{"id":"778deabf-4f03-40e9-baab-cfd04ffb7163","arxiv_id":"1908.01542","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A framework with signed angle constraints is infinitesimally angle rigid exactly when its rigidity matrix has rank 2N-4, and this condition is used to build an angle-only formation stabilization law.","lead":"This paper develops a theory of angle rigidity for planar point configurations, where only signed angles between rays to neighbors are constrained, and uses it to design formation controllers that need only angle measurements. A smart generalist might read it because it could let robot teams stabilize shapes with cheap vision sensors and without aligned global coordinate systems.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The angle rigidity matrix is defined inconsistently: rows (11)-(13) and (14)-(15) give different first-order conditions, so the rank test in Theorem 4 is not executable as written.","rationale":"The reader's conditional verdict identified the omitted Lemma 3 and the need for f_A to be a smooth submersion. I agree that the sufficiency direction of Theorem 4 needs that machinery and that the proof is delegated rather than supplied. However, the more immediately load-bearing issue is that the paper gives two incompatible definitions of the angle rigidity matrix. Equations (11)-(13) are not scalar multiples of the rows constructed in (14)-(15) unless the two incident edges have equal length; the concrete 3-vertex example above shows different kernels. The rank criterion in Theorem 4 is therefore not well-defined as written. This is a correctness risk in the central claim, not merely a missing detail: a reader implementing the rank test must guess which matrix is intended. Because the matrix in (14)-(15) is the one that matches the standard first-order derivative of the signed angle, the theory may be repairable by declaring (11)-(13) typographical and deleting them, but that repair is not present in the manuscript. The final recommendation remains conditional acceptance: the authors should clarify the definition, correct or remove the inconsistent formulas, and provide the omitted proof of the level-set/manifold step. I do not see grounds to reject the mathematical program outright, since the correct rigidity matrix and rank condition are plausible and the control section is explicit; the concern is about precision and reproducibility of the stated criterion.","tokens_in":20878,"tokens_out":32219,"duration_ms":320861,"concrete_test":"Use the 3-vertex configuration p1 = (2,0), p2 = (0,0), p3 = (0,1) with the single angle constraint (1,2,3) = pi/2. Compute the row according to (11)-(13) and the row according to (14)-(15), then evaluate both on v = (0,2,0,0,-1,0). The (14)-(15) row gives 0 and the (11)-(13) row gives 3/2, so the two proposed rigidity matrices are not equivalent. Then, for the generic two-angle triangular angularity with A = {(1,2,3), (2,3,1)} at a scalene triangle, compute the rank of each matrix: if the rank under (11)-(13) is not 2 = 2N-4, Theorem 4 is false under that definition; if it is 2, the author must state explicitly which matrix defines B(p) and re-derive the proof of Theorem 4 with that definition.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Equations (11)-(13) define a row whose coefficients are A = a_perp/(l_ij l_kj sin beta), B = (-a_perp + b_perp)/(l_ij l_kj sin beta), C = -b_perp/(l_ij l_kj sin beta), where a = pi - pj and b = pk - pj. The matrix display in (14), with N_ij = (pi-pj)_perp / l_ij^2, defines a different row: (a_perp/l_ij^2, -a_perp/l_ij^2 + b_perp/l_kj^2, -b_perp/l_kj^2). These two rows are proportional only when l_ij = l_kj. For the concrete configuration p1 = (2,0), p2 = (0,0), p3 = (0,1) and angle (1,2,3) = pi/2, the tangent vector v = (dot p1 = (0,2), dot p2 = 0, dot p3 = (-1,0)) is annihilated by the row from (14)-(15), which is exactly the first-order condition for angle preservation, but the row from (11)-(13) evaluates to 3/2 on v. Hence the two definitions have different kernels, and the rank test in Theorem 4 cannot be valid for both. The paper does not resolve this ambiguity, and no example or simulation tells the reader which matrix is meant. Since the central claim is literally 'rank of B(p) equals 2N-4', this inconsistency is load-bearing: one reading makes the criterion false, while the other may make it