REVIEW 4 major objections 4 minor 20 references
Electromagnetic Formation Flying with State and Input Constraints Using Alternating Magnetic Field Forces
T0 review · 4 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash
Pith's one-line read This paper establishes that electromagnetic formation flying can enforce collision, speed, and apparent-power limits simultaneously through one relaxed control barrier function built on frequency-multiplexed magnetic forces.
desk verdict A competent, narrow extension of the authors' own CBF machinery to EMFF, with a real formal gap: the safety theorems hold for a time-averaged model, not the physical plant. 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
Three pieces carry the argument. First, the alternating magnetic field force decomposition: with magnetic moments $u_i(t)=\sum_{j\ne i}p_{ij,k}\sin(\omega_{ij}t)$ and pair-unique frequencies $\omega_{ij}=\omega_{ji}$, Proposition 1 gives $\frac{1}{T}\int_{kT}^{kT+T} f(r,u_i,u_j)\,dt=\frac{1}{2}f(r,p_{ij,k},p_{ji,k})$ when $r$ is held fixed, so pair forces decouple in the time average; the explicit amplitude pair $(c_1,c_2)$ in (13)-(22) realizes any prescribed $f^*$ (Proposition 2), with the magnitude relation in Proposition 3 enabling the power constraint. Second, the control dynamics $\dot{\nu}=-a\nu+a\mu$ (40) promote the control $\nu$ to a state, converting the apparent-power input constraint into a state constraint and raising the relative degrees of the collision and speed barriers. Third, the log-sum-exponential soft minimum $\operatorname{softmin}_\rho(z_1,\dots,z_N)=-\frac{1}{\rho}\log\sum_i e^{-\rho z_i}$ composes all higher-order barrier functions into one relaxed CBF $h$, and the closed-form projection $\mu^*=\mu_d+\lambda L_Gh^T$ with $\lambda$ from (58) enforces $h\ge 0$ while staying as close as possible to the MPC's desired force.
What would settle it
Simulate the unaveraged force model (1)-(3) with the same feedback law (4), (40), (57)-(59) for a formation whose relative positions change by more than a small fraction of their distance within one period $T$, and check whether $\|r_{ij}\|\ge\bar{r}$, $\|v_i-v_j\|\le\bar{v}$, and the apparent-power bound in (O4) hold; a violation while the approximate trajectory stays in $S$ would show that Theorem 2 does not transfer to the physical dynamics.
Extended reading notes
Core claim
The central claim is that all the safety and input constraints can be enforced by one scalar inequality, and the control that respects it can be written in closed form. On the approximate dynamics (9)-(10) obtained by time-averaging the piecewise-sinusoidal magnetic moments, the control (40) with $\mu = \mu^*$ from (57)-(59) minimizes a quadratic cost that penalizes deviation from the formation-seeking force, subject to $b(x,\nu,\hat{\mu},\hat{\eta})\ge 0$, where $b$ is the relaxed control-barrier-function condition built from a soft-minimum composition of higher-order barriers $R_{ij,2}$, $V_{ij,1}$, and $Q_i$. Theorem 2 states that if $(x_0,\nu_0)\in S$, the closed-loop solution remains in $S\subset S_s$ for as long as it is defined, which means objectives (O2)-(O4), namely no collision, bounded relative speed, and bounded apparent power, hold. The numerical example then shows that, with these constraints active, the relative positions $r_{ij}$ converge to the desired offsets $d_{ij}$, achieving objective (O1).
Load-bearing premise
The load-bearing premise is that the physical intersatellite force (1)-(3) can be replaced by its time average over one modulation period $T$, justified only by the statement that the relative position $r_{ij}$ should not change significantly during that period; the safety theorems are proved for this averaged model, and no quantitative bound ties $T$ to the formation's speed, so a fast or close pair could violate a constraint in the physical system even though the averaged model says it is safe.
Editorial extensions
If this is right
- For the approximate time-averaged dynamics, any trajectory starting in $S$ remains in $S\subset S_s$, so no pair comes closer than $\bar{r}$, no pair exceeds relative speed $\bar{v}$, and no satellite exceeds apparent power $\bar{Q}$, while the control continues to seek the formation.
- The constraint-enforcing control (57)-(59) is closed-form, so enforcing safety does not require solving a new optimization at every sample; the only online optimization is the linear MPC that produces the desired force.
- The amplitude construction $(c_1,c_2)$ realizes any prescribed intersatellite force, and Proposition 3's magnitude relation lets the apparent-power limit be expressed through the prescribed force, so the power constraint can be enforced by choosing that force.
- The three-satellite simulation shows that straight-line paths to the desired formation would have caused collisions, while the safe controller reaches the formation with the collision and power barriers active, indicating (O1) is compatible with (O2)-(O4).
