sparse particles to dense Gaussians, and image-space silhouette and depth losses are injected into the physics adjoint through a multi-pass interleaved optimization. On challenging sequences, the depth-supervised variant reduces Chamfer distance by about 2.5 percent relative to a physics-only baseline, while the full model captures thin features such as ears and tails better than the baseline (Song et al., 21 Nov 2025).

A plausible implication is that Design-by-Morphing is moving from explicit low-dimensional parameterizations toward hybrid representations in which morphing variables can be geometric, material, latent, or image-supervised, provided the forward map remains sufficiently learnable or differentiable to support inverse design.

6. Morphing vehicles, propulsors, and control-coupled design

In robotic and aerospace systems, Design-by-Morphing frequently appears as co-design: morphology is optimized jointly with control, mission profile, or actuation strategy. In "Co-Design Optimisation of Morphing Topology and Control of Winged Drones" (Bergonti et al., 2023), the hardware variables include wing chord, span, mounting position, rest-angle rotation, number and axes of revolute joints, servo model, and propulsion unit, while the trajectory variables include joint angles, torques, thrust, base pose, and time step. For each candidate hardware design, direct multiple-shooting trajectory optimization in CasADi/Ipopt is solved over multiple scenarios, and an outer-loop NSGA-II search on $$J_1(x_h)=\frac1{n_s}\sum E_i,\;J_2(x_h)=\frac1{n_s}\sum T_i$$2 and $$J_1(x_h)=\frac1{n_s}\sum E_i,\;J_2(x_h)=\frac1{n_s}\sum T_i$$3 returns a Pareto front over energy and mission time. In 972 parametric scenarios, four representative morphing drones were compared to the commercial fixed-wing "bixler3," with reported average improvements per meter of 37% to 74% in energy saving and 22% to 33% in time saving (Bergonti et al., 2023).

Mechanical implementations make the morphology-control coupling explicit. "Design and Control of an Actively Morphing Quadrotor with Vertically Foldable Arms" (Yeh et al., 4 Aug 2025) uses a parallelogram arm linkage so that propeller orientation remains constant during morphing, driven by a single central servomotor with gears and racks. The prototype shrinks to 67% of its original size when fully folded, and an adaptive sliding-mode controller with a disturbance observer reduces tracking errors during folding, grasping, and narrow-gap passage. Reported experiments include circular-flight tracking with positional error below 0.05 m during repeated fold-unfold cycles (Yeh et al., 4 Aug 2025).

"FAST-Hex -- A Morphing Hexarotor: Design, Mechanical Implementation, Control and Experimental Validation" (Ryll et al., 2020) takes a different route: six propellers are synchronously tilted by one additional actuator, allowing transition from an under-actuated collinear configuration at $$J_1(x_h)=\frac1{n_s}\sum E_i,\;J_2(x_h)=\frac1{n_s}\sum T_i$$4 to a fully actuated configuration for $$J_1(x_h)=\frac1{n_s}\sum E_i,\;J_2(x_h)=\frac1{n_s}\sum T_i$$5. The force-efficiency index simplifies to $$J_1(x_h)=\frac1{n_s}\sum E_i,\;J_2(x_h)=\frac1{n_s}\sum T_i$$6 for balanced hovering, so morphing directly trades actuation richness against efficiency. The platform uses one additional control input to achieve configurability and full actuation in a vast state space, with a maximum tilt of $$J_1(x_h)=\frac1{n_s}\sum E_i,\;J_2(x_h)=\frac1{n_s}\sum T_i$$7 (Ryll et al., 2020).

Morphing can also be tuned at the level of propulsive kinematics rather than vehicle topology. "Propulsive performance of morphing and heaving foil" (Singh et al., 2022) varies the extent of morphing $$J_1(x_h)=\frac1{n_s}\sum E_i,\;J_2(x_h)=\frac1{n_s}\sum T_i$$8 and the morph initiation point $$J_1(x_h)=\frac1{n_s}\sum E_i,\;J_2(x_h)=\frac1{n_s}\sum T_i$$9 in a sinusoidally heaving NACA 0015 foil at $N$0 and $N$1. The reported optimum is maximum mean thrust $N$2 at $N$3 and $N$4, while maximum propulsive efficiency $N$5 occurs at $N$6 under the same onset condition (Singh et al., 2022).

7. Scope, limitations, and directions of development

A common misconception is that Design-by-Morphing refers to a single design algorithm. The literature instead shows several distinct practices: baseline interpolation-extrapolation in fluid design, inverse mechanics of morphing sheets and kirigami, energy shaping for bistability, and co-design of morphology and control in vehicles (Sheikh et al., 2022, Wang et al., 2023, Bergonti et al., 2023, Bharaj et al., 2018). A second misconception is that it necessarily requires active materials. The record is broader: morphing can be driven by magnetic actuation, thermal expansion, residual stress, plastic deformation, simple stretching-and-release, or purely geometric superposition of pre-existing shapes (Wang et al., 2023, Charpentier et al., 2024, Yang et al., 2020, Sheikh et al., 2022).

The literature also identifies recurring limitations. Extrapolative baseline morphing can generate self-intersections or omit shape families not spanned by the chosen baselines (Sheikh et al., 2022). CFD- or FEA-driven frameworks remain expensive unless paired with surrogates or differentiable models (Sheikh et al., 2022, Yang et al., 2020). Learned simulators and neural operators inherit training-domain limitations; SimuLearn omits collision handling for its material system, and extension to larger topologies requires more data (Yang et al., 2020). Morpho-plastic programming is most accurate under force control, although displacement control also works (Charpentier et al., 2024). In morphing aerial vehicles, increased actuation capability can reduce thrust efficiency, as in the $N$7 trade-off of FAST-Hex (Ryll et al., 2020).

At the same time, several papers explicitly point to broader applicability. The differentiable kirigami framework is described as extensible to thermal and hygroscopic systems and to other hinged structures such as origami, lattices, and tensegrities (Wang et al., 2023). Neural-operator design is reported on irregular-boundary shapes, porous structures, thin-walled structures, modular assemblies, and joint training across multiple geometries (Chen et al., 16 Jan 2026). The metamaterials review frames shape morphing through unit-cell mechanism and tessellation strategy rather than through any specific actuation route, which suggests a durable conceptual foundation for future work (Dudek et al., 14 Jan 2025).

Across these strands, Design-by-Morphing has evolved into a research program centered on one principle: the design variables are chosen so that they already encode feasible transformation pathways. Whether those variables are morphing weights, hinge rotations, porosity fields, modulus gradients, plastic residual angles, or control-coupled topology parameters, the objective is the same—to make morphing not an afterthought of design, but its organizing representation.