pith. sign in

arxiv: 2605.29021 · v1 · pith:VNRGGXFPnew · submitted 2026-05-27 · 💻 cs.LG

Designing Active Tether-Net Systems for Space Debris Capture with Graph-Learning-Aided Mixed-Combinatorial Optimization

Feng Liu , Achira Boonrath , Gishnu Madhu , Eleonora M. Botta , Souma Chowdhury This is my paper

Pith reviewed 2026-06-29 14:12 UTC · model grok-4.3

classification 💻 cs.LG
keywords space debris capturetether-net systemsgraph neural networksmixed combinatorial optimizationmaneuverable unitsnonlinear programmingparticle swarm optimization
0
0 comments X

The pith

A graph neural network recommends net and thruster combinations to turn mixed combinatorial design of tether-net capture systems into a standard nonlinear program.

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

The paper shows that the design of active tether-net systems for space debris capture involves a mixed combinatorial nonlinear program with continuous choices like masses and aiming points alongside discrete choices for net connectivity and maneuverable unit components. By representing those discrete choices as nodes in a graph, a graph neural network can be trained to score promising combinations when given the continuous variables as input. The optimization then reduces to solving an ordinary nonlinear program over the continuous variables for each recommended candidate, using a particle swarm optimizer with gradient fine-tuning. On the concurrent design of net morphology, mass and thruster selections, and controller aiming points, this recommender reaches solutions of comparable quality to direct solution of the full mixed problem but with markedly faster convergence.

Core claim

The combinatorial space of net connectivity patterns and component choices is represented as a graph whose nodes are candidate designs; a graph neural network trained to output scores for those nodes, conditioned on the continuous design variables, supplies a short list of candidates that are then optimized as ordinary nonlinear programs, yielding solutions of similar quality to the original mixed combinatorial nonlinear program but with substantially faster convergence.

What carries the argument

Graph neural network that scores candidate design nodes (net connectivity and component choices) given continuous inputs, reducing the mixed combinatorial nonlinear program to repeated solution of a standard nonlinear program.

If this is right

  • The same graph-learning reduction applies to any engineering design task whose discrete decisions form a graph whose nodes can be scored from continuous parameters.
  • Once trained, the graph neural network can be reused across multiple continuous optimization runs without retraining, amortizing the cost of learning the combinatorial structure.
  • Because the approach is solver-agnostic, any improved nonlinear program solver can be swapped in to further accelerate the reduced problems.
  • The method separates the combinatorial recommendation step from the continuous optimization step, allowing independent scaling of each.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If the graph neural network generalizes across different target sizes or orbital conditions, the same trained model could support rapid redesign for new debris objects without repeating the full mixed optimization.
  • Extending the graph to include failure modes or sensor placement choices would let the same recommender jointly optimize capture reliability alongside capture performance.
  • The reduction may also apply to other capture mechanisms such as harpoon or robotic arm systems whose topology choices are likewise graph-structured.

Load-bearing premise

The graph structure on the discrete design choices is rich enough that a trained graph neural network will surface combinations whose optimized continuous solutions are close to the global optimum of the full mixed problem.

What would settle it

Run both the GNN-recommender pipeline and a direct mixed-integer or integer-coded solver on the same tether-net design instance for a fixed computational budget and compare the best objective value achieved; if the direct solver consistently finds a materially better design, the reduction claim is falsified.

Figures

Figures reproduced from arXiv: 2605.29021 by Achira Boonrath, Eleonora M. Botta, Feng Liu, Gishnu Madhu, Souma Chowdhury.

