Structure and classification results for the $\infty$-elastica problem

Roger Moser

REVIEW 1 major objections 5 minor 30 references

Structure and classification results for the $\infty$-elastica problem

T0 review · 1 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Curves minimizing the $L^\infty$-norm of curvature are exactly the solutions of a differential system, and every solution is either a finite planar arc-and-line chain around a line or a three-dimensional helical-type curve.

desk verdict First complete treatment of the L∞-elastica problem; the characterization and classification hold up, with one fixable technical gap in Lemma 19. read the letter →

arxiv 1908.01569 v1 pith:73HOH6SB submitted 2019-08-05 math.DG

Roger Moser This is my paper · ORCID

classification math.DG MSC 53A04

keywords infinity-elasticamaximumcurvatureminimizationL-infinityvariationalproblemscurveclassificationMarkov-Dubinsproblemboundedvariationweightdifferentialequationsforcurves

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper asks how much a curve of fixed length must bend to connect two points with prescribed tangent directions, and solves the version in which the cost is the largest curvature (in a weighted $L^\infty$ norm) rather than an integral. Because that functional is not differentiable, the paper introduces a weakened notion, the $\infty$-elastica, and proves that it is characterized by a system of differential equations involving a unit vector and a nonnegative multiplier function. It then classifies all solutions: apart from the degenerate straight segment, every $\infty$-elastica is either a finite chain of planar circular arcs and line segments wound around a single line, or a curve contained in a three-dimensional affine subspace that solves the system with a strictly positive multiplier. The classification matters because it converts a nondifferentiable variational problem into an explicit description of all possible optimal shapes, with direct analogues for shortest-path problems with curvature bounds.

What carries the argument

An $L^p$ approximation with a deliberately added penalization term carries the argument. For finite $p$, minimizers of $K_p(\tau)+\frac{\mu}{2L}\int_0^L\beta|\tau-\tau_0|^2\,dt$ satisfy an explicit Euler-Lagrange equation on the sphere, and the penalization forces the approximating tangents to converge back to a chosen pseudo-minimizer $\tau_0$ as $p\to\infty$, despite nonuniqueness. Dividing the renormalized Euler-Lagrange equation by $1+|\Lambda_p|$ and passing to the limit yields the system $u'+ (u\cdot\tau')\tau=\beta(\lambda-(\lambda\cdot\tau)\tau)$ and $|u|\tau'=ku$, which reparametrization and projection convert into the system of Theorem 2. The structural dichotomy comes from the zero set of $f=k|u|$: wherever $f>0$ the ODE is nondegenerate and produces three-dimensional solutions; wherever $f=0$ the curve must run along circular arcs on great circles through $\lambda$, collapsing onto a line; Lemma 19 uses bounded variation of $\alpha$ to force only finitely many such intervals.

What would settle it

Take a bounded but non-$(BV)$ weight $\alpha$ and boundary data for which the multiplier $g$ from the weak system would vanish at a sequence of points accumulating inside the interval; numerically solve the system and check whether a curve with infinitely many planar arcs of curvature magnitude $k$ satisfies the weak equations. If such a curve exists, Theorem 4's 'finitely many intervals' is false without bounded variation.

Watch

Extended reading notes

Core claim

The central claim, Theorem 2, is that a curve $\gamma$ with unit tangent $T=\gamma'$ and curvature norm $k=K_\alpha(\gamma)$ is an $\infty$-elastica if and only if there exist a unit vector $\lambda$ and a nonzero, nonnegative function $g$ such that $g((\alpha T')' + k^2T/\alpha)=k^2\,\mathrm{proj}^{\perp}_{T,T'}(\lambda)$ and $g'=\alpha\,\lambda\cdot T'$ hold weakly. Theorem 4 then separates the solutions into two regimes. In the first, the curve leaves a single line $L\parallel \lambda$ only on finitely many relatively open intervals; on each such interval it is planar, its weighted curvature $\alpha\gamma''$ is continuous with magnitude exactly $k$, and the sign of $\lambda\cdot\gamma''$ flips at the ends of the interval. In the second, the curve lies in a three-dimensional affine subspace, $\alpha\gamma''$ is $W^{1,\infty}$ with $|\alpha\gamma''|\equiv k$, and the system holds with $g>0$ almost everywhere. The same system yields a sufficient condition for genuine minimizers, and examples show that $\infty$-elasticas can fail to be minimizers.

