Pith. sign in

REVIEW 3 major objections 5 minor 14 references

Genetic Fuzzy System-based Control for Final Approach of Spacecraft Rendezvous and Proximity Operations

T0 review · 3 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read A genetic-algorithm-tuned fuzzy controller can perform the final approach of a spacecraft rendezvous, and a test with an untrained disturbance indicates it does so while keeping energy use low.

desk verdict A clean, reproducible GFS controller for final-approach rendezvous, but 'energy minimality' is asserted rather than demonstrated; no baseline comparison appears. read the letter →

arxiv 2608.12760 v1 pith:GKOKZCDE submitted 2026-08-13 physics.space-ph cs.ROmath.OC

Daegyun Choi , Donghoon Kim , Henzeh Leeghim This is my paper · ORCID
classification physics.space-phcs.ROmath.OC
keywords fuzzyinferencesystemgeneticalgorithmspacecraftrendezvousproximityoperationsenergyminimizationClohessy-Wiltshireequationsinterpretablecontrol
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

This paper proposes a controller for the final approach phase of spacecraft rendezvous: a fuzzy inference system, tuned offline by a genetic algorithm, that maps relative position error and relative velocity error into thrust for each axis. The authors aim to get the chaser to a cooperative target on a circular orbit while nearly minimizing energy consumption, measured by the integrated squared control force $\int_0^{t_f} f^T f\,dt$ plus a penalty for risky behavior. Training covers multiple initial positions without disturbances, yet a separate test case with a sinusoidal disturbance not seen in training shows the chaser converging to within $\pm 0.5$ m in position and $\pm 0.01$ m/s in velocity while the control force stays below the disturbance level. The design matters because it combines the interpretability of If-Then rules with the optimizing power of a genetic algorithm, an attribute the paper notes is absent from reinforcement-learning controllers.

What carries the argument

The central object is a genetic fuzzy system (GFS): a Mamdani-type fuzzy inference system whose membership functions and rules are tuned by a genetic algorithm. Each axis has its own two-input, one-output FIS, with normalized relative position error and normalized relative velocity error as inputs and normalized force as the output. Five triangular membership functions (large negative, small negative, zero, small positive, large positive) keep the rule base simple, and symmetry is used to reduce the tunable parameters to 30, stored in a single vector that the GA optimizes. The fitness function $J = \int_0^{t_f} f^T f\,dt + \rho$ is what makes the training concrete: it rewards low energy and penalizes unsafe or jerky approaches, and the penalty threshold is set high enough to exclude undesirable trajectories.

What would settle it

Run the same Clohessy-Wiltshire simulation for the paper's initial conditions and solve the constrained minimum-energy control problem numerically to obtain a benchmark cost; if the fuzzy controller's $J$ is substantially larger than that benchmark, or if a second untrained disturbance profile (for example, a constant bias or a different frequency) drives the chaser outside the reported convergence bands, the central claim of near-minimal energy and robustness is not supported.

Watch

Extended reading notes

Core claim

The central claim is that a compact fuzzy rule base—five triangular membership functions per input and output and a 30-parameter design—can realize a near-minimal-energy final-approach controller once a genetic algorithm tunes it. The controller is built on the Clohessy-Wiltshire relative-motion equations, and the fitness function $J = \int_0^{t_f} f^T f\,dt + \rho$ combines the energy objective with penalties for approaching too fast or commanding abrupt force changes. In the reported training results the relative position, velocity, and force all converge smoothly to zero, and in the testing scenario, with a disturbance not included in training, the chaser stays within $\pm 0.5$ m of the target over the final 200 seconds with relative velocity below $\pm 0.01$ m/s. The authors take this as evidence that a fuzzy controller trained without uncertainties is robust to them while still producing an interpretable rule table.

Load-bearing premise

The load-bearing premise is that a five-by-five triangular fuzzy rule base with only 30 tunable parameters can represent the true minimum-energy control policy closely enough to be near-optimal; if that rule base is too coarse, the claimed energy savings and robustness do not follow.