Figure 1
Figure 1. Figure 1: Illustrations of the tether-net system speed. For each i-th MU, solving the optimal minimum-energy control problem with the initial and final states defined above results in the states xr,i(t) = [r T r,i(t), v T r,i(t)]T defined for the entire reference path, consisting of the reference positions rr,i(t) and the reference velocities vr,i(t). To track the reference path, the control force for each i-th MU i… view at source ↗
Figure 2
Figure 2. Figure 2: GNN-aided Optimization Framework the induced node fitness scores. This yields an approximate global ranking over nodes [35]. After this, a greedy filter is applied to select the node with the lowest score. The selected candidate combination, Z ∗ q , is used in the function evaluation with the continuous vector provided by the optimizer, and it is also stored in the cache, so it becomes the new initial cand… view at source ↗
Figure 3
Figure 3. Figure 3: In this research, the population of the MDPSO is set to 100, and the max iteration limit is set to 200. The optimization is terminated early if either of the following conditions is met: i) the population remains entirely infeasible and the total constraint violation of the global best does not decrease (within a tolerance of 10−5 ) over 15 consecutive iterations; or ii) the global best is feasible but its… view at source ↗
Figure 3
Figure 3. Figure 3: Sketch of the GNN-NavCo-aided optimization using population-based optimizer as an example. The graphs are simplified [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Shows the performance on training data used to train [PITH_FULL_IMAGE:figures/full_fig_p010_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: The top figure shows the training and validation loss [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Objective convergence history comparison between the optimization with and without the aid of GNN-NavCo [PITH_FULL_IMAGE:figures/full_fig_p012_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Rendered simulation screenshots of the GNN-NavCo-aided optimized design. The dashed lines are the minimum-energy [PITH_FULL_IMAGE:figures/full_fig_p013_7.png] view at source ↗
read the original abstract

Active tether-net systems are a promising solution for capturing large non-cooperative targets, such as space debris, by deploying a flexible net manipulated by maneuverable units (MUs). However, concurrent systematic explorations of design and control choices of the tether-net system to understand its full potential remain limited, partly due to the complex, constrained, nonlinear optimization problem that it presents -- one that involves a mixture of continuous, integer and categorical variables, with the latter two arising from net connectivity and component choices, respectively. Classical binary encoding methods are often ineffective for solving highly nonlinear and multimodal Mixed Combinatorial Nonlinear Programmings (MCNLPs) in engineering design, while integer coding approaches can introduce spurious relations among combinations. Given the graph-structured characteristics of the combinatorial space, this paper adopts and extends a new graph-learning-aided optimization approach to solve this MCNLP problem. Here, a Graph Neural Network (GNN) is trained to score (as output) and thereof recommend candidate combinations represented as nodes in a graph, with the continuous variable vector portion of a candidate design given as input. As a result, the MCNLP optimization reduces to an NLP, which can be solved using standard solvers. While this reduction approach is agnostic to the choice of the NLP solver, here a state-of-the-art Particle Swarm Optimization (PSO) algorithm with gradient-based fine-tuning is used as the solver. Demonstrated on the problem of concurrently designing the morphology of the net, choice of mass and thrusters in the MUs and aiming points used by the controller of the tether-net system, the GNN-based recommender is shown to provide significantly faster convergence to similar optimal solutions, compared to direct solution of the MCNLP problem.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The manuscript presents a graph neural network (GNN)-aided method to solve mixed combinatorial nonlinear programming (MCNLP) problems arising in the concurrent design of morphology, component choices, and control parameters for active tether-net systems used in space debris capture. By representing the combinatorial space as a graph and training a GNN to score nodes given continuous inputs, the approach reduces the MCNLP to a standard NLP solved using particle swarm optimization (PSO) with gradient-based fine-tuning. The paper demonstrates that this method achieves significantly faster convergence to solutions of comparable quality compared to directly solving the full MCNLP.

Significance. If the empirical claims hold under rigorous validation, the reduction technique could offer an efficient route for graph-structured mixed-variable design optimization in aerospace applications. The solver-agnostic framing is a constructive feature. No machine-checked proofs or parameter-free derivations are present, so significance rests entirely on the quality and reproducibility of the numerical comparisons.

major comments (3)
  1. [§5] §5 (results): the central claim of 'significantly faster convergence to similar optimal solutions' is stated without reported wall-clock times, objective-value tables, convergence curves, baseline solver settings for the direct MCNLP, or any statistical test confirming that the attained optima are statistically indistinguishable.
  2. [§4.2] §4.2 (GNN training): no small-scale exhaustive enumeration or hold-out coverage test is described to verify that the learned node scores, conditioned only on the continuous vector, recover the global-optimum discrete combinations; this directly bears on whether the reduced NLP can miss superior designs.
  3. [§3.1] §3.1 (graph construction): the mapping from net-connectivity and MU-component choices to graph nodes is defined without a formal argument that every feasible discrete combination appears as a node, leaving open the possibility that the GNN recommender operates on an incomplete search space.
minor comments (2)
  1. [Figure 2] Figure 2 caption does not specify the exact continuous-variable inputs fed to the GNN at inference time.
  2. [§5] The PSO hyper-parameters used for both the reduced NLP and the direct MCNLP baseline should be listed in a single table for direct comparison.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We address each major comment below and will revise the manuscript to incorporate the suggested improvements for clarity and rigor.