Load-bearing premise

The proof that the planar pieces are finite relies on the weight $\alpha$ having bounded variation with $1/\alpha$ bounded; if $\alpha$ is merely bounded, nothing in the argument stops the zero set of the multiplier from accumulating and the curve from having infinitely many alternating pieces.

Editorial extensions

If this is right

Every $R$-geodesic in the bounded-curvature shortest-path problem minimizes the unweighted functional $K_1$, so Theorem 4 reproves and sharpens the classical planar and three-dimensional classifications of that problem.
With unit weight, planar $\infty$-elasticas are exactly two shapes: a circular arc, then line segments and equal-radius full circles, then a circular arc, or several equal-radius arcs of equal length with alternating orientation.
A circular arc of radius $r$ is a minimizer of $K_1$ whenever its length is at most $2\pi r/3$ under the sufficient condition; the older chord result extends this bound to $2\pi r$, showing the sufficient condition is not necessary.
There are $\infty$-elasticas that are not minimizers, including a one-parameter family of three-arc curves that are not even local minimizers in $W^{1,2}$.
The theory covers weights $\alpha$ of bounded variation with bounded reciprocal, so the same classification applies to weighted maximum-curvature problems after reparametrization by the weight.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

If one drops bounded variation and only keeps $\alpha$ bounded, the proof's finiteness step fails; a natural extension would be to build a weight whose zero set accumulates and to check whether infinitely many planar pieces satisfy the weak system, which would mark the exact boundary of the classification.
The same penalization trick, adding a distance term to select one solution in the limit, could transfer to other nonunique $L^\infty$ variational problems, where classical $L^p$ approximations only recover one of many minimizers.
Because every $R$-geodesic minimizes $K_\alpha$, any uniqueness or regularity result proved for $K_\alpha$ minimizers automatically constrains shortest-path solutions, so the differential-equation description may offer a new route to stability questions in motion planning.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

You should know two things about this paper. First, it really is the first study of the L∞-elastica problem, and the central results are good: Theorem 2 gives an if-and-only-if ODE characterization, and Theorem 4 gives a two-case classification (planar concatenations around a line, or a genuinely 3D curve solving a nondegenerate ODE system). Second, the proof strategy is the interesting part: approximate by L^p functionals with an added penalization term that selects the target pseudo-minimiser, then pass to the limit. The penalization is more than a technical trick; it is what makes the characterization crisp.

The paper does a lot well. The pseudo-minimiser definition is a sensible replacement for a non-existent Euler-Lagrange equation. The passage from the L^p Euler-Lagrange equation to the limit system (10)–(11), then to (2)–(3), is carefully done. The regularity analysis in Lemma 14 is solid. I found no circularity: the system is derived from the variational inequality, not assumed, and the Markov-Dubins comparison provides an external benchmark. The examples are useful, especially Example 22, which shows the ∞-elastica condition is genuinely weaker than being a minimiser.

The soft spots are minor. The bounded-variation assumption on α, stated at the start, is used exactly once, in Lemma 19, to control the sum |1 − b'_i/b_i| and force the zero set of g to be finite. Without it, Theorem 4(i) could allow infinitely many planar pieces; that is a scoping statement, not a hidden gap. There is a small technical gap inside Lemma 19: since β is only BV, the weighted mean value theorem used to pick ω_i, ω_i' is not automatic. The same bound follows by comparing weighted averages of β on the two subintervals and using the total variation, so this is fixable. Section 7's alternative proof of Dubins's and Sussmann's results is a sketch and depends on some of Dubins's lemmas; it is presented as such and does not affect the main theorem.

This paper deserves a serious referee. I would accept it after minor revision, with the referee asked to check Lemma 19 carefully and perhaps request a fuller treatment of the Dubins recovery. It should be sent to someone with taste in L∞ variational problems or geometric control. I would cite it as the reference for the L∞-elastica and use the penalized L^p approximation as a model for similar problems.

