{"id":"d3dac832-ccc1-4751-9d9d-a5fd7f445a66","arxiv_id":"2608.10726","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"An LQR controller for formation flight near halo orbits is computed for a single orbital period and extended to all later times by a fixed rotation, with simulations showing up to 40% less fuel use than an impulsive baseline.","lead":"This paper designs a fuel-efficient controller for spacecraft flying in formation near halo orbits in the Earth-Moon system. The controller's feedback gains only need to be computed for one orbital period and then extended to all later times with a simple rotation, saving onboard computation and memory.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The Proposition's uniqueness step depends on uniform stabilizability of the toroidal LTV pair, which is asserted but never verified; a one-period controllability Gramian check would settle whether the central single-period gain storage claim actually holds.","rationale":"The reader identified the same weakest assumption: uniform stabilizability and detectability are asserted but not verified. My read of the Proposition confirms that this is the single most load-bearing step in the central claim. The proof itself is internally consistent: Eq. (26) follows from the toroidal transformation relation T(t+T)=T(t)Gamma, the orthogonality of Gamma is correctly used in the shifted DRE, and the uniqueness argument is the right mechanism for this kind of quasi-periodic LTV system. The missing piece is not algebraic but existential: without a concrete verification that the actual toroidal pair is uniformly stabilizable, the theorem's conclusion is conditional. Because Q is positive definite, detectability is not a concern, so the decisive numerical check is the one-period controllability Gramian. Since the system is periodic up to an orthogonal similarity, a full-rank Gramian over one period implies uniform controllability on the infinite horizon, which resolves the concern. Conversely, rank deficiency would require a PBH analysis and could invalidate the single-period storage claim. The reader's verdict of CONDITIONAL is therefore appropriate: the manuscript should add this check before the central claim is accepted as fully established. My recommendation is UNCHANGED because the concern reinforces, rather than moves, the reader's conditional verdict.","tokens_in":11683,"tokens_out":9990,"duration_ms":108715,"concrete_test":"Compute the one-period controllability Gramian W_c = sum_{k=0}^{Np-1} Phi_A(Np,k+1) B_k B_k^T Phi_A(Np,k+1)^T using the same A_k and B_k from Eqs. (12)-(13) with the paper's Delta t = 26.89 min on the L1 Northern halo orbit. If rank(W_c)=6, the pair is uniformly controllable over the period, hence uniformly stabilizable, and the Proposition's hypothesis is satisfied. If rank(W_c)<6, examine the uncontrollable subspace via a PBH test to determine whether it is constrained to exponentially stable modes. Also evaluate det R(t) along the orbit to confirm T(t) remains invertible throughout the period.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is the Proposition following Eq. (25): under uniform stabilizability and detectability, the unique stabilizing solution of the difference Riccati equation satisfies P_{k+Np}=Gamma^T P_k Gamma, so the infinite-horizon LQR gains are determined by one orbital period. The algebra of the proof is sound: Eq. (26) is correctly derived from T(t+T)=T(t)Gamma and the orthogonality of Gamma, and the shifted-sequence argument is valid. The load-bearing gap is the hypothesis itself. The paper states that the toroidal LTV pair (A_k,B_k) is uniformly stabilizable and detectable, but it never checks this for the specific L1 halo orbit considered. Because Q is positive definite, detectability is automatic; the real question is uniform stabilizability. Since the system is quasi-periodic up to the orthogonal similarity (A_{k+Np},B_{k+Np})=(Gamma^T A_k Gamma, Gamma^T B_k), uniform stabilizability over the infinite horizon is equivalent to controllability of the lifted system over a single period. The numerical simulations do not settle this: a successful closed-loop run is consistent with stabilizability but does not prove it. If the one-period controllability Gramian were rank-deficient on an uncontrollable unstable or neutrally stable mode, the uniqueness argument would collapse and Eq. (27), and with it the claimed reduction to one period of offline computation, would not be guaranteed. A secondary supporting assumption, invertibility of R(t) along the orbit, is stated but also not verified in the paper; it is less likely to fail but should be monitored in the same check.