read point-by-point responses

  1. Referee: [§5] §5 (results): the central claim of 'significantly faster convergence to similar optimal solutions' is stated without reported wall-clock times, objective-value tables, convergence curves, baseline solver settings for the direct MCNLP, or any statistical test confirming that the attained optima are statistically indistinguishable.

    Authors: We agree that the results section would benefit from more comprehensive reporting to support the central claim. In the revised manuscript, we will add wall-clock time measurements, objective-value tables comparing both approaches, convergence curves, explicit details on baseline solver settings for the direct MCNLP, and statistical tests (such as paired t-tests or Wilcoxon signed-rank tests with p-values) to assess whether the attained optima are statistically indistinguishable. These will be included in an expanded Section 5. revision: yes

  2. Referee: [§4.2] §4.2 (GNN training): no small-scale exhaustive enumeration or hold-out coverage test is described to verify that the learned node scores, conditioned only on the continuous vector, recover the global-optimum discrete combinations; this directly bears on whether the reduced NLP can miss superior designs.

    Authors: This is a valid concern regarding the validation of the GNN recommender. While the manuscript focuses on end-to-end performance, we will add a new subsection to §4.2 that includes a small-scale exhaustive enumeration on a reduced instance of the combinatorial space, along with hold-out coverage tests. This will quantify how well the learned scores recover globally optimal discrete combinations and address the risk of missing superior designs. revision: yes

  3. Referee: [§3.1] §3.1 (graph construction): the mapping from net-connectivity and MU-component choices to graph nodes is defined without a formal argument that every feasible discrete combination appears as a node, leaving open the possibility that the GNN recommender operates on an incomplete search space.

    Authors: We will strengthen §3.1 by adding a formal argument and proof sketch demonstrating that the graph construction procedure generates a node for every feasible discrete combination. This will explicitly show that the mapping from net-connectivity patterns and MU-component choices exhaustively covers the feasible space under the problem constraints, ensuring the GNN operates on a complete search space. revision: yes

Circularity Check

0 steps flagged

No significant circularity in derivation or performance claims

full rationale

The paper presents an empirical comparison between a GNN-based reduction of the MCNLP (training a GNN to score graph nodes for discrete choices given continuous inputs, then solving the resulting NLP with PSO) and direct solution of the full MCNLP on the tether-net design task. No equation or claim reduces the reported convergence speed or solution quality to a quantity defined by the GNN outputs themselves; the benchmark remains an external direct solver. The adoption of the graph-learning approach is described as an extension of prior methodology but does not serve as a load-bearing self-citation for the performance result, nor does any step match self-definitional, fitted-input, or ansatz-smuggling patterns. The derivation chain is therefore self-contained against the stated external benchmark.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Review is abstract-only; no explicit free parameters, invented entities, or additional axioms are stated beyond the core modeling assumption.

axioms (1)
  • domain assumption The combinatorial space of net connectivity and component choices can be represented as a graph for GNN scoring.
    This representation is required to reduce the MCNLP to an NLP via GNN recommendations.

pith-pipeline@v0.9.1-grok · 5863 in / 1271 out tokens · 38174 ms · 2026-06-29T14:12:03.071523+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

38 extracted references · 2 canonical work pages

  1. [1]

    Analysis and Consequences of the Irid- ium 33-Cosmos 2251 Collision,

    P. Anz-Meador and J.-C. Liou, “Analysis and Consequences of the Irid- ium 33-Cosmos 2251 Collision,” in38th COSPAR Scientific Assembly, Bremen, Germany, July 2010

    2010

  2. [2]