Referee Report

1 major / 5 minor

Summary. The paper studies the L-infinity elastica problem: among arc-length parametrized curves of fixed length with prescribed endpoints and endpoint tangents, minimize the essential supremum of alpha |gamma''|. After reparametrization by the weight alpha, the problem is reformulated in terms of tangent vector fields tau, minimizing ||tau'||_infinity under endpoint and integral constraints. The main results are Theorem 2, which characterizes infinity-elasticas (defined through a quadratic penalization inequality) by the ODE system (2)-(3); Theorem 3, a sufficient condition for genuine minimizers; and Theorem 4, a classification into finitely many planar circular-arc/line-segment pieces around a single line, or a three-dimensional curve solving the ODE system with a positive multiplier. The proof strategy is L^p approximation with a penalization term, passage p -> infinity, and then detailed analysis of the resulting ODE system. Section 7 connects the results to the Markov-Dubins problem and sketches an alternative proof of Dubins's and Sussmann's theorems.

Significance. If correct, these results are a substantial and novel contribution. The paper provides a usable Euler-Lagrange-type characterization for a non-differentiable L-infinity geometric variational problem and a complete structural classification of its solutions, going substantially beyond existing work on second-order L-infinity variational problems. The L^p approximation with penalization is well designed and the ODE analysis is detailed. The connection to Dubins's R-geodesics provides a valuable independent check on the classification. The main proofs are presented in considerable detail, and the treatment is largely self-contained. The classification theorem is significant enough to merit publication once the one load-bearing proof gap identified below is repaired.

major comments (1)

[Section 6, Lemma 19] The paragraph beginning 'If b'_i <= b_i, then we may choose omega_i ...' applies a weighted mean value theorem to conclude that the integral of beta against a nonnegative weight is comparable to the weight evaluated at a point. This step requires continuity (or at least a Darboux-type property) of beta, whereas the standing assumption only gives beta of bounded variation, which permits jumps. As written, the inequalities displayed there are not justified. This is load-bearing: the estimate on the sum of |1 - b'_i/b_i| is exactly what forces I\Omega to be finite, and hence what yields the finite-decomposition conclusion in Theorem 4(i). The gap is repairable: replace the pointwise choices of omega_i and omega'_i by weighted averages of beta over the two subintervals; the identity between the two integrals gives b'_i/b_i as a ratio of two weighted averages of beta, and the difference of those averages is bounded by Var(beta, [t_i, t_{i+1}]) / inf beta. I recommend that the authors rewrite this paragraph accordingly.

minor comments (5)

[Section 2, proof of Proposition 7(2)] The proof assumes an inequality with 'm > 0', while Definition 6 permits any real m; the argument goes through after replacing m by max{m,0}, but this reduction should be stated explicitly.
[Section 3, Proposition 11] In the display following (18), 'beta|tau~ - tau'|^2' should read 'beta|tau~ - tau|^2', since sigma = tau~ - tau; the same typo appears two lines below.
[Section 6, Lemma 17] The sentence 'It follows that Omega'\Omega is a null set, and so is Omega'\Omega' contains a duplication; the second clause should state that Omega'\Omega is open relative to [0,L], which is why it must be empty.
[Section 7] The alternative proof of Dubins's theorem is only sketched and explicitly invokes Dubins's Lemma 2; the authors should list precisely which facts from [8] are used, so that the claimed alternative proof is checkable.
[Section 2, Proposition 7(1)] The Euler-Lagrange derivation is omitted with a reference to standard harmonic-map-style constrained variation computations; a short derivation or a precise citation would improve self-containedness, though this is not a substantive issue.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the derivation is self-contained from the pseudo-minimiser definition through explicit L^p approximation and ODE analysis.