Editorial extensions

If this is right

  • A chaser using the trained controller can start the final approach from the tested positions in the $\pm 5$ m to $\pm 100$ m range without retraining, because the FIS inputs and output are normalized.
  • The controller respects the 0.5 N thrust limit and produces force that converges smoothly to zero, so the simulated close approach avoids high closing speeds and abrupt actuator commands.
  • The trained rule table remains human-readable after optimization, so a mission designer can inspect the If-Then logic that produced a given maneuver.
  • Because the design and training loop depend only on the explicit Clohessy-Wiltshire dynamics and fitness function, the same pipeline can be rerun for another target orbit or chaser mass to obtain a new tuned rule base.

Reading between the lines

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

  • A direct extension would be to sweep disturbance frequencies, amplitudes, and directions during testing, since the single sinusoidal profile used here leaves the boundary of the controller's robust region unspecified.
  • The objective $\int f^T f\,dt$ is integrated squared force, not propellant mass; for thrusters with non-quadratic specific-impulse curves, the near-minimal-energy solution may not be the near-minimal-fuel solution.
  • Because the FIS is per-axis and uses only local relative states, it could sit beneath a higher-level collision-avoidance planner without changing its internal structure; the paper lists obstacle avoidance as future work.
  • Including disturbances in the GA training itself, rather than only in testing, would likely widen the set of disturbances the controller can reject, at the cost of more computation.
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.

Referee Report

3 major / 5 minor

Summary. The paper proposes a genetic fuzzy system (GFS) controller for the final approach phase of spacecraft rendezvous and proximity operations. The chaser dynamics are modeled with the Clohessy-Wiltshire equations in a local-vertical/local-horizontal frame, and a fuzzy inference system with two inputs (relative position error and relative velocity error) per axis is used to generate control force. The fuzzy membership functions and rule base contain 30 parameters that are tuned offline by a genetic algorithm using a fitness function J = ∫ f^T f dt + ρ, where ρ penalizes undesirable behaviors such as sudden control changes and high terminal velocity. Training is conducted over multiple initial relative positions without disturbances, and testing is performed on a single scenario with a deterministic sinusoidal disturbance. The paper claims that the trained controller provides proper control inputs that make the chaser approach the target while minimizing energy consumption, and that the controller is robust to uncertainties.

Significance. If the energy-minimality and robustness claims were fully supported, the work would offer a useful, explainable alternative to reinforcement-learning-based proximity operations controllers. The GFS framework is coherent, the use of multiple training scenarios is a reasonable step toward generalization, and the numerical simulations demonstrate feasible convergence for the presented training and testing cases. However, the evidence as presented is limited: the central claims of energy minimization and robustness rest on a single deterministic test, no comparison with any baseline controller or optimal-control lower bound is given, and the fitness function is not the control-effort objective that the abstract claims to minimize. The paper therefore currently establishes feasibility rather than optimality or robustness.