true but is not the definition stated in (11)-(13).","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript develops a theory of signed-angle rigidity for planar frameworks. It introduces an 'angularity' as a vertex set together with ordered triplet angle constraints, defines angle rigidity and global angle rigidity, and shows that angle rigidity does not imply global angle rigidity through a four-vertex example. It then defines an angle rigidity matrix and claims, in Theorem 4, that an angularity is infinitesimally angle rigid if and only if this matrix has rank 2N-4. The paper also gives sufficient conditions for global rigidity via Type-I vertex additions and a necessary combinatorial condition for minimal infinitesimal angle rigidity. The final section applies the theory to formation control: a triangular formation controller and a sequential vertex-addition controller are proposed, with claims of exponential convergence and collision avoidance using only local angle measurements.","tokens_in":21280,"tokens_out":9866,"duration_ms":97179,"significance":"If the main results are correct, the paper offers a useful coordinate-free rigidity notion for angle-only formation control, together with a computable rank test. The example in Fig. 2 correctly illustrates that signed-angle rigidity is a local, not global, property, which is a useful distinction from bearing rigidity. The control law in (77) is simple and naturally expressed in local coordinates. The paper also makes explicit contributions: an infinitesimal rigidity rank condition, vertex-addition constructions, and an exponential-convergence proof for the triangular formation. However, the central definition of the angle rigidity matrix is internally inconsistent as written, and several load-bearing proofs are omitted or circular, so the current manuscript does not establish its claims.","major_comments":[{"comment":"The two displayed definitions of the angle rigidity matrix row are not consistent. Equations (11)-(13) define a row with coefficients A = (pi-pj)_perp/(l_ij l_jk sin beta), B = -(pi-pj)_perp/(l_ij l_jk sin beta) + (pk-pj)_perp/(l_kj l_ij sin beta), and C = (pj-pk)_perp/(l_kj l_ij sin beta), while the matrix display (14)-(15) defines a row with entries N_ij = (pi-pj)_perp/l_ij^2 at vertex i, N_ji + N_kj at vertex j, and N_jk at vertex k. These rows are not proportional except in special cases. For the concrete configuration p1=(2,0), p2=(0,0), p3=(0,1) with angle (1,2,3)=pi/2, the rotation vector v=( (0,2), 0, (-1,0) ) is annihilated by the row from (14)-(15), which is the correct first-order condition for angle preservation, but the row from (11)-(13) evaluates to 3/2 on v. Consequently, under the definition in (11)-(13), Lemma 2 is false because the rotation vector is not in the null space. Since Theorem 4 is literally a rank test on B(p), this inconsistency is load-bearing: the paper must state which row is the angle rigidity matrix and correct the derivation of (10)-(13).","section":"Section III.A, Eqs. (11)-(13) and (14)-(15)"},{"comment":"Lemma 3 is stated without proof, with only a reference to the analogous distance-rigidity result [1, Proposition 5.1]. That analogue requires the measurement map to be a smooth map with constant-rank derivative near p, which is not verified for the signed-angle function f_A and in fact fails at collinear configurations where sin(beta)=0. Theorem 6 depends on Lemma 3, and its sufficiency proof uses Lemma 3 to assert that M and f_A^{-1}(f_A(p)) coincide near p before angle rigidity has been established, making that step circular. The necessity proof also assumes, without proof, that any nearby equivalent configuration can be connected to p by a continuous path along which f_A remains constant. In addition, the converse direction of Theorem 4 has a related gap: a nonzero vector in the null space of B(p) that is not in the span of the four trivial motions gives only a first-order infinitesimal motion, but the paper does not prove that such a vector can be integrated to a curve along which f_A is exactly constant. These gaps should be addressed by proving the needed submersion/constant-rank property of f_A at generic p or by restricting the claims accordingly.","section":"Section III.B, Lemma 3 and Theorem 6"},{"comment":"The induction step in the proof of Theorem 8 is described only as 'similar to the proof from (54) to (74)' and does not verify that the hypotheses--small initial distances, small initial angle errors, and the inequalities l*_{j1 i} > l*_{j2 i}, l*_{j3 i} > l*_{j2 i}--are satisfied for the (k+1)-agent step. The collision-avoidance argument in Lemma 8 extends T to infinity by citing [35, Theorem 2.1], but the hypotheses of that theorem are not checked; inequality (76) only proves positivity at a finite time and does not establish the absence of finite escape or the uniform boundedness needed for extension to all t > 0. The exponential-convergence theorem for the full N-agent formation is therefore not fully supported.","section":"Section IV.B, Theorem 8 and Lemma 8"},{"comment":"The conditions l*_{ij1} > l*_{ij2} and l*_{ij3} > l*_{ij2} are introduced without geometric justification. For a desired formation obtained through Type-I vertex additions, nothing in the construction guarantees that such an ordering holds, and relabeling agents to enforce it may disrupt the triangular sub-formation assumptions. Since these inequalities are used in the determinant computation in Lemma 8, the applicability of the control theorem to the intended class of angle-rigid formations is narrower than claimed and should be discussed explicitly.","section":"Section IV.B, Theorem 8 assumptions"}],"minor_comments":[{"comment":"In Definition 5, the text says 'pj is constrained to be on a ray starting from pj', which appears to be a typo; it should say that pi is constrained to lie on a ray starting from pj.","section":"Section II.A, Definition 5"},{"comment":"The symbols l_ij, l_ji, l_jk, and l_kj are used without a single consistent definition; since l_ij and l_ji denote the same length, the derivation would be easier to follow if all edge lengths were defined once.","section":"Section III.A, Eq. (9)"},{"comment":"Proposition 3 is a formal result, but its proof is merely described as 'can be easily constructed following similar arguments'; this is not a proof and should be replaced with a complete argument or explicitly labeled as a sketch.","section":"Section III, Proposition 3"},{"comment":"The boundary argument in Lemma 6 sets epsilon1 = epsilon2 + epsilon3 without explaining why this exhausts the relevant boundary of the set M_a, and the definitions of alpha_i and alpha*_i near this boundary are not fully specified.","section":"Section IV.A, Lemma 6"},{"comment":"There are several typographical errors, including 'sectoin' in the Section III header and 'minimaly' in Lemma 5; these should be corrected.","section":"Throughout"},{"comment":"The proof of Theorem 1 asserts local uniqueness of the intersection of the ray and the circle; this is plausible but would benefit from a short analytic argument rather than relying only on the figure.","section":"Section II.B, Theorem 1"}],"recommendation":"major_revision","confidential_remarks":"The core idea is promising and likely publishable after a careful revision, but the inconsistency between the rigidity matrix in Eqs. (11)-(13) and the matrix display in Eqs. (14)-(15) is exactly the kind of load-bearing defect that must be fixed before acceptance. The omitted proof of Lemma 3 is also a serious gap, as Theorem 6 and the converse of Theorem 4 rely on it. I found no inappropriate citation patterns or self-citation concerns. The paper fits the journal's scope and, once the technical issues are repaired, would be a useful contribution."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This paper introduces \"angle rigidity\" for planar frameworks with signed angle constraints, and the core idea is genuinely new: the rank condition for infinitesimal angle rigidity and an angle-only formation controller that does not require aligned local frames. The local-vs-global contrast with bearing rigidity is real, and the flex-ambiguity example is instructive. The formalization of an angularity as a set of ordered triples is clean, and the controller in (77) is a sensible bearing-based law that could plausibly work with only local angle measurements.\n\nThe stress-test note is correct, and it matters. Equations (11)-(13) and the matrix display (14)-(15) define different rows. For p1=(2,0), p2=(0,0), p3=(0,1), beta=pi/2, the row from (14)-(15) gives the correct first-order condition for angle preservation, while the row from (11)-(13) evaluates differently on the same velocity. The two rows are proportional only when the adjacent edge lengths are equal. As written, the rank test in Theorem 4 is not executable because the paper does not say which matrix is meant. Since (14)-(15) is the actual Jacobian of the signed-angle function, I believe the intended result is salvageable, but this is a load-bearing ambiguity, not a cosmetic typo.