full rationale

The paper's central derivation is self-contained. Definition 1 introduces the ∞-elastica/pseudo-minimiser as an inequality with a quadratic penalty; this is the starting point, not an assumed conclusion. Section 2 reparametrises the problem and defines the L^p approximations J^µ_p with a penalisation term, and Proposition 7 proves existence and convergence using the direct method and the pseudo-minimiser inequality. The Euler-Lagrange equation (7) is derived from J^µ_p; Lemma 8 bounds the Lagrange multipliers; Proposition 9 passes p → ∞ to obtain the first-order system (10)-(11). Propositions 11 and 12 prove the converse and show equivalence with the second-order ODE system (19)-(20), giving Theorem 2. The classification in Theorem 4 follows from the ODE analysis in Lemmas 14-20, not from any prior result assumed as an input. The extra hypothesis that α has bounded variation and 1/α is bounded is an explicit standing assumption and is used transparently in Lemma 19; removing it would weaken the finite-interval conclusion of Theorem 4(i), but that is a scoping statement, not a circular step. The self-citations [13, 23, 26] are methodological comparisons and are not load-bearing for the main theorems. The Markov-Dubins connection in Proposition 5 is proved in the paper and is checked externally against Dubins's and Sussmann's results; it is not used to derive Theorem 2 or Theorem 4. No fitted parameter is renamed as a prediction, and no load-bearing uniqueness theorem is imported from the author's prior work.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

No free parameters are fitted; the constants (length, curvature bound k) are either prescribed or derived from the functional. The only invented objects are the auxiliary function g and vector lambda, which are Lagrange multipliers arising from the variational analysis, not independent postulates. The paper relies on standard background theory and a feasibility assumption.

assumptions (3)

domain assumption The admissible set G of curves satisfying (1) is nonempty for the given boundary data.
The paper fixes endpoints, tangents, and length without proving existence of a feasible curve; the entire variational analysis presupposes that the constraint set is not empty.
standard math Standard existence, regularity, and ODE theory used throughout (direct method, Sobolev embedding, Picard-Lindelof) is valid.
Invoked in Proposition 7, Lemma 8, Lemma 14, and elsewhere; these are background results not proved in the paper.
standard math The Euler-Lagrange equation (7) for minimizers on the sphere S^{n-1} holds as stated.
Proposition 7(1) is asserted without proof, citing standard harmonic-map techniques (Simon); this is a standard computation but still an unproved background step.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Structure and classification results for the $\infty$-elastica problem." pith.science (2026). https://pith.science/paper/73HOH6SB

@misc{pith2026190801569,
  author       = {Pith},
  title        = {Pith review of: Structure and classification results for the $\infty$-elastica problem},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/73HOH6SB}},
  note         = {Machine review of arXiv:1908.01569}
}

abstract

Consider the following variational problem: among all curves in $\mathbb{R}^n$ of fixed length with prescribed end points and prescribed tangents at the end points, minimise the $L^\infty$-norm of the curvature. We show that the solutions of this problem, and of a generalised version, are characterised by a system of differential equations. Furthermore, we have a lot of information about the structure of solutions, which allows a classification.

Figures

Figures reproduced from arXiv: 1908.01569 by the authors.

**Figure 2.** Construction of an ∞-elastica that is not a minimiser For r ∈ [1/3, 1), we also construct some comparison curves including three circular arcs of radius r. To this end, define ω(r) = arccos((1 − r)/2r). For h ∈ R, there is a curve comprising three circular arcs of radius r, with centres (r − 1, h), (0, h + 2r sin ω(r)), (1 − r, h), that connects the points (−1, h) and (1, h). The length of this curve is ˜`(r) = r(3π… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

30 extracted references · 29 canonical work pages

[1]

Aronsson, Minimization problems for the functional supxF (x, f(x), f′(x)), Ark

G. Aronsson, Minimization problems for the functional supxF (x, f(x), f′(x)), Ark. Mat. 6 (1965), 33–53

work page 1965
[2]

, Minimization problems for the functional supxF (x,f (x),f′(x)). II, Ark. Mat. 6 (1966), 409–431

work page 1966
[3]

, Extension of functions satisfying Lipschitz conditions , Ark. Mat. 6 (1967), 551–561

work page 1967
[4]

, On certain minimax problems and Pontryagin’s maximum princi- ple, Calc. Var. Partial Diﬀerential Equations 37 (2010), no. 1-2, 99–109

work page 2010
[5]

Blaschke, Vorlesungen ¨ uber Diﬀerentialgeometrie und geometrische Grundlagen von Einsteins Relativit¨ atstheorie

