Scale-Translation Rigidity: Principles & Applications

• Scale-translation rigidity is defined as the property where a structure’s inter-neighbor bearings or dihedral angles uniquely determine its embedding up to translation and uniform scaling.
• The framework’s rigidity is verified via the bearing rigidity matrix, where the rank condition (rank = dn - (d+1)) confirms that only trivial motions (translation and scaling) are allowed.
• This concept is pivotal in applications such as formation control, network localization, and shape morphing, offering robust computational methods even in higher dimensions.

Scale-translation rigidity denotes the property of a geometric structure—typically a framework or a mesh embedding—being uniquely determined, up to global translation and uniform scaling (and, in some generalizations, up to rotation), by a particular invariant. In the context of frameworks, this invariant is a collection of inter-neighbor bearing directions; in triangulated surface meshes, it can be the set of dihedral angles. The essential feature is that equivalence of these invariants implies the two underlying realizations differ by no more than a composition of translation, scaling, (and possibly rotation). This concept is a core object of study in rigidity theory, formation control, and shape analysis; it is often termed "bearing rigidity" in the framework literature (Zhao et al., 2014) and is closely related to "dihedral rigidity" in shape processing (Amenta et al., 2018).

1. Frameworks, Bearings, and the Rigidity Matrix

Given an undirected graph $G = (V, E)$ with $n$ vertices and $m$ edges, a framework $G(p)$ consists of an assignment $$p = [p_1^\top, ..., p_n^\top]^\top \in \mathbb{R}^{dn}$$ with each $p_i \in \mathbb{R}^d$. For every edge $(i,j)\in E$, the unit-bearing vector from $i$ to $j$ is $b_{ij} = (p_j - p_i) / \|p_j - p_i\|$. The collection of all such bearings defines the bearing map $n$0, and its Jacobian with respect to the configuration $n$1 is the bearing rigidity matrix $n$2.

An infinitesimal motion $n$3 is said to be bearing-preserving if $n$4. The trivial such motions consist of global translations ($n$5, $n$6) and uniform scalings ($n$7, $n$8), both of which maintain all bearings fixed.

The explicit structure of $n$9 is determined via incidence matrices and orthogonal projections. For an oriented edge $m$0, the projection $m$1 is used to ensure that only directions orthogonal to the bearing are encoded. The matrix thus captures the infinitesimal effects of vertex movements on edge bearings (Zhao et al., 2014).

2. Scale-Translation Rigidity: Definition and Characterization

A framework $m$2 is scale-translation rigid (also called bearing rigid) if every other embedding $m$3 with the same set of edge bearings satisfies

$m$4

for some $m$5 (nonzero scale) and $m$6 (translation vector). Thus, coincident inter-neighbor bearings uniquely specify the realization, except for global scaling and translation.

A key equivalence result states that in $m$7, a framework is scale-translation rigid if and only if it is infinitesimally bearing rigid. The latter means the only bearing-preserving infinitesimal motions are translations and scalings, as formalized through the bearing rigidity matrix (Zhao et al., 2014).

Main Rigidity Theorem (Frameworks)

The following are equivalent for a framework $m$8 in $m$9:

• (a) $G(p)$0 is infinitesimally bearing rigid.
• (b) $G(p)$1.
• (c) $G(p)$2. Consequently, the rank condition $G(p)$3 is both necessary and sufficient for scale-translation rigidity (Zhao et al., 2014).

3. Relation to Classical and Distance Rigidity

Scale-translation rigidity (in the bearing sense) contrasts with classical (distance) rigidity, which is usually up to Euclidean transformations (translations and rotations, without scaling). In the plane, the rank condition $G(p)$4 holds, where $G(p)$5 is the distance rigidity matrix, so infinitesimal bearing and distance rigidity coincide in $G(p)$6. However, in higher ambient dimensions, this equivalence does not generally hold.

Distance-based rigidity in $G(p)$7 requires $G(p)$8 for generic configurations, with indeterminacy arising from the $G(p)$9 translational and $$p = [p_1^\top, ..., p_n^\top]^\top \in \mathbb{R}^{dn}$$0 rotational degrees of freedom. For bearing (scale-translation) rigidity, the only ambiguities are the $$p = [p_1^\top, ..., p_n^\top]^\top \in \mathbb{R}^{dn}$$1 translations and one scaling, so full rank is $$p = [p_1^\top, ..., p_n^\top]^\top \in \mathbb{R}^{dn}$$2 (Zhao et al., 2014).