Reading between the lines
- Editorial inference: transferring Theorem 2's guarantee to the physical system requires a quantitative bound on how much relative positions drift within one modulation period $T$; the paper leaves that gap open, so the formal safety certificate currently applies to the time-averaged model only.
- Editorial inference: the controller uses full-formation state for both the MPC and the composite barrier, so although the AMFF force decoupling is decentralized, the safety filter as presented is not; a fully decentralized safe version using only neighbor-relative measurements would be a natural extension.
- Editorial inference: Proposition 4 enforces the power limit through a smooth upper bound $\psi$ on $\|p_{ij}\|^2$, so the enforced constraint is conservative; tuning $\epsilon_1$ and $\epsilon_2$ trades conservatism for smoothness, and a numerical sensitivity study could quantify how much performance is lost by that conservatism.
- Editorial inference: the proofs of Propositions 2 and 3 are omitted for brevity, so the force-realization construction is supported by direct computation that a reader would have to re-check, especially because any error there would break the input-constraint argument.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper presents a feedback control algorithm for electromagnetic formation flying (EMFF) with state and input constraints. The approach uses alternating magnetic field forces (AMFF) with piecewise-sinusoidal magnetic moments to decouple intersatellite forces, an explicit construction for choosing the sinusoidal amplitudes that produce a prescribed time-averaged force, and a composite relaxed control barrier function (CBF) that combines collision avoidance, relative-speed limits, and apparent-power limits. The main formal result, Theorem 2, proves forward invariance of a safe set for the approximate time-averaged dynamics (25)-(29) under the stated transversality and Lipschitz assumptions. A three-satellite simulation demonstrates the formation and constraint satisfaction. The central limitation is that the formal safety guarantee is proved only for the approximate model, not for the original physical dynamics (1)-(3), and formation convergence is shown only in simulation.
Significance. The paper has a valuable formal core: for the approximate time-averaged dynamics (25)-(29), it constructs an explicit closed-form control (40), (57)-(59) and proves (Theorem 2) forward invariance of a set S that implies (O2)-(O4), under the stated transversality and Lipschitz assumptions. The amplitude-to-force construction in Section IV and the use of a composite soft-minimum relaxed CBF are nontrivial and original. If the gap between the averaged model and the physical plant were closed, this would be a significant advance for EMFF with safety and input constraints. However, as it stands the headline safety guarantee is not established for the physical EMFF plant, and the formation-convergence claim rests on simulation only. The contribution is nonetheless sufficiently promising to warrant major revision rather than rejection.
major comments (4)
- [Section III-B, Eqs. (7)-(10); Theorem 2] The safety theorem is proved only for the approximate time-averaged model, not for the original plant (1)-(3). Proposition 1 gives exact equality of averaged force only when r_ij is constant on [kT, kT+T); Section III-B replaces (3) by (9)-(10) under the qualitative condition that r_ij does not change significantly over each period. No quantitative bound on the approximation error is provided, and the modulation period T is not part of the hypotheses of Theorem 2. Moreover, the implemented amplitudes are p_ij,k = p_ij(kT), so the actual actuation is a zero-order-hold version of the continuous-time nu(t) analyzed in Theorem 2. Consequently, Theorem 2 does not certify (O2)-(O4) for the physical system; a trajectory that stays in S in the approximate model can violate the constraints when the formation moves appreciably within one interval. The paper should either provide an approximation-error bound and prove robust forward invariance for (1)-(3), or explicitly state that the formal guarantee applies only to the time-averaged model.
- [Section IV, Proposition 2 and Eq. (20)] Proposition 2 asserts that f(r,c1(r,f*),c2(r,f*)) = f* for all r in R^3\{0} and f* in R^3, but the rotation matrix R in (20) contains Phi_2(r,f*) in a denominator, and Phi_2(r,f*) = 0 whenever r and f* are collinear. Thus the construction is undefined exactly on the set of radial force commands. Since the MPC and CBF modules can in principle command radial forces (e.g., along the line between two satellites during braking or collision avoidance), the proposition is false as stated and the implementation lacks a well-defined value on that set. A limiting construction or separate treatment of the collinear case is needed.
- [Section V-D, Theorem 2 and Proposition 5] The central invariance result is conditional on the transversality assumption dh/dnu != 0 for all (x,nu) in B, and on local Lipschitzness of h'. Neither is verified analytically or checked in the simulation; B is defined implicitly through h and the dynamics. Because h is a soft-minimum of many functions, the gradient can vanish when several arguments are equal or when the active argument has zero derivative. The manuscript should provide conditions under which the assumption holds, or verify it on the simulated trajectory, or relax the theorem.
