REVIEW 2 major objections 4 minor
A Study on Effective Initial Guess Finding Method Based on B\'ezier Curves: Orbit Determination Applications
T0 review · 2 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash
Pith's one-line read This paper proposes fitting a quadratic Bézier curve to a two-point boundary value problem and using its initial slope as the shooting-method starting guess, reporting faster convergence and a successful highly elliptical orbit solve…
desk verdict Likely factor-of-2 error in Eq. (16) undermines the numerical evidence, but the core collocation idea is sound and worth a serious referee. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The central object is the quadratic non-rational Bézier curve, defined through Bernstein polynomial basis functions with control points that shape the curve. The paper represents time $t(s)$ and position $\mathbf{r}(s)$ as such curves, keeps the endpoint control points fixed by the boundary conditions, and treats the intermediate control points $t_1$ and $\mathbf{r}_1$ as unknowns. The mechanism is the residual integral $L(t_1,\mathbf{r}_1)=\int_0^1 g_s^\top g_s\,ds$, where $g_s$ is the two-body equation written in Bézier form using the chain-rule derivative identities; minimizing $L$ selects control points whose curve approximately satisfies the dynamics. The initial velocity guess is then the derivative $\mathbf{r}'(s)/t'(s)$ evaluated at $s=0$.
What would settle it
Re-run the orbit cases with the second derivative taken as $\mathbf{r}''(s)=2(\mathbf{r}_0-2\mathbf{r}_1+\mathbf{r}_2)$ in Eq. (16) and check whether the corrected Bézier initial guess for case 3-3 still lets the shooting method converge in the reported seven iterations; if the general shooting method then matches or beats it, the central claim fails.
Extended reading notes
Core claim
On the paper's own terms, the central discovery is that the control points of a quadratic non-rational Bézier curve, selected to minimize the residual integral $L(t_1,\mathbf{r}_1)=\int_0^1 g_s^\top g_s\,ds$ of the transformed two-body equation, yield an initial velocity close enough to the true initial condition that the shooting method converges quickly. For each simulation case the paper computes the intermediate control points $t_1$ and $\mathbf{r}_1$ by minimizing $L$, substitutes them into the Bézier expressions for $t(s)$ and $\mathbf{r}(s)$, evaluates the chain-rule derivative $\mathbf{r}'(s)/t'(s)$ at $s=0$, and uses that vector as the initial guess. The reported comparisons show fewer shooting iterations and shorter total time in all tested cases, with improvements in total time between roughly 9% and 44% for the orbit cases, and convergence for case 3-3, the Molniya near-perigee case, where the general shooting method with a cross-product-based initial guess does not converge.
Load-bearing premise
The numerical results rest on the control-point optimization being set up correctly, which requires the printed second-derivative identity for the quadratic Bézier curve to be exact; if that identity is wrong, the optimized initial guesses and the reported convergence gains change.
Editorial extensions
If this is right
- Shooting-based solvers for orbit determination can be given a principled starting velocity instead of an arbitrary one, reducing the interval-finding and root-finding iterations in the shooting loop.
- For highly elliptical orbits with short time of flight near perigee, the method may find solutions that a general shooting method with a geometric initial guess cannot reach.
- The same residual-minimization idea transfers to other two-point boundary value problems, since the one-dimensional test problem also shows a large reduction in total computation time, about 72.6%.
- The fitted Bézier curve itself gives a smooth approximate trajectory between the boundary positions, so it can be compared against the integrated orbit as a sanity check before shooting.
Reading between the lines
- The residual integral could be evaluated on a collocation grid rather than exactly, which would make the control-point optimization cheaper and easier to differentiate for higher-order Bézier curves.
- The method could be combined with multiple shooting: the Bézier fit could provide segment-wise initial states instead of a single initial velocity, potentially stabilizing long-arc or high-eccentricity transfers.
- Because the paper's derivative identity for $\mathbf{r}''(s)$ is printed inconsistently with the one-dimensional case, the numerical tables should be re-run with the correct factor; this is our check, not a claim the paper makes.
- For Lambert-type problems, the fitted Bézier arc could be used to initialize iterative Lambert solvers, not just shooting methods.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper proposes an initial-guess generation method for two-point boundary value problems (TPBVPs) in orbit determination. The method represents the time and position along the trajectory as quadratic non-rational Bézier curves, fixes the endpoint control points to the given boundary conditions, and chooses the intermediate control points t1 and r1 by minimizing the integral of the squared residual of the transformed two-body equation. The initial velocity is then computed from the Bézier curve at s=0 and used to seed a shooting method. The authors validate the approach on a scalar nonlinear BVP and on circular, tundra, and Molniya orbit cases, reporting fewer iterations, lower computational time, and success in a highly elliptical case where the comparison shooting method fails.
Significance. If the method works as reported, it would provide a simple, systematic way to initialize shooting methods for orbit determination, with particular value for highly eccentric orbits where naive initialization fails. The idea of using Bézier control points as optimization variables is clear, and the formulation of the residual in transformed time is a natural collocation-style approach. The paper also includes numerical results for several orbit geometries, which is a useful first demonstration. However, the central numerical evidence currently rests on a derivative formula that is internally inconsistent with the paper's own scalar derivation, so the reported improvements cannot be taken at face value until the formula is corrected and the experiments are rerun or explicitly confirmed.
major comments (2)
- [§3.2, Eq. (16)] The printed second-derivative formula for the quadratic Bézier position curve is wrong by a factor of two. For r(s) = (1-s)^2 r0 + 2s(1-s) r1 + s^2 r2, direct differentiation gives r''(s) = 2(r0 - 2r1 + r2), not (r0 - 2r1 + r2). This is inconsistent with Eq. (10), where the identical quadratic form is correctly differentiated as x''(s) = 2(x0 - 2x1 + x2). Since Eq. (17) uses r'' in the residual g_s, the objective L in Eq. (18) is not the residual of the two-body problem if the printed formula was used. The control points reported in Table 3 and the comparisons in Table 4 therefore do not currently establish the claimed method. The authors must correct Eq. (16), rerun the optimization with the correct derivative, and report whether the numerical results change, or explicitly state that the implementation used the correct factor despite the printed formula.