\n\nOther soft spots are real but less severe. Lemma 3 is asserted without proof, and Theorem 6 leans on it; the equivalence between angle rigidity and infinitesimal angle rigidity at generic configurations needs the angle function to be a well-behaved submersion, which the paper does not verify, though generic positions should avoid the collinear singularities. The induction in Theorem 8 is only a sketch; the 4-agent base case is worked out, but the step from k to k+1 is waved through. The collision-avoidance argument extends from a finite interval to infinity informally by citing [35, Theorem 2.1], and the \"sufficiently small\" assumptions are not quantified. There are also no simulations or numerical checks, which would help a reader trust the controller.\n\nWho this is for: people working on vision-based multi-robot formation control and researchers interested in rigidity variants beyond distance and bearing. It deserves a serious referee, but the referee should demand a corrected rigidity matrix derivation, a real proof of Lemma 3, and at least one numerical example. If (11)-(13) are a typo, the authors should say so and remove them; if they are not, the theory as stated is wrong.","headline":"A promising angle-rigidity framework for angle-only formation control, but the paper contains a load-bearing inconsistency between its two definitions of the angle rigidity matrix that must be fixed before the rank test is trustworthy.","tokens_in":21702,"tokens_out":15346,"would_cite":false,"duration_ms":129080,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["52C25","93C85"],"pacs":[],"model":"deepseek-v4-flash","headline":"Angle rigidity gives a rank test for when signed angles determine a planar formation up to translation, rotation, and scaling.","keywords":["angle rigidity","angularity","planar framework","formation control","rigidity matrix","signed angles","multi-agent systems","infinitesimal rigidity"],"falsifier":"Take the collinear three-vertex example the paper itself gives (vertices on a line, single angle constraint equal to $0$, Fig. 3(b)). The angle rigidity matrix formulas contain $\\sin(\\beta)$ in every denominator, so at $\\beta=0$ the matrix is undefined; any limiting construction yields rank $0$, far below $2N-4=2$, yet the paper calls this configuration angle rigid. Computing $B(p)$ explicitly for this case and checking whether a nontrivial angle-preserving flex exists would settle the scope of Theorem 4 and whether a non-collinearity hypothesis is needed.","tokens_in":20715,"feed_emoji":"📐","tokens_out":8699,"duration_ms":89612,"temperature":0.7,"pith_summary":"This paper invents a notion of rigidity for planar point formations whose shape is prescribed by signed angles measured counterclockwise from rays between neighbors—no distances, no relative positions, and no common reference frame. It proves that a formation is infinitesimally angle rigid exactly when a certain angle rigidity matrix $B(p)$ has rank $2N-4$, giving a computational certificate that the angles determine the shape up to translation, rotation, and scaling. Because angle constraints can be quadratic (arcs) as well as linear (rays), angle rigidity turns out to be local, not global: a rigid formation can still have a distant mirror- or alternate realization. The paper uses the rank theory to design a formation-control law that needs only angle measurements in each agent's local coordinate system and is shown locally exponentially stable.","feed_headline":"Rank test decides when angles lock a formation","feed_subtitle":"New angle-rigidity theory lets robot teams stabilize shapes using only local signed angles, no distances.","key_machinery":"The angle rigidity matrix $B(p) \\in \\mathbb{R}^{M \\times 2N}$ is the central object: for each angle constraint $(i,j,k)$ it records the gradient of the signed angle with respect to the three involved vertices, with blocks of the form $(p_i-p_j)^\\perp/l_{ij}^2$ and a factor $1/\\sin(\\beta)$; its null space always contains the four similarity infinitesimal