- [Section III-C and Section VI] Objective (O1) is not covered by any theorem; convergence to the desired formation is demonstrated only in the numerical example (Figs. 3-5). The abstract's claim of achieving formation is therefore stronger than the formal results. Please qualify the claim, e.g., achieve formation in simulation, or provide a convergence result for the closed loop under (40), (57)-(59).
minor comments (4)
- [Section IV, Proposition 3] Proposition 3(2) appears to contain a typo: from (13)-(16), when r^T f* = 0 one obtains ||c1||^2 = 2 ||c2||^2, not 1/sqrt(2) times; if the printed ratio is instead intended, Proposition 4 contradicts it. Please correct.
- [Abstract] The phrase 'from a optimal control' should be 'from an optimal control'.
- [Section III-B] The notation introducing the approximate variables is unclear; define \tilde{r}_i and \tilde{v}_i explicitly before (8) so that the relationship between (5)-(7) and (8)-(10) is unambiguous.
- [Section IV] The proof of Proposition 2 is omitted; given that this proposition underpins the entire amplitude construction, consider including the proof or a reference to a full derivation.
Circularity Check
No significant circularity: the EMFF-specific derivation is self-contained, and the main caveat is a physical-vs-averaged-model validity gap, not a circular reduction.
full rationale
The paper's derivation chain is not circular. The EMFF-specific contributions—the closed-form amplitude construction in Section IV (Proposition 2, verified by direct computation: f(r,c1(r,f*),c2(r,f*))=f*) and the application of relaxed-CBF control to the averaged EMFF dynamics (25)-(29)—do not fit any parameter to the conclusion, rename a known result, or define a quantity in terms of the target result. Theorem 2, the main safety claim, is inherited from the authors' prior CBF results ([15, Corollary 3]; also Propositions 5-6), but those prior results are general, parameter-free theorems about relaxed control barrier functions with explicitly stated assumptions (the transversality condition ∂h/∂ν ≠ 0 on B and local Lipschitz continuity of h'), and they do not assume the EMFF safety outcome. Under the stated evaluation rules, such a citation counts as independent support rather than circular self-justification. The soft-minimum and CBF-composition machinery from [15], [16], [18], [19] is invoked as lemmas, not as the claim being derived. The genuine weakness is not circularity: Theorem 2 is proved only for the time-averaged approximate model (25)-(29), with no quantitative bound linking it to the physical force model (1)-(3) when rij changes appreciably over the modulation period T. That is a modeling and validity gap that limits the force of the headline safety claim for the real system, but it does not make the derivation equivalent to its inputs.
Assumptions & free parameters
free parameters (7)
- Control-dynamics bandwidth a =
0.7
- Tracking gain sigma =
3
- Soft-minimum sharpness rho =
10
- Higher-order CBF gains alpha0, alpha1, alpha_v =
5
- CBF gain alpha and slack penalty gamma =
0.02 and 1040
- Smoothing parameters epsilon1, epsilon2 =
not specified
- MPC weights and horizon Tf =
not specified
assumptions (6)
- domain assumption The time-averaged intersatellite force decouples according to Proposition 1 of [8] when relative positions are nearly constant over a modulation period T.
- domain assumption Dipole-dipole magnetic force model (1)-(2) from [6], [7].
- domain assumption Apparent power input constraint (O4) is described by the given quadratic form in the amplitudes.
- domain assumption The composite soft-minimum relaxed CBF results of [15], [16], [18], [19] carry over.
- ad hoc to paper The transversality condition dh/dnu != 0 on the boundary set B and local Lipschitzness of h' hold.
- domain assumption The MPC minimizer exists and is computed in real time.
Cite this review
Pith. "Pith review of Electromagnetic Formation Flying with State and Input Constraints Using Alternating Magnetic Field Forces." pith.science (2026). https://pith.science/paper/XTF5W7WT
@misc{pith2026241116908,
author = {Pith},
title = {Pith review of: Electromagnetic Formation Flying with State and Input Constraints Using Alternating Magnetic Field Forces},
year = {2026},
howpublished = {\url{https://pith.science/paper/XTF5W7WT}},
note = {Machine review of arXiv:2411.16908}
}
read the original abstract
This article presents a feedback control algorithm for electromagnetic formation flying with constraints on the satellites' states and control inputs. The algorithm combines several key techniques. First, we use alternating magnetic field forces to decouple the electromagnetic forces between each pair of satellites in the formation. Each satellite's electromagnetic actuation system is driven by a sum of amplitude-modulated sinusoids, where amplitudes are controlled in order to prescribe the time-averaged force between each pair of satellites. Next, the desired time-averaged force is computed from a optimal control that satisfies state constraints (i.e., no collisions and an upper limit on intersatellite speeds) and input constraints (i.e., not exceeding satellite's apparent power capability). The optimal time-averaged force is computed using a single relaxed control barrier function that is obtained by composing multiple control barrier functions that are designed to enforce each state and input constraint. Finally, we demonstrate the satellite formation control method in a numerical simulation.