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper designs an infinite-horizon LQR for relative motion control near periodic halo orbits in the CR3BP, formulated in non-singular toroidal coordinates. The main theoretical contribution is a Proposition showing that, for a quasi-periodic LTV system whose state transition and input matrices satisfy A_{k+Np}=Γ^T A_k Γ and B_{k+Np}=Γ^T B_k, and for a state weight Q satisfying Γ^T Q Γ=Q, the unique stabilizing solution of the difference Riccati equation satisfies P_{k+Np}=Γ^T P_k Γ. Consequently the optimal gains are determined by one orbital period of offline computation and can be extended online by a fixed rotation. The controller is tested on the full nonlinear CR3BP dynamics for an L1 Northern halo orbit in the Earth-Moon system and on a high-fidelity ephemeris model with solar perturbation, and is compared with an impulsive targeting plus station-keeping baseline, reporting a roughly 40% reduction in total ΔV at comparable tracking accuracy.","tokens_in":12024,"tokens_out":6682,"duration_ms":66655,"significance":"If the Proposition holds for the specific orbit used, the paper removes the apparent need to store an infinite non-repeating gain sequence for quasi-periodic toroidal LTV dynamics. The proof itself is a clean direct substitution into the Riccati recursion and is correct conditional on uniform stabilizability and detectability; no fitted or data-driven quantity enters the periodicity argument. The paper also demonstrates a working implementation on a realistic ephemeris model, and the comparison against an impulsive baseline gives a concrete, falsifiable performance claim. The main significance is therefore the compact gain representation and the accompanying numerical demonstration, both of which are valuable for practical cislunar formation flying.","major_comments":[{"comment":"The Proposition is load-bearing for the single-period gain storage claim, but its key hypothesis, uniform stabilizability and detectability of the discrete LTV pair (A_k,B_k), is asserted and never verified for the L1 halo orbit used in §5. Since Q is positive definite, detectability is automatic; the real condition to check is uniform stabilizability. For the quasi-periodic structure (A_{k+Np},B_{k+Np})=(Γ^T A_k Γ,Γ^T B_k), this is equivalent to a rank condition on the single-period controllability Gramian of the lifted system, and a short numerical check would settle it. As written, the uniqueness argument in Eqs. (28)-(30) guarantees P_{k+Np}=Γ^T P_k Γ only if that condition holds. A successful closed-loop simulation is consistent with stabilizability but does not prove it, and if the condition failed, the central claim that one period of offline computation suffices would not be guaranteed.","section":"§4.3, Proposition after Eq. (25)"},{"comment":"The abstract and conclusions state that the proposed controller attains a tracking accuracy 'comparable' to the baseline, but the numbers in Table 1 do not directly support that wording: after the first revolution, the LQR RMS error is 0.394 km versus 0.174 km for the baseline, and the maximum error is 2.060 km versus 1.687 km. The 40% ΔV reduction is meaningful, but 'comparable' needs an operational definition, for example the same RMS or maximum-error threshold, or a Pareto comparison from the sensitivity analysis in Fig. 9. As reported, the baseline is roughly twice as accurate on RMS, so the comparison should be framed as a trade-off rather than an equivalence.","section":"§6, Table 1 and Fig. 8"},{"comment":"The ephemeris validation consists of a single reconfiguration trajectory with gains obtained by cubic-spline interpolation of CR3BP-derived eigenvectors and gains. This introduces an approximation that is not quantified: the ephemeris trajectory is not exactly periodic, and the interpolation is performed over one CR3BP revolution while the actual revolution durations vary. The claim that the single-period gain representation remains valid on the ephemeris model would be substantially stronger if the authors reported a second epoch or reference phasing, or directly checked that the interpolated gains stabilize the linearized ephemeris model over several revolutions. As it stands, the validation demonstrates feasibility but not robustness of the one-period storage claim outside the CR3BP model.","section":"§6.2, LQR controller performance on ephemeris model"}],"minor_comments":[{"comment":"The sign convention for the rotation matrix R_{θλ} and the resulting Γ is not made explicit in the implementation. Since Eq. (31) gives K_{k+Np}=K_k Γ while Eq. (37) uses K(t)=K(s_curr)Γ^p, a reader cannot immediately tell whether the online rotation is by +mθ or -mθ; please state the convention clearly.","section":"Eq. (17) and Eq. (37)"},{"comment":"The bottom plots are labeled 'Acc. ΔV [mm/s]' but the text describes the quantity as accumulated control effort; please clarify whether the plotted value is cumulative and why the unit is mm/s rather than m/s.","section":"Fig. 2 and Fig. 6"},{"comment":"There is a typo in 'Tuning appropiately the values' and the sensitivity figure would benefit from an explicit statement of whether the plotted curves correspond to the same simulation time window and the same initial/final states as Table 1.","section":"§6, Sensitivity analysis"}],"recommendation":"major_revision","confidential_remarks":"The central Proposition is a correct but conditional calculation; the missing verification of uniform stabilizability is the main technical gap and should be required before acceptance. The practical comparisons are suggestive but need a clearer definition of 'comparable accuracy.' The paper is otherwise well organized and the application is timely for the cislunar formation-flying community."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a real result. The single-period gain storage theorem is clean, the algebra checks out, and the 40% fuel saving appears genuine. The main gap is that uniform stabilizability is asserted, not verified, and the validation is thinner than the abstract suggests. With a controllability check and a more honest accuracy statement, this becomes a solid paper.\n\nThe genuinely new thing is the Proposition after Eq. (25). The toroidal LTV system has the quasi-periodic structure A_{k+Np} = Gamma^T A_k Gamma and B_{k+Np} = Gamma^T B_k, where Gamma is an orthogonal rotation. The paper shows by direct substitution that if Q is invariant under Gamma, the difference Riccati solution satisfies P_{k+Np} = Gamma^T P_k Gamma, and hence the optimal gains are K_{k+Np} = K_k Gamma. I traced the algebra; the shifted-sequence uniqueness argument is legitimate. This is a real extension of periodic LQR (Lyapunov-Floquet) to the toroidal quasi-periodic case, and it is not present in the cited literature. No fitted parameters hide in the result. Credit where it is due.\n\nThe soft spot is not the algebra but the hypothesis. The Proposition relies on uniform stabilizability and detectability to identify the unique bounded stabilizing solution. Detectability is automatic because Q > 0. Uniform stabilizability is never checked for the specific L1 halo orbit. Because the system is quasi-periodic up to orthogonal similarity, uniform stabilizability is equivalent to controllability of the lifted single-period system; a rank check of the one-period controllability Gramian would settle it. The closed-loop simulations suggest stabilizability holds for the tested case, but that is an empirical hint, not proof. If an uncontrollable mode were neutrally stable or unstable, the uniqueness argument would break, and with it the single-period storage claim. The invertibility of R(t) is also stated but not verified; it is unlikely to fail for a non-rectilinear mode, but it should be checked in the same pass.\n\nThe numerical validation is single-orbit and single-baseline. The headline 'comparable tracking accuracy' is generous. Table 1 gives LQR RMS 0.394 km versus baseline 0.174 km, and max errors 2.06 km versus 1.69 km. That is roughly a factor of two worse accuracy. The honest statement is that the LQR trades accuracy for a 40% fuel saving. The ephemeris test also uses spline-interpolated gains, which is a reasonable engineering workaround but adds another layer of approximation. No code or data is shipped, which makes reproduction harder.\n\nWho is this for: people working on formation flying near halo orbits, and anyone interested in quasi-periodic LQR reductions. It deserves a serious referee. My recommendation: conditional acceptance, with a required one-period controllability/Gramian check, a comment on R(t) invertibility, a softened accuracy comparison, and ideally a code/data release.","headline":"A genuinely new quasi-periodic LQR reduction with a clean proof, but the load-bearing stabilizability hypothesis is asserted rather than verified and the validation is too thin to fully support the accuracy claims.","tokens_in":12575,"tokens_out":2297,"would_cite":true,"duration_ms":24108,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Quasi-periodic symmetry reduces infinite-horizon LQR for halo-orbit formation flying to a single-period computation.","keywords":["linear quadratic regulator","toroidal coordinates","quasi-periodic symmetry","halo orbits","circular restricted three-body problem","difference Riccati equation","formation flying","cislunar