motions, and infinitesimal angle rigidity means the null space is exactly that span. The paper also introduces the combinatorial structure of an angularity—a vertex set $V$, an angle set $\\mathcal{A}$ of ordered triples, and an embedding $p$—and defines Type-I and Type-II vertex addition operations that give sufficient conditions for global and local rigidity, respectively. The rank identity $\\operatorname{rank}(B(p)) = 2N-4$ is the load-bearing algebraic criterion that converts the geometric question of shape determination into a checkable linear-algebra test.","core_discovery":"The central claim is that an angularity $\\mathcal{A}(V,\\mathcal{A},p)$—a planar framework whose constraints are ordered triples of vertices specifying signed angles—is infinitesimally angle rigid if and only if the rank of its angle rigidity matrix $B(p)$ equals $2N-4$. Since the null space of $B(p)$ always contains the four similarity motions (two translations, one rotation, one scaling), reaching this maximal rank means no other infinitesimal deformations preserve the listed angles. The paper further shows that, unlike bearing rigidity, angle rigidity is only a local property: an angle-rigid angularity may admit a non-congruent but equivalent realization far away (flex ambiguity), and this distinction persists even for generic configurations. For generic positions, angle rigidity and infinitesimal angle rigidity are equivalent, so the rank test doubles as a test for finite rigidity of typical configurations. A minimal angle-rigid angularity needs exactly $2N-4$ angles, and any vertex in such a minimal configuration is incident to at least two and at most five angle constraints.","pith_inferences":["The rank test suggests a practical sensor-placement algorithm: add angle constraints incrementally until the angle rigidity matrix of a random realization reaches rank $2N-4$, relying on the genericity of rank to certify all generic positions.","The paper's minimality result is only a necessary condition; proving a Laman-type combinatorial characterization for generic infinitesimal angle rigidity, or finding a counterexample, is a natural next step that the paper explicitly leaves open.","The convergence analysis is sequential (agents 1–3 first, then agent 4 onward) and local; a distributed simultaneous version might inherit exponential convergence for any angle-rigid graph with small initial errors, but the paper does not assert this.","The angle-only controllers use cosines of angles via dot products of local bearings, so a plausible extension is robustness to bounded noise in angle measurements; this is testable but not analyzed in the paper."],"forward_implications":["Any angle-constrained planar framework whose angle rigidity matrix has full rank $2N-4$ is locally shaped-locked up to translation, rotation, and scaling, so the rank test provides a direct algebraic certificate for angle-based shape determination.","For generic configurations, angle rigidity and infinitesimal angle rigidity coincide (Theorem 6), meaning the rank test also decides finite rigidity for typical embeddings.","A minimally angle-rigid planar formation requires exactly $2N-4$ angles, and in such minimal formations every vertex must be incident to at least two and at most five angle constraints—a concrete structural bound on minimal angle-only sensing graphs.","Using the Type-I vertex addition construction, a globally angle-rigid $N$-agent formation can be grown from a generically rigid triangle, and the proposed angle-only control law $u_i = -\\sum_{(j,i,k)\\in\\mathcal{A}}(\\alpha_{jik}-\\alpha^*_{jik})(z_{ij}+z_{ik})$ locally exponentially stabilizes the desired shape in each agent's local coordinate frame.","The control law depends only on local bearing measurements converted to angles, so it removes the usual requirements of GPS, inter-agent distance sensing, or aligned coordinate frames for formation stabilization."],"supporting_citations":[{"why":"The omitted proof of Lemma 3 is stated to be analogous to the distance-rigidity result in this reference, providing the local-manifold argument linking finite and infinitesimal rigidity.","marker":"[1]"},{"why":"The paper follows this reference's approach to derive the angle rigidity matrix from angle constraints by differentiating squared dot-product identities.","marker":"[8]"},{"why":"Provides the contrasting bearing