major comments (3)
  1. [Optimization of Controller, Eq. (4)] The central claim that the controller minimizes energy consumption is not supported. The fitness function is J = ∫ f^T f dt + ρ, where ρ is a penalty for sudden control changes and high velocity near the target. The paper itself states that the true energy-optimal solution is bang-off-bang and is deliberately avoided, so the optimized quantity is a smoothness-penalized control effort, not energy consumption. Without an optimal-control lower bound, an LQR or MPC baseline, or a comparison against bang-off-bang control with the same terminal constraints, the abstract's claim of energy minimization is unverifiable; any smooth stabilizing controller could produce the displayed time histories.
  2. [Simulation Study, Tables 4 and 5] The conclusion that the fuzzy controller is 'robust to uncertainties' is supported by only one deterministic disturbance profile and one initial condition. A single trajectory with one sinusoidal disturbance cannot establish robustness to untrained disturbances. The authors should either add Monte Carlo simulations with varied disturbance amplitudes, frequencies, and phases, or temper the robustness claim to state that the controller performed well in one tested disturbance scenario.
  3. [Simulation Study, Figures 5 and 6] No comparison is made with any existing rendezvous controller or guidance law, so the claim that the trained FIS 'can generate near-minimal control force' is not demonstrated. The figures show convergence and bounded control effort, which are useful feasibility results, but they do not quantify the gap between the achieved performance and the true minimum control effort. Adding a baseline controller (e.g., a linear-quadratic regulator or a standard CW guidance law) and reporting the fitness and control-effort values would make the optimality claim testable.
minor comments (5)
  1. [Title page] The word 'SP ACECRAFT' in the title appears to be a typo for 'SPACECRAFT'.
  2. [Table 5] The disturbance vector notation is ambiguous; the three components should be listed with explicit separators, e.g., d = [d_x, d_y, d_z]^T.
  3. [Relative dynamic model, Eq. (3)] The definition of the mean motion n is typeset as 'n = p µ/r2t'; this should be corrected to n = sqrt(µ/r_t^3).
  4. [Simulation Study, penalty conditions] Table 4 states that the first penalty condition applies to the last 30 seconds, but the text reports an average over the last 50 seconds; these values should be harmonized.
  5. [References] Reference [2] contains a corrupted bibliographic string, '125doi=3–1264', which should be corrected.

Circularity Check

1 steps flagged · score 6.0 of 10

Energy-minimality claim reduces to the GA fitness J; robust disturbance rejection is independently tested.

  1. fitted input called prediction [Optimization of Controller, Eq. (4); Simulation Study, paragraph beginning 'One can obtain the trained FIS...']
    "the fitness function is defined as J = ∫_0^{t_f} f^T f dt + ρ. ... One can obtain the trained FIS that can generate near-minimal control force after the training process."

    The GA's only objective is to minimize J in Eq. (4), and J's leading term is the same ∫ f^T f dt energy measure that the abstract says is minimized. Thus the statement that the trained FIS generates 'near-minimal control force' is not an independent finding; it is the training objective restated. No optimal-control lower bound, baseline, or testing-scenario energy comparison is provided, so the energy-minimality assertion is forced by the fitting setup. The disturbance-rejection test does independently support convergence, so the circularity covers only the energy claim.

full rationale

The paper's main robustness claim — that a GA-trained fuzzy controller can bring the chaser to the target under an untrained disturbance — is tested on a separate scenario and is not circular. The training scenarios use different initial positions from the test, and the test includes a time-varying disturbance absent from training. The self-citations [6]-[8] are motivational rather than load-bearing; the GFS methodology is supported by external references [13] and [14], and the dynamics are the standard Clohessy-Wiltshire equations [12]. However, the energy-minimization component of the central claim is essentially the training objective restated: the GA directly minimizes J in Eq. (4), and the paper then asserts that the trained FIS generates 'near-minimal control force' without any independent optimal-control benchmark or lower bound. That particular sub-claim reduces by construction to the fitted fitness function. The non-constructive universal-approximation citation [13] and the lack of expressiveness bounds are correctness risks, not circularity. Overall, the circularity is partial: the robust approach is independently demonstrated, but the energy-minimality claim is supported only by the optimization setup itself.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The central claim rests on standard CW dynamics, a cooperative target assumption, and the representational capacity of the chosen FIS. The GA-tuned parameters and hand-chosen penalty weights are free parameters. The testing disturbance is ad hoc and not justified as representative.

free parameters (3)
  • FIS parameters p (30 values) = GA-optimized
    Membership function edges/centers and rule consequents are tuned by GA to minimize J; no independent constraint.
  • Penalty weight rho = 100
    Chosen by hand as 'greater than total control force'; affects the trade-off between energy and smoothness.
  • Penalty thresholds = 0.05 m/s and 0.1 N/s
    Hand-chosen to enforce smooth approach; not derived from mission requirements.
assumptions (5)
  • standard math Clohessy-Wiltshire equations
    Used to model relative motion under circular orbit and small separation; cited [12].
  • domain assumption Target on circular orbit and relative distance small
    Stated in Section 'Relative Dynamic Model'; required for CW equations.
  • domain assumption Target is cooperative and shares state information
    Assumed in Section 'Relative Dynamic Model', enabling measurement of relative state.
  • standard math Fuzzy inference system is a universal approximator
    Cited [13] to justify using FIS as a controller; no constructive bounds given.
  • ad hoc to paper Disturbance in testing is representative
    The single sinusoidal disturbance in Table 5 is chosen by authors; no justification for coverage of real orbital disturbances.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Genetic Fuzzy System-based Control for Final Approach of Spacecraft Rendezvous and Proximity Operations." pith.science (2026). https://pith.science/paper/GKOKZCDE