    A review and comparison of active space debris capturing and removal methods,

    M. Shan, J. Guo, and E. Gill, “A review and comparison of active space debris capturing and removal methods,”Progress in Aerospace Sciences, vol. 80, pp. 18–32, 2015

    2015

  3. [3]

    Robustness and safety of net-based debris capture under deployment and environmental uncertainties,

    A. Boonrath and E. M. Botta, “Robustness and safety of net-based debris capture under deployment and environmental uncertainties,”Journal of Spacecraft and Rockets, pp. 1–17, 2025

    2025

  4. [4]

    Dynamic computation of a tether-net system capturing a space target via discrete elastic rods and an energy-conserving integrator,

    Y . Hou, C. Liu, H. Hu, W. Yang, and J. Shi, “Dynamic computation of a tether-net system capturing a space target via discrete elastic rods and an energy-conserving integrator,”Acta Astronautica, vol. 186, pp. 118–134, 2021

    2021

  5. [5]

    Contact dynamic analysis of tether-net system for space debris capture using incremental potential formulation,

    W. Huang, H. Zou, H. Liu, W. Yang, J. Gao, and Z. Liu, “Contact dynamic analysis of tether-net system for space debris capture using incremental potential formulation,”Advances in Space Research, 2023

    2023

  6. [6]

    Approach modeling and control of an autonomous maneuverable space net,

    Z. Meng, P. Huang, and J. Guo, “Approach modeling and control of an autonomous maneuverable space net,”IEEE Transactions on Aerospace and Electronic Systems, vol. 53, no. 6, pp. 2651–2661, 2017

    2017

  7. [7]

    Impulsive super-twisting sliding mode control for space debris capturing via tethered space net robot,

    Y . Zhao, F. Zhang, P. Huang, and X. Liu, “Impulsive super-twisting sliding mode control for space debris capturing via tethered space net robot,”IEEE transactions on industrial electronics (1982), vol. 67, no. 8, pp. 6874–6882, 2020

    1982

  8. [8]

    Learning constrained corner node trajectories of a tether net system for space debris capture,

    F. Liu, A. Boonrath, P. KrisshnaKumar, E. M. Botta, and S. Chowdhury, “Learning constrained corner node trajectories of a tether net system for space debris capture,” inAIAA AVIATION 2023 Forum, 2023, p. 3920

    2023

  9. [9]

    Learning-aided control of robotic tether-net with maneuverable nodes to capture large space debris,

    A. Boonrath, F. Liu, E. M. Botta, and S. Chowdhury, “Learning-aided control of robotic tether-net with maneuverable nodes to capture large space debris,”IEEE 2024 International Conference on Robotics and Automation, 2024

    2024

  10. [10]

    Multi-debris capture by tethered space net robot via redeployment and assembly,

    W. Zhu, Z. Pang, Z. Du, G. Gao, and Z. H. Zhu, “Multi-debris capture by tethered space net robot via redeployment and assembly,”Journal of Guidance, Control, and Dynamics, pp. 1–18, 2024

    2024

  11. [11]

    Surrogate-aided learning of active tether-net maneuver to capture rotating space debris,

    F. Liu, A. Boonrath, E. M. Botta, and S. Chowdhury, “Surrogate-aided learning of active tether-net maneuver to capture rotating space debris,” IEEE Transactions on Aerospace and Electronic Systems, vol. 62, pp. 630 – 645, 2025

    2025

  12. [12]

    Mixed integer programming modeling for the satellite three-dimensional com- ponent assignment and layout optimization problem,

    Y . Xia, X. Chen, Z. Liu, W. Zhou, W. Yao, and Z. Zhang, “Mixed integer programming modeling for the satellite three-dimensional com- ponent assignment and layout optimization problem,”Chinese Journal of Aeronautics, p. 103415, 2025. 12 TABLE IV: Comparison between optimized variables and objective with and without the aid of GNN-NavCo Continuous Variable...