rigidity theory, which is global; the paper uses this contrast to establish that its angle rigidity is only a local property.","marker":"[9]"},{"why":"Prior work on planar angle rigidity that constrains cosines of angles but requires relative displacement measurements; the present paper's contribution is an angle-only control law built on the new rank test.","marker":"[21]"},{"why":"Supplies the definition of algebraic independence used for generic positions, which is central to the generic angle rigidity results.","marker":"[25]"},{"why":"Provides the description of trivial motions (translation, rotation, scaling) that defines infinitesimal angle rigidity and the null-space vectors of the rigidity matrix.","marker":"[26]"},{"why":"Laman's theorem is the classical minimal-counting baseline that the paper contrasts with its own necessary combinatorial condition for minimal angle rigidity.","marker":"[29]"}],"fun_headline_variants":["Angle-only rank test locks planar formations","Local angle rigidity: rank test stabilizes robot teams","Signed angles alone: rank condition for formation lock","Angle-only control: rank test ensures shape lock","Rank test for infinitesimal angle rigidity in formations"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The rank test assumes that near the chosen formation the set of configurations with the same listed angles is a smooth family consisting exactly of the translations, rotations, and scalings; this smoothness breaks down when one of the constrained angles is exactly $0$ or $\\pi$, and the genericity condition is not spelled out for every theorem.","fun_headline_variants_meta":{"raw":{"variants":["Angle-only rank test locks planar formations","Local angle rigidity: rank test stabilizes robot teams","Signed angles alone: rank condition for formation lock","Angle-only control: rank test ensures shape lock","Rank test for infinitesimal angle rigidity in formations"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000632,"raw_usage":{"total_tokens":2950,"prompt_tokens":1008,"completion_tokens":1942,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":624,"completion_tokens_details":{"reasoning_tokens":1872}},"tokens_in":624,"tokens_out":1942,"duration_ms":14475,"temperature":1.0,"reasoning_tokens":1872,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:10:05.616449+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the collinear three-vertex example the paper itself gives (vertices on a line, single angle constraint equal to $0$, Fig. 3(b)). The angle rigidity matrix formulas contain $\\sin(\\beta)$ in every denominator, so at $\\beta=0$ the matrix is undefined; any limiting construction yields rank $0$, far below $2N-4=2$, yet the paper calls this configuration angle rigid. Computing $B(p)$ explicitly for this case and checking whether a nontrivial angle-preserving flex exists would settle the scope of Theorem 4 and whether a non-collinearity hypothesis is needed.","supporting_citations":[{"cited_title":"Rigid and flexible frameworks,","cited_arxiv_id":null,"evidence_quote":"The omitted proof of Lemma 3 is stated to be analogous to the distance-rigidity result in this reference, providing the local-manifold argument linking finite and infinitesimal rigidity."},{"cited_title":"Sensor and network topologies of formations with direction, bearing, and angle information between agents,","cited_arxiv_id":null,"evidence_quote":"The paper follows this reference's approach to derive the angle rigidity matrix from angle constraints by differentiating squared dot-product identities."},{"cited_title":"Generic global rigidity,","cited_arxiv_id":null,"evidence_quote":"Supplies the definition of algebraic independence used for generic positions, which is central to the generic angle rigidity results."},{"cited_title":"Frameworks, tensegrities and symmetry: understanding stable structures,","cited_arxiv_id":null,"evidence_quote":"Provides the description of trivial motions (translation, rotation, scaling) that defines infinitesimal angle rigidity and the null-space vectors of the rigidity matrix."},{"cited_title":"On graphs and rigidity of plane skeletal structures,","cited_arxiv_id":null,"evidence_quote":"Laman's theorem is the classical minimal-counting baseline that the paper contrasts with its own necessary combinatorial condition for minimal angle rigidity."}],"review_version":1}