@misc{pith2026260812760,
  author       = {Pith},
  title        = {Pith review of: Genetic Fuzzy System-based Control for Final Approach of Spacecraft Rendezvous and Proximity Operations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GKOKZCDE}},
  note         = {Machine review of arXiv:2608.12760}
}
read the original abstract

In-space servicing has been receiving great attention to extend the operation of spacecraft with defective components. This requires rendezvous and proximity operations for a chaser to provide service to a target. This work constructs a fuzzy inference system-based controller for the chaser to reach the cooperative target on a circular orbit in the final approach phase while minimizing the energy consumption of the chaser. The offline training process performed by a genetic algorithm deals with multiple initial relative positions of the chaser, and the trained controller is validated using a testing environment with disturbances, which differs from the training scenarios.

Figures

Figures reproduced from arXiv: 2608.12760 by the authors.

Figure 1
Figure 1. Phasing and Rendezvous Process proposed control laws combine the APF with non-singular terminal sliding mode control,3 sub￾optimal sliding mode control,4 and backstepping control.5 With the advancement of computing technologies, some researchers have applied artificial intelligence (AI) techniques to diverse control problems and platforms, such as explainable AI-based spacecraft attitude control,6 decentralized cont… view at source ↗
Figure 2
Figure 2. Reference Frames angular momentum vector of the target’s orbit, as shown in [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Block Diagram of the Genetic Fuzzy System [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: MFs for the FIS Model [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Chaser’s State (Training) [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: Chaser’s State (Testing) position and velocity error. Also, the proposed fuzzy controller is trained by a genetic algorithm. The offline training process deals with multiple initial relative positions of the chaser while the trained controller is applied to a testing e…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

14 extracted references · 11 canonical work pages

  1. [1]

    Attitude Dynamics of Spacecraft with Time-Varying Inertia Dur- ing On-Orbit Refueling,

    Z. Guang, Z. Heming, and B. Liang, “Attitude Dynamics of Spacecraft with Time-Varying Inertia Dur- ing On-Orbit Refueling,”Journal of Guidance, Control, and Dynamics, V ol. 41, No. 8, 2018, pp. 1744– 1754, 10.2514/1.G003474. 9

    work page doi:10.2514/1.g003474 2018

  2. [2]

    Hybrid Control of Space Robot in On-Orbit Screw- Driving Operation,

    L. Shi, J. Katupitiya, and N. M. Kinkaid, “Hybrid Control of Space Robot in On-Orbit Screw- Driving Operation,”IEEE Transactions on Aerospace and Electronic Systems, V ol. 54, No. 3, 2018, pp. 125doi=3–1264, 10.1109/TAES.2017.2780558

    work page arXiv 2018

  3. [3]

    Non-cooperative autonomous rendezvous and docking using artificial potentials and sliding mode control,

    X. Li, Z. Zhu, and S. Song, “Non-cooperative autonomous rendezvous and docking using artificial potentials and sliding mode control,”Proceedings of the Institution of Mechanical Engineers, Part G: Journal of Aerospace Engineering, V ol. 233, No. 4, 2019, pp. 1171–1184, 10.1177/0954410017748988

    work page doi:10.1177/0954410017748988 2019

  4. [4]

    Suboptimal artificial potential function sliding mode control for spacecraft rendezvous with obstacle avoidance,