- [§4.2, Table 4] The comparison baseline is a 'general shooting method' initialized with the cross product of the orbit normal unit vector and the initial position unit vector. This is a deliberately rough guess and does not represent typical practical initialization strategies for orbit determination, such as a Lambert solver or an analytic conic estimate. The reported improvements in computational time and the failure of the general shooting method in case 3-3 may therefore be attributable to the weakness of this baseline rather than to a general advantage of the proposed method. To support the practical claim, the authors should compare against a standard initialization method and, ideally, report sensitivity to the baseline choice.
minor comments (4)
- [§4.2, Eq. (18)] The paper states that control points are obtained by solving a system of nonlinear equations, but the actual optimization procedure is not described. Details such as the solver used, tolerances, initial guesses for the control points, and whether a global or local minimum is sought should be provided.
- [Table 3] The column 'Error (%)' is not defined precisely. It appears to be a per-component relative error, but for case 3-3 the y and z components show 90.21% error while x shows 15.28%, and a single norm-based error metric would be more informative. Please clarify the definition.
- [§3.1, Eq. (11)] The transition from Eq. (6) to the scalar example is clear, but the notation f_s in Eq. (6) is never explicitly defined for the scalar case; it would help to spell out the transformed right-hand side for Eq. (11).
- [§4.2, Figs. 2–4] The figures compare the Bézier approximation with numerical integration using the shooting solution, but the reported initial velocity errors in Table 3 are quite large in some components. A short discussion of how the approximation error translates into shooting convergence would improve the presentation.
Circularity Check
No significant circularity: the Bézier initial-guess method is a standard residual-minimization/collocation approach, and its claims are not forced by the inputs or by self-citation.
full rationale
The paper's core derivation is not circular. It defines the residual g_s in Eq. (17) from the two-body equation, parameterizes the trajectory and time as quadratic Bézier curves, and selects the unknown control points t_1 and r_1 by minimizing L in Eq. (18). The resulting initial velocity is then read off from the derivative of the fitted curve at s = 0. This is a legitimate weighted-residual/collocation strategy: the minimization uses only the governing ODE, the boundary conditions, and the chosen polynomial basis, and it does not use the true initial velocity as an input. The later shooting step solves the same ODE with that independently generated initial guess, so the comparison in Table 4 is a genuine test of whether the guess lies in the convergence basin. The citations to Kim and Mortari [7,8] provide standard derivative identities and prior Bézier applications, but the central formulas in Eqs. (5), (9), and (10) are elementary calculus and are not load-bearing self-citations; no uniqueness theorem or fitted parameter is imported from those papers. The factor-of-2 discrepancy between the second derivative in Eq. (10), x''(s) = 2(x0 - 2x1 + x2), and the printed r''(s) = (r0 - 2r1 + r2) in Eq. (16) is a potentially serious implementation/correctness issue that may invalidate the numerical results if the printed formula was used, but it is not a circularity: an incorrect residual functional is still an independent, falsifiable computation rather than an equivalence between input and output. Therefore no circular step is exhibited.
Assumptions & free parameters
free parameters (2)
- t1 (intermediate time control point) =
not reported (varies per case)
- r1 (intermediate position control point, 3 components) =
not reported (varies per case)
assumptions (3)
- domain assumption Two-body dynamics with μ=398600 km^3/s^2
- ad hoc to paper Quadratic non-rational Bézier curves are adequate to approximate the trajectory
- standard math Bernstein polynomials form a complete basis on [0,1]
Cite this review
Pith. "Pith review of A Study on Effective Initial Guess Finding Method Based on B\'ezier Curves: Orbit Determination Applications." pith.science (2026). https://pith.science/paper/OFYWIHQA
@misc{pith2026250613921,
author = {Pith},
title = {Pith review of: A Study on Effective Initial Guess Finding Method Based on B\'ezier Curves: Orbit Determination Applications},
year = {2026},
howpublished = {\url{https://pith.science/paper/OFYWIHQA}},
note = {Machine review of arXiv:2506.13921}
}
read the original abstract
In celestial mechanics, proper orbits related to missions are obtained by solving two-point boundary value problems. Since a selection method of initial value affects the convergence of the solution, developing an effective method to find an initial guess is required. In this work, B\'ezier curves, which can describe complicated curves and surfaces, are utilized to find the initial guess. First, the given problems are transformed into B\'ezier curves forms, and B\'ezier curves' control points, which can handle the shape of curves, are selected by solving the system of nonlinear equations. Finally, the initial guess is obtained by substituting the calculated control points to B\'ezier curves. To validate the performance of the proposed method, numerical simulations are conducted with respect to three kinds of orbits, which are from circular to highly elliptical orbit (HEO). The proposed method is compared to the general shooting method. The comparison results show that the initial guess calculated by B\'ezier curves makes finding the solution more efficient in terms of computational time and iterations. Also, it shows that the proposed method finds the solution for the HEO while the general shooting method fails to find the solution.
Reviewed August 7, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.