W. Blaschke, Vorlesungen ¨ uber Diﬀerentialgeometrie und geometrische Grundlagen von Einsteins Relativit¨ atstheorie. Band I. Elementare Diﬀer- entialgeometrie, Dover Publications, New York, N. Y., 1945, 3d ed

work page 1945
[6]

Bryant and P

R. Bryant and P. Griﬃths, Reduction for constrained variational problems and ´ 1 2k2ds, Amer. J. Math. 108 (1986), no. 3, 525–570

work page 1986
[7]

S. S. Chern, Curves and surfaces in Euclidean space , Studies in Global Geometry and Analysis, Math. Assoc. Amer. (distributed by Prentice-Hall, Englewood Cliﬀs, N.J.), 1967, pp. 16–56

work page 1967
[8]

L. E. Dubins, On curves of minimal length with a constraint on average curvature, and with prescribed initial and terminal positions and tangents , Amer. J. Math. 79 (1957), 497–516. 26

work page 1957

Show all 30 references

[9]

L. Euler, Methodus inveniendi lineas curvas maximi minimive proprietate gaudentes sive solutio problematis isoperimetrici latissimo sensu accepti , Marc-MichelBousquet & Co., Lausanne–Geneva, 1744
[10]

Ferone, B

V. Ferone, B. Kawohl, and C. Nitsch, Generalized elastica problems under area constraint, Math. Res. Lett. 25 (2018), no. 2, 521–533

2018
[11]

Huang, A note on the p-elastica in a constant sectional curvature man- ifold, J

R. Huang, A note on the p-elastica in a constant sectional curvature man- ifold, J. Geom. Phys. 49 (2004), no. 3-4, 343–349

2004
[12]

Katzourakis, An introduction to viscosity solutions for fully nonlinear PDE with applications to calculus of variations in L∞, SpringerBriefs in Mathematics, Springer, Cham, 2015

N. Katzourakis, An introduction to viscosity solutions for fully nonlinear PDE with applications to calculus of variations in L∞, SpringerBriefs in Mathematics, Springer, Cham, 2015

2015
[13]

Katzourakis and R

N. Katzourakis and R. Moser, Existence, uniqueness and structure of second order absolute minimisers , Arch. Ration. Mech. Anal. 231 (2019), no. 3, 1615–1634

2019
[14]

Katzourakis and E

N. Katzourakis and E. Parini, The eigenvalue problem for the ∞- Bilaplacian, NoDEA Nonlinear Diﬀerential Equations Appl. 24 (2017), no. 6, Art. 68, 25

2017
[15]

Katzourakis and T

N. Katzourakis and T. Pryer, Second-order L∞ variational problems and the ∞-polylaplacian, Adv. Calc. Var. (Ahead of Print) (2018), DOI:10.1515/acv-2016-0052

2018 doi
[16]

Methods Partial Diﬀerential Equations 35 (2019), no

, On the numerical approximation of p-biharmonic and ∞- biharmonic functions , Numer. Methods Partial Diﬀerential Equations 35 (2019), no. 1, 155–180

2019
[17]

Langer and D

J. Langer and D. A. Singer, Knotted elastic curves in R3, J. London Math. Soc. (2) 30 (1984), no. 3, 512–520

1984
[18]

Diﬀerential Geom

, The total squared curvature of closed curves , J. Diﬀerential Geom. 20 (1984), no. 1, 1–22

1984
[19]

Laumond, S

J. Laumond, S. Sekhavat, and F. Lamiraux, Guidelines in nonholonomic motion planning for mobile robots , Robot Motion Planning and Control, Lecture Notes in Control and Information Sciences, vol. 229, Springer- Verlag, Berlin–Heidelberg, 1998, pp. 1–53

1998
[20]

Linn´ er, Explicit elastic curves , Ann

A. Linn´ er, Explicit elastic curves , Ann. Global Anal. Geom. 16 (1998), no. 5, 445–475

1998
[21]

A. A. Markov, Some examples of the solution of a special kind of problem on greatest and least quantities, Soobshch. Kharkovsk. Mat. Obshch. 1 (1887), 250–276, in Russian
[22]