Figures
Figures from the paper (4 more)
Reference graph
Works this paper leans on
- [8]
- [15]
- [16]
-
[1]
E. M. C. Kong, D. W. Kwon, S. A. Schweighart, L. M. Elias, R. J. Sedwick, D. W. Miller, Electromagnetic formation flight for multisatellite arrays, J. Spacecraft and Rockets 41 (4) (2004) 659–666
2004
-
[2]
D. W. Kwon, Propellantless formation flight applications using electro- magnetic satellite formations, Acta Astronautica 67 (9) (2010) 1189– 1201
work page 2010
-
[3]
A. K. Porter, D. J. Alinger, R. J. Sedwick, J. Merk, R. A. Opperman, A. Buck, G. Eslinger, P. Fisher, D. W. Miller, E. Bou, Demonstration of electromagnetic formation flight and wireless power transfer, J. Spacecraft and Rockets 51 (6) (2014) 1914–1923
work page 2014
-
[4]
L. M. Elias, D. W. Kwon, R. J. Sedwick, D. W. Miller, Electromag- netic formation flight dynamics including reaction wheel gyroscopic stiffening effects, J. Guid., Contr., Dyn. 30 (2) (2007) 499–511
work page 2007
-
[5]
D. W. Kwon, R. J. Sedwick, S.-i. Lee, J. L. Ramirez-Riberos, Electromagnetic formation flight testbed using superconducting coils, J. Spacecraft and Rockets 48 (1) (2011) 124–134. -2 0 2r12(m) i -2 0 2 j -0.5 0 0.5 k rijdij -3 0 3r13(m) -3 0 3 -1 0 1 0 100 200 300 400 t(s) -2 0 2r23(m) 0 100 200 300 400 t(s) -2 0 2 0 100 200 300 400 t(s) -0.5 0 0.5 Fig. 4...
work page 2011
Show all 20 references
-
[6]
Ahsun, D
U. Ahsun, D. Miller, Dynamics and control of electromagnetic satellite formations, in: Proc. Amer. Contr. Conf., 2006, pp. 1730–1735
2006
-
[7]
S. A. Schweighart, R. J. Sedwick, Explicit dipole trajectory solution 0 20 40h 0 80 160Rij;2 249 249.5 250 Vij;1 (i;j)=(1;2)(i;j)=(1;3)(i;j)=(2;3) 0 50 100 150 200 250 300 350 400 t(s) 25 27 29Qi i=1 i=2 i=3 Fig. 8. Soft-minimum and its constituent barrier functions. 0 2 4 6 8...
2010
-
[9]
R. C. Youngquist, M. A. Nurge, S. O. Starr, Alternating magnetic field forces for satellite formation flying, Acta Astronautica 84 (2013) 197–205
2013
-
[10]
M. A. Nurge, R. C. Youngquist, S. O. Starr, A satellite formation flying approach providing both positioning and tracking, Acta Astronautica 122 (2016) 1–9
2016
-
[11]
Wieland, F
P. Wieland, F. Allg ¨ower, Constructive safety using control barrier functions, IFAC Proceedings V olumes 40 (2007) 462–467
2007
-
[12]
A. D. Ames, X. Xu, J. W. Grizzle, P. Tabuada, Control barrier function based quadratic programs for safety critical systems, IEEE Trans. Autom. Contr. 62 (8) (2017) 3861–3876
2017
-
[13]
W. Xiao, C. Belta, High-order control barrier functions, IEEE Trans. Autom. Contr. 67 (7) (2022) 3655–3662
2022
-
[14]
X. Tan, W. S. Cortez, D. V . Dimarogonas, High-order barrier functions: Robustness, safety, and performance-critical control, IEEE Trans. Autom. Contr. 67 (6) (2022) 3021–3028
2022
-
[17]
Iskandarani, S
M. Iskandarani, S. N. Givigi, G. Fusina, A. Beaulieu, Unmanned aerial vehicle formation flying using linear model predictive control, in: Proc. IEEE Int. Sys. Conf., 2014, pp. 18–23
2014
-
[18]
Rabiee, J
P. Rabiee, J. B. Hoagg, Soft-minimum and soft-maximum barrier functions for safety with actuation constraints, Automatica (2024). doi:10.1016/j.automatica.2024.111921
2024
-
[19]
Safari, J
A. Safari, J. B. Hoagg, Time-varying soft-maximum barrier functions for safety in unmapped and dynamic environments (2024). arXiv: 2409.01458
2024 arXiv
-
[20]
W. S. Cortez, D. V . Dimarogonas, On compatibility and region of attraction for safe, stabilizing control laws, IEEE Trans. Autom. Contr. 67 (9) (2022) 4924–4931
2022
Reviewed August 12, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.