space"],"falsifier":"Compute sliding-window controllability and observability Gramians of $(A_k, B_k)$ along the L1 halo orbit over many periods: if their singular values are not uniformly bounded away from zero and infinity, or if $R(t)$ becomes singular, the uniqueness argument for the Riccati solution fails and the relation $P_{k+N_p} = \\Gamma^T P_k \\Gamma$ is not guaranteed.","tokens_in":11463,"feed_emoji":"🛰️","tokens_out":5347,"duration_ms":47293,"temperature":0.7,"pith_summary":"The paper aims to show that an infinite-horizon linear-quadratic regulator for relative motion near halo orbits can be designed with only one orbital period of offline computation. The trick is to write the relative dynamics in non-singular toroidal coordinates, where the time-varying coordinate frame rotates by a fixed angle every period. The paper proves, via a uniqueness argument on the stabilizing solution of the difference Riccati equation, that the Riccati matrices and optimal gains at later periods are just rotated copies of the first-period ones. If correct, this removes the need to store an infinite, non-repeating gain sequence and makes the optimal controller practical for onboard cislunar formation flying. Simulations on the nonlinear CR3BP and on an ephemeris model with solar perturbation support the claim, with a 40% reduction in total control effort compared with an impulsive baseline.","feed_headline":"One orbit period yields every optimal gain for halo formation flight","feed_subtitle":"Toroidal-coordinate rotation makes the infinite-horizon controller a single stored period plus a rotation, saving about 40 percent delta-v.","key_machinery":"The central object is the orthogonal matrix $\\Gamma = \\mathrm{blockdiag}(R_{\\theta_\\lambda}, R_{\\theta_\\lambda})$, describing how the non-singular toroidal frame returns after one reference-orbit period: $R(t+T) = R(t) R_{\\theta_\\lambda}$, where $R_{\\theta_\\lambda}$ is a rotation by the center-mode angle $\\theta_\\lambda$. This quasi-periodic symmetry of the coordinate transformation turns the otherwise non-periodic linear time-varying dynamics into a quasi-periodic system with $A_{k+N_p} = \\Gamma^T A_k \\Gamma$ and $B_{k+N_p} = \\Gamma^T B_k$. The argument then invokes the standard uniqueness of the bounded stabilizing solution of the difference Riccati equation under uniform stabilizability and detectability, and shows the rotated sequence $\\Gamma^T P_k \\Gamma$ satisfies the same recursion, forcing $P_{k+N_p} = \\Gamma^T P_k \\Gamma$ and $K_{k+N_p} = K_k \\Gamma$.","core_discovery":"On the discrete-time linear time-varying system obtained in toroidal coordinates, the state transition and input matrices satisfy $A_{k+N_p} = \\Gamma^T A_k \\Gamma$ and $B_{k+N_p} = \\Gamma^T B_k$, where $\\Gamma$ is an orthogonal matrix that rotates the toroidal frame by the center-mode angle over one orbital period. When the state weight $Q$ is invariant under this rotation, $\\Gamma^T Q \\Gamma = Q$, the paper proves that the unique stabilizing solution of the infinite-horizon difference Riccati equation obeys $P_{k+N_p} = \\Gamma^T P_k \\Gamma$, and consequently the optimal feedback gains satisfy $K_{k+N_p} = K_k \\Gamma$. Therefore all future gains are obtained from the gains of a single period by a fixed rotation. The paper validates this controller on an L1 Northern halo orbit in the Earth-Moon system, first against the full nonlinear CR3BP and then against a high-fidelity ephemeris model, showing successful reconfiguration on the invariant torus with about 40% lower total delta-v than an impulsive targeting-plus-station-keeping baseline at comparable tracking accuracy.","pith_inferences":["By the same logic, any quasi-periodic system whose dynamics are invariant up to an orthogonal congruence should admit a one-period Riccati computation; the proof does not use specific CR3BP structure beyond the $\\Gamma$ relation.","A numerical verification of uniform stabilizability and detectability on the sampled halo orbit would close the gap between the Proposition's hypotheses and the tested case.","The compact gain representation could be combined with model predictive control terminal ingredients to obtain receding-horizon stability guarantees, as the paper itself suggests as future work.","The 40% delta-v comparison is for one reconfiguration scenario; testing across phasings near perilune, as done for the CR3BP case, would show whether the saving persists on the ephemeris model."],"forward_implications":["Only one orbital period of Riccati recursion needs to be computed offline; every later gain is a rotation of that period's gains.","Onboard storage shrinks from an infinite gain sequence