Masnou and J.-M

S. Masnou and J.-M. Morel, Level lines based disocclusion , Proceedings 1998 IEEE International Conference on Image Processing, 1998, pp. 259– 263

1998
[23]

Moser and H

R. Moser and H. Schwetlick, Minimizers of a weighted maximum of the Gauss curvature, Ann. Global Anal. Geom. 41 (2012), no. 2, 199–207. 27

2012
[24]

A Oldfather, C

W. A Oldfather, C. A. Ellis, and D. M. Brown, Leonhard Euler’s elastic curves, Isis 20 (1933), 72–160

1933
[25]

J. A. Reeds and L. A. Shepp, Optimal paths for a car that goes both forwards and backwards, Paciﬁc J. Math. 145 (1990), no. 2, 367–393

1990
[26]

Z. N. Sakellaris, Minimization of scalar curvature in conformal geometry , Ann. Global Anal. Geom. 51 (2017), no. 1, 73–89

2017
[27]

Schmidt, ¨Uber das Extremum der Bogenl¨ ange einer Raumkurve bei vergeschriebenen Einschr¨ ankungen ihrer Kr¨ ummung, Sitzungsberichte preuss

E. Schmidt, ¨Uber das Extremum der Bogenl¨ ange einer Raumkurve bei vergeschriebenen Einschr¨ ankungen ihrer Kr¨ ummung, Sitzungsberichte preuss. Akad. Wissenschaften Berlin 1925 (1925), 485–490

1925
[28]

Schur, ¨Uber die Schwarzsche Extremaleigenschaft des Kreises unter den Kurven konstanter Kr¨ ummung, Math

A. Schur, ¨Uber die Schwarzsche Extremaleigenschaft des Kreises unter den Kurven konstanter Kr¨ ummung, Math. Ann. 83 (1921), 143–148

1921
[29]

Simon, Theorems on regularity and singularity of energy minimizing maps, Lectures in Math

L. Simon, Theorems on regularity and singularity of energy minimizing maps, Lectures in Math. ETH Z¨ urich, Birkh¨ auser, Basel, 1996

1996
[30]

H. J. Sussmann, Shortest 3-dimensional paths with a prescribed curvature bound, Proceedings of 1995 34th IEEE Conference on Decision and Control, vol. 4, 1995, pp. 3306–3312. 28

1995

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [1]

Aronsson, Minimization problems for the functional supxF (x, f(x), f′(x)), Ark

G. Aronsson, Minimization problems for the functional supxF (x, f(x), f′(x)), Ark. Mat. 6 (1965), 33–53

work page 1965

[2] [2]

, Minimization problems for the functional supxF (x,f (x),f′(x)). II, Ark. Mat. 6 (1966), 409–431

work page 1966

[3] [3]

, Extension of functions satisfying Lipschitz conditions , Ark. Mat. 6 (1967), 551–561

work page 1967

[4] [4]

, On certain minimax problems and Pontryagin’s maximum princi- ple, Calc. Var. Partial Diﬀerential Equations 37 (2010), no. 1-2, 99–109

work page 2010

[5] [5]

Blaschke, Vorlesungen ¨ uber Diﬀerentialgeometrie und geometrische Grundlagen von Einsteins Relativit¨ atstheorie

W. Blaschke, Vorlesungen ¨ uber Diﬀerentialgeometrie und geometrische Grundlagen von Einsteins Relativit¨ atstheorie. Band I. Elementare Diﬀer- entialgeometrie, Dover Publications, New York, N. Y., 1945, 3d ed

work page 1945

[6] [6]

Bryant and P

R. Bryant and P. Griﬃths, Reduction for constrained variational problems and ´ 1 2k2ds, Amer. J. Math. 108 (1986), no. 3, 525–570

work page 1986

[7] [7]

S. S. Chern, Curves and surfaces in Euclidean space , Studies in Global Geometry and Analysis, Math. Assoc. Amer. (distributed by Prentice-Hall, Englewood Cliﬀs, N.J.), 1967, pp. 16–56

work page 1967

[8] [8]

L. E. Dubins, On curves of minimal length with a constraint on average curvature, and with prescribed initial and terminal positions and tangents , Amer. J. Math. 79 (1957), 497–516. 26

work page 1957