An important advantage is that checking infinitesimal bearing rigidity is dimension-invariant: the criterion is the same in all $$p = [p_1^\top, ..., p_n^\top]^\top \in \mathbb{R}^{dn}$$3, $$p = [p_1^\top, ..., p_n^\top]^\top \in \mathbb{R}^{dn}$$4, and a $$p = [p_1^\top, ..., p_n^\top]^\top \in \mathbb{R}^{dn}$$5-dimensional framework remains bearing rigid when embedded in a higher-dimensional ambient space.

4. Dihedral Rigidity and Extension to Surface Embeddings

For triangulated surface meshes, scale-translation rigidity can also be defined in terms of invariance under the group of similarities (translations, rotations, scaling). Specifically, for a combinatorial triangulation $$p = [p_1^\top, ..., p_n^\top]^\top \in \mathbb{R}^{dn}$$6 of the sphere with $$p = [p_1^\top, ..., p_n^\top]^\top \in \mathbb{R}^{dn}$$7 vertices and $$p = [p_1^\top, ..., p_n^\top]^\top \in \mathbb{R}^{dn}$$8 edges, an embedding $$p = [p_1^\top, ..., p_n^\top]^\top \in \mathbb{R}^{dn}$$9 is determined (modulo similarities) by its vector of edge dihedral angles.

The dihedral rigidity theorem states: for a generic (non-degenerate) embedding, the mapping from mesh embeddings (modulo similarity) to the space of dihedral angle assignments is locally injective, that is, the only infinitesimal motions preserving all dihedrals are infinitesimal similarities. Formally, for almost every point in embedding space (a full-measure Zariski-open set), $p_i \in \mathbb{R}^d$0, and $p_i \in \mathbb{R}^d$1 (Amenta et al., 2018).

The linearized system characterizing infinitesimal dihedral-preserving deformations involves vertex equations, face sum constraints, and cotangent cycle equations. Nontrivial solutions outside the similarity group occur only for degenerate configurations, a set of measure zero (Amenta et al., 2018).

5. Algorithmic Implications and Applications

Scale-translation rigidity underpins several computational methods:

• Formation Control and Network Localization: In multi-agent systems, the ability to stabilize formations or localize nodes using only bearing (directional) information relies on the system's scale-translation rigidity. The bearing-only control laws proposed in (Zhao et al., 2014) exploit this structure.
• Shape Interpolation and Morphing: For triangulated meshes, morphing and interpolation in "dihedral space" is justified because dihedral angles determine the shape up to similarity. The algorithm involves linear interpolation of dihedral vectors, followed by least-squares reconstruction of vertex positions, enforcing centroid and average edge-length normalization to fix translation and scale (Amenta et al., 2018).
• Statistical Shape Analysis: Principal component analysis in dihedral space provides a similarity-invariant basis for statistical studies of shape. This approach demonstrates substantial effectiveness in decomposing complex shape spaces, as shown with human body scans (Amenta et al., 2018).

Domain	Invariant Used	Ambiguity Modulo
Frameworks	Bearings	Translation, Scaling
Meshes	Dihedral Angles	Similarity (Trans., Rot., Scale)

6. Examples and Illustrative Cases

Several canonical examples clarify the rigidity criterion:

• A non-degenerate triangle ($p_i \in \mathbb{R}^d$2) in $p_i \in \mathbb{R}^d$3 is scale-translation rigid, as $p_i \in \mathbb{R}^d$4.
• Three collinear points in $p_i \in \mathbb{R}^d$5 have $p_i \in \mathbb{R}^d$6, so the framework is not rigid.
• In triangulated surfaces, convex polyhedra realize generic (non-degenerate) points where the scale-translation rigidity holds (Zhao et al., 2014, Amenta et al., 2018).

Smooth shape morphs—such as the transition from a dinosaur to a camel—computed by interpolating dihedral angles and reconstructing consistent embeddings, demonstrate the ability of dihedral-based methods to avoid shearing or collapse, producing visually coherent shape sequences (Amenta et al., 2018).

7. Context, Significance, and Limitations

Scale-translation rigidity offers a natural language for applications where distance data are unavailable or unreliable, and only directional or angular information can be measured, as in vision-based formation control, wireless sensor network localization, or computer graphics. Its invariant-based characterization enables efficient algebraic tests and unifies treatment across dimensions.

It should be noted that in higher dimensions ($p_i \in \mathbb{R}^d$7 for frameworks), the equivalence between bearing rigidity and classical (distance) rigidity breaks down, and that degenerate configurations (e.g., collinearity, coplanarity) can lead to loss of rigidity. In mesh embeddings, the Zariski-open genericity caveat means that rare configurations can still fail to be rigid despite sharing all dihedral angles (Zhao et al., 2014, Amenta et al., 2018).