to one period of gains plus the scalar rotation angle $\\theta_\\lambda$.","The quasi-periodic LTV system admits a well-defined infinite-horizon LQR with finite memory, which is not possible for a generic LTV system without such symmetry.","On the ephemeris model, the controller keeps RMS position error below 400 m after the first revolution and reduces total delta-v from 2.99 m/s to 1.79 m/s compared with the impulsive baseline.","Tuning the weight matrices trades control effort against perilune error spikes, so the same controller can be made more or less aggressive."],"supporting_citations":[{"why":"Defines the non-singular toroidal coordinates and the transformation matrix $T$ whose quasi-periodicity is exploited throughout the paper.","marker":"[5]"},{"why":"Supplies the uniqueness theorem for the bounded stabilizing solution of the difference Riccati equation used in the Proposition.","marker":"[10]"},{"why":"Provides the impulsive targeting and station-keeping baseline controller used for the 40% delta-v comparison.","marker":"[9]"},{"why":"Establishes that nearby relative motion near periodic orbits is governed by invariant 2-tori, motivating the toroidal formulation.","marker":"[4]"},{"why":"Provides the planetary ephemeris data used for the high-fidelity validation.","marker":"[11]"},{"why":"Provides the geometry-handling toolkit used to build the ephemeris model.","marker":"[12]"}],"fun_headline_variants":["One orbit of gains steers all halo formation maneuvers","LQR on a torus cuts formation-keeping delta-v by 40%","Quasi-periodic symmetry packs all halo gains into one pass","Rotate one period of gains, get every halo-orbit LQR gain","Toroidal-symmetry LQR saves 40% delta-v for halo flight"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof assumes the discrete linear time-varying pair $(A_k, B_k)$ in toroidal coordinates is uniformly stabilizable and detectable, and that the toroidal frame $R(t)$ stays invertible along the orbit; the paper does not verify either condition for the specific halo orbit.","fun_headline_variants_meta":{"raw":{"variants":["One orbit of gains steers all halo formation maneuvers","LQR on a torus cuts formation-keeping delta-v by 40%","Quasi-periodic symmetry packs all halo gains into one pass","Rotate one period of gains, get every halo-orbit LQR gain","Toroidal-symmetry LQR saves 40% delta-v for halo flight"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000204,"raw_usage":{"total_tokens":1390,"prompt_tokens":947,"completion_tokens":443,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":563,"completion_tokens_details":{"reasoning_tokens":348}},"tokens_in":563,"tokens_out":443,"duration_ms":4733,"temperature":1.0,"reasoning_tokens":348,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T18:34:48.550330+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute sliding-window controllability and observability Gramians of $(A_k, B_k)$ along the L1 halo orbit over many periods: if their singular values are not uniformly bounded away from zero and infinity, or if $R(t)$ becomes singular, the uniqueness argument for the Riccati solution fails and the relation $P_{k+N_p} = \\Gamma^T P_k \\Gamma$ is not guaranteed.","supporting_citations":[{"cited_title":"Describing relative motion n ear periodic orbits via local toroidal coordi- nates,","cited_arxiv_id":null,"evidence_quote":"Defines the non-singular toroidal coordinates and the transformation matrix $T$ whose quasi-periodicity is exploited throughout the paper."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the uniqueness theorem for the bounded stabilizing solution of the difference Riccati equation used in the Proposition."},{"cited_title":"Impulsive control of formati ons near invariant tori via local toroidal co- ordinates,","cited_arxiv_id":null,"evidence_quote":"Provides the impulsive targeting and station-keeping baseline controller used for the 40% delta-v comparison."},{"cited_title":"Mo del predictive control for formation recon- figuration exploiting quasi-periodic tori in the cislunar e nvironment,","cited_arxiv_id":null,"evidence_quote":"Establishes that nearby relative motion near periodic orbits is governed by invariant 2-tori, motivating the toroidal formulation."},{"cited_title":"Ancillary data services of NASA ’s navig ation and ancillary information facility,","cited_arxiv_id":null,"evidence_quote":"Provides the planetary ephemeris data used for the high-fidelity validation."},{"cited_title":"A look t owards the future in the handling of space science mission geometry,","cited_arxiv_id":null,"evidence_quote":"Provides the geometry-handling toolkit used to build the ephemeris model."}],"review_version":1}