    2025

  13. [13]

    On the time transformation of mixed integer optimal control problems using a consistent fixed integer control function,

    M. Ringkamp, S. Ober-Bl ¨obaum, and S. Leyendecker, “On the time transformation of mixed integer optimal control problems using a consistent fixed integer control function,”Mathematical programming, vol. 161, no. 1-2, pp. 551–581, 2017

    2017

  14. [14]

    A compre- hensive survey on graph neural networks,

    Z. Wu, S. Pan, F. Chen, G. Long, C. Zhang, and P. S. Yu, “A compre- hensive survey on graph neural networks,”CoRR, vol. abs/1901.00596, 2019

    work page arXiv 1901

  15. [15]

    Attention is all you need,

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin, “Attention is all you need,”Advances in neural information processing systems, vol. 30, 2017

    2017

  16. [16]

    Solve routing problems with a residual edge-graph attention neural network,

    K. Lei, P. Guo, Y . Wang, X. Wu, and W. Zhao, “Solve routing problems with a residual edge-graph attention neural network,”Neurocomputing, vol. 508, pp. 79–98, 2022

    2022

  17. [17]

    Learning scalable policies over graphs for multi-robot task allocation using capsule attention net- works,

    S. Paul, P. Ghassemi, and S. Chowdhury, “Learning scalable policies over graphs for multi-robot task allocation using capsule attention net- works,” in2022 International Conference on Robotics and Automation (ICRA), 2022, pp. 8815–8822

    2022

  18. [18]

    Learning to allocate time-bound and dynamic tasks to multiple robots using covariant attention neural net- works,

    S. Paul and S. Chowdhury, “Learning to allocate time-bound and dynamic tasks to multiple robots using covariant attention neural net- works,”Journal of Computing and Information Science in Engineering, vol. 24, no. 9, p. 091005, 08 2024

    2024

  19. [19]

    Augmenting optimization-based molecular design with graph neural networks,

    S. Zhang, J. S. Campos, C. Feldmann, F. Sandfort, M. Mathea, and R. Misener, “Augmenting optimization-based molecular design with graph neural networks,”Computers & chemical engineering, vol. 186, pp. 108 684–, 2024

    2024

  20. [20]

    Efficient de- sign optimization over mixed-combinatorial spaces enabled by graph- learning,

    F. Liu, S. Chowdhury, A. Boonrath, and E. M. Botta, “Efficient de- sign optimization over mixed-combinatorial spaces enabled by graph- learning,” inASME 2025 International Design Engineering Technical Conferences and Computers and Information in Engineering Confer- ence, 08 2025, p. V03AT03A021

    2025

  21. [21]

    Graph convolutional networks: a comprehensive review,

    S. Zhang, H. Tong, J. Xu, and R. Maciejewski, “Graph convolutional networks: a comprehensive review,”Computational Social Networks, vol. 6, no. 1, p. 11, Nov 2019

    2019

  22. [22]

    Graph attention networks,

    P. Veli ˇckovi´c, G. Cucurull, A. Casanova, A. Romero, P. Li `o, and 13 Y . Bengio, “Graph attention networks,” inInternational Conference on Learning Representations, vol. 6, no. 2. International Conference on Learning Representations, 2018

    2018

  23. [23]

    Graph capsule convolutional neural net- works,

    S. Verma and Z.-L. Zhang, “Graph capsule convolutional neural net- works,” inarXiv, 2018

    2018

  24. [24]

    On the simulation of tether-nets for space debris capture with V ortex Dynamics,

    E. M. Botta, I. Sharf, A. K. Misra, and M. Teichmann, “On the simulation of tether-nets for space debris capture with V ortex Dynamics,” Acta Astronautica, vol. 123, pp. 91–102, 2016

    2016

  25. [25]

    Simulation of tether-nets for capture of space debris and small asteroids,

    E. M. Botta, I. Sharf, and A. K. Misra, “Simulation of tether-nets for capture of space debris and small asteroids,”Acta Astronautica, vol. 155, pp. 448–461, 2019

    2019

  26. [26]

    A mission concept for removing multiple debris using formation-flying slingshot spacecraft,

    A. Boonrath, F. Rossi, I. A. Nesnas, and E. M. Botta, “A mission concept for removing multiple debris using formation-flying slingshot spacecraft,”Acta Astronautica, vol. 237, pp. 443–459, 2025

    2025

  27. [27]

    Active debris removal simulations using spider-web space-nets for kitsat-1 satellite,