    L. Cao, D. Qiao, and J. Xu, “Suboptimal artificial potential function sliding mode control for spacecraft rendezvous with obstacle avoidance,”Acta Astronautica, V ol. 143, 2018, pp. 133–146, https://doi.org/10.1016/j.actaastro.2017.11.022

    work page doi:10.1016/j.actaastro.2017.11.022 2018

  5. [5]

    Artificial potential field based robust adaptive control for spacecraft rendezvous and docking under motion constraint,

    Q. Li, J. Yuan, B. Zhang, and H. Wang, “Artificial potential field based robust adaptive control for spacecraft rendezvous and docking under motion constraint,”ISA Transactions, V ol. 95, 2019, pp. 173– 184, https://doi.org/10.1016/j.isatra.2019.05.018

    work page doi:10.1016/j.isatra.2019.05.018 2019

  6. [6]

    Genetic Algorithm-Aided Fuzzy Controller for Spacecraft Attitude Maneuver with Uncertainties,

    D. Choi, A. Chhabra, and D. Kim, “Genetic Algorithm-Aided Fuzzy Controller for Spacecraft Attitude Maneuver with Uncertainties,”ASCEND 2021, 2021, 10.2514/6.2021-4174

    work page doi:10.2514/6.2021-4174 2021

  7. [7]

    Intelligent Multi-Robot System for Collaborative Object Transportation Tasks in Rough Terrains,

    D. Choi and D. Kim, “Intelligent Multi-Robot System for Collaborative Object Transportation Tasks in Rough Terrains,”Electronics, V ol. 10, No. 12, 2021, 10.3390/electronics10121499

    work page doi:10.3390/electronics10121499 2021

  8. [8]

    Intelligent cooperative collision avoidance via fuzzy potential fields,

    D. Choi, A. Chhabra, and D. Kim, “Intelligent cooperative collision avoidance via fuzzy potential fields,”Robotica, 2021, p. 1–20, 10.1017/S0263574721001454

    work page doi:10.1017/s0263574721001454 2021

Show all 14 references
  1. [9]

    Deep Learning Techniques for Autonomous Spacecraft Guidance During Proximity Operations,

    L. Federici, B. Benedikter, and A. Zavoli, “Deep Learning Techniques for Autonomous Spacecraft Guidance During Proximity Operations,”Journal of Spacecraft and Rockets, V ol. 58, No. 6, 2021, pp. 1774–1785, 10.2514/1.A35076

    2021 doi

  2. [10]

    Autonomous Six-Degree-of-Freedom Spacecraft Docking with Rotating Targets via Reinforcement Learning,

    C. E. Oestreich, R. Linares, and R. Gondhalekar, “Autonomous Six-Degree-of-Freedom Spacecraft Docking with Rotating Targets via Reinforcement Learning,”Journal of Aerospace Information Sys- tems, V ol. 18, No. 7, 2021, pp. 417–428, 10.2514/1.I010914

    2021 doi

  3. [11]

    Applications of fuzzy algorithms for control of a simple dynamic plant,

    E. H. Mamdani, “Applications of fuzzy algorithms for control of a simple dynamic plant,”Proceedings of the IEEE, 1974

    1974

  4. [12]

    Terminal Guidance System for Satellite Rendezvous,

    W. H. CLOHESSY and R. S. WILTSHIRE, “Terminal Guidance System for Satellite Rendezvous,” Journal of the Aerospace Sciences, V ol. 27, No. 9, 1960, pp. 653–658, 10.2514/8.8704

    1960 doi

  5. [13]

    Fuzzy systems as universal approximators,

    B. Kosko, “Fuzzy systems as universal approximators,”IEEE Transactions on Computers, V ol. 43, No. 11, 1994, pp. 1329–1333, 10.1109/12.324566

    1994 doi

  6. [14]

    Herrera, F

    F. Herrera, F. Hoffmann, and L. Magdalena,Genetic Fuzzy Systems: Evolutionary Tuning and Learning of Fuzzy Knowledge Bases. Singapore: World Scientific Publishing Company, 2001. 10

    2001

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.