    M. Jang, H.-C. Shin, C.-H. Sim, J.-S. Park, and H. Cho, “Active debris removal simulations using spider-web space-nets for kitsat-1 satellite,” International Journal of Aeronautical and Space Sciences, vol. 24, no. 5, pp. 1311–1322, 2023

    2023

  28. [28]

    Concurrent design optimization of tether-net system and actions for reliable space-debris capture,

    C. Zeng, G. R. Hecht, S. Chowdhury, and E. M. Botta, “Concurrent design optimization of tether-net system and actions for reliable space-debris capture,”Journal of Spacecraft and Rockets, vol. 0, no. 0, pp. 1–11, 0. [Online]. Available: https://doi.org/10.2514/1.A35812

    work page doi:10.2514/1.a35812

  29. [29]

    State- of-the-art small spacecraft technology 2024,

    S. V . Weston, C. D. Burkhard, J. M. Stupl, R. L. Ticknor, B. D. Yost, R. A. Austin, P. Galchenko, L. K. Newman, and L. S. Soto, “State- of-the-art small spacecraft technology 2024,” National Aeronautics and Space Administration (NASA), Ames Research Center, Moffett Field, California, Tech. Rep. NASA/TP—20250000142, February 2025. [Online]. Available: htt...

    2024

  30. [30]

    Ecaps 5n hpgp thruster brochure,

    ECAPS, “Ecaps 5n hpgp thruster brochure,” 2023, ac- cessed: 2025-03-20. [Online]. Available: https://satsearch.co/products/ ecaps-5n-hpgp-thruster

    2023

  31. [31]

    5n low throughput (lt) thruster,

    R. S. Systems, “5n low throughput (lt) thruster,” 2023, accessed: 2025-03-20. [Online]. Available: https://www.rubicon.space/product/19/ 5n-lt-thruster

    2023

  32. [32]

    Deployment and Capture Dynamics of Tether-Nets for Active Space Debris Removal,

    E. M. Botta, “Deployment and Capture Dynamics of Tether-Nets for Active Space Debris Removal,” Ph.D. dissertation, McGill University, Montreal, Qu ´ebec, 2017

    2017

  33. [33]

    [Online]

    DuPont,Kevlar® Aramid Fiber Technical Guide, 2023, accessed: 2025- 03-20. [Online]. Available: https://www.dupont.com/products/kevlar

    2023

  34. [34]

    A gentle introduction to graph neural networks,

    B. Sanchez-Lengeling, E. Reif, A. Pearce, and A. B. Wiltschko, “A gentle introduction to graph neural networks,”Distill, 2021, https://distill.pub/2021/gnn-intro

    2021

  35. [35]

    Graph laplacians and least squares on graphs,

    A. N. Hirani, K. Kalyanaraman, and S. Watts, “Graph laplacians and least squares on graphs,” in2015 IEEE International Parallel and Distributed Processing Symposium Workshop, 2015, pp. 812–821

    2015

  36. [36]

    A mixed- discrete particle swarm optimization algorithm with explicit diversity- preservation,

    S. Chowdhury, W. Tong, A. Messac, and J. Zhang, “A mixed- discrete particle swarm optimization algorithm with explicit diversity- preservation,”Structural and Multidisciplinary Optimization, vol. 47, pp. 367–388, 2013

    2013

  37. [37]

    A multi-objective mixed- discrete particle swarm optimization with multi-domain diversity preser- vation,

    W. Tong, S. Chowdhury, and A. Messac, “A multi-objective mixed- discrete particle swarm optimization with multi-domain diversity preser- vation,”Structural and Multidisciplinary Optimization, vol. 53, pp. 471– 488, 2016

    2016

  38. [38]

    Data and video repository for the paper – designing active tether-net system for space debris capture with graph learning aided mixed-combinatorial optimization,

    F. Liu, A. Boonrath, G. Madhu, E. M. Botta, and S. Chowdhury, “Data and video repository for the paper – designing active tether-net system for space debris capture with graph learning aided mixed-combinatorial optimization,” https://github.com/adamslab-ub/GNN-ReCo-Benchmark/ tree/GNN NavCo, 2026. 14

    2026