pith. sign in

arxiv: 2605.20644 · v1 · pith:KQSJ7PWWnew · submitted 2026-05-20 · 💻 cs.LG · cs.AI· cs.RO

Design for Manufacturing: A Manufacturability Knowledge-Integrated Reinforcement Learning Framework for Free-Form Pipe Routing in Aeroengines

Caicheng Wang , Zili Wang , Shuyou Zhang , Yongzhe Xiang , Zheyi Li , Liangyou Li , Jianrong Tan This is my paper

Pith reviewed 2026-05-21 06:17 UTC · model grok-4.3

classification 💻 cs.LG cs.AIcs.RO
keywords free-form pipe routingreinforcement learningmanufacturability constraintsFrenet framecurvature torsionaeroengine designdesign for manufacturingproximal policy optimization
0
0 comments X

The pith

A reinforcement learning framework embeds manufacturing limits on curvature and torsion to generate fabricable free-form pipe routes for aeroengines directly from design.

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

The paper proposes the FPRO framework to link pipe routing design in aeroengines with actual fabrication constraints. It models paths in the Frenet frame as curvature and torsion profiles interpolated by cubic Hermite functions, then restricts those profiles to the ranges a six-axis bending machine can achieve. Optimization proceeds through proximal policy optimization with a stage-guided reward that balances alignment, length, collision avoidance, and manufacturability. If the approach holds, designers could produce routes that require no separate trial-and-error manufacturing checks, cutting iteration cycles in complex engine assembly.

Core claim

FPRO formulates routing as a boundary-value problem in the Frenet frame, generates curvature and torsion profiles through cubic Hermite interpolation, enforces manufacturability by bounding those profiles to the permissible ranges of a six-axis free-bending machine, optimizes the profiles with proximal policy optimization using stochastic exploration and a staged reward, and supplies a direct mapping from the resulting path to machine motion trajectories for immediate fabrication.

What carries the argument

Curvature and torsion profiles in the Frenet frame, bounded to machine-derived limits and optimized by the proximal policy optimization algorithm.

If this is right

  • The method produces collision-free paths with smoother geometric profiles than Cartesian-based routing techniques.
  • It reaches terminal alignment, shorter path length, better obstacle avoidance, and higher manufacturability scores with faster convergence than existing reinforcement-learning baselines.
  • Optimized paths translate directly into bending-die trajectories that the six-axis machine can execute without additional adjustment.
  • Real-world trials confirm that the fabricated pipe matches the digital geometry within close tolerance.

Where Pith is reading between the lines

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

  • The same curvature-torsion constraint pattern could be adapted to route other conduits or harnesses inside tightly packed mechanical systems.
  • Closing the loop from design to machine trajectory may allow incremental updates when new manufacturing data arrives from the shop floor.
  • Extending the reward function to include additional machine-specific dynamics could further reduce the gap between simulated and physical outcomes.

Load-bearing premise

All manufacturability requirements for the pipe are captured by simple upper and lower bounds on curvature and torsion derived from the six-axis bending machine.

What would settle it

A generated path that satisfies the curvature and torsion bounds but cannot be bent on the physical six-axis machine without defects or deviates measurably from the digital design after fabrication.

read the original abstract

Design for manufacturing plays a critical role in advanced aeroengine development, where complex components necessitate careful consideration of manufacturability. However, current practices in pipe routing remain largely decoupled from down-stream manufacturing, leading to labor-intensive, trial-and-error iterations to achieve manufacturable designs. To address this problem, this study proposes the Frenet-based pipe routing optimization (FPRO) framework, a manufacturability knowledge-integrated reinforcement learning approach for free-form pipe design in aeroengines. FPRO formulates the routing problem as a boundary value problem in the Frenet frame. In this framework, the pipe path is represented by curvature and torsion profiles, which are generated using cubic Hermite interpolation. To integrate design and manufacturing, domain-specific manufacturing knowledge is embedded as constraints on the permissible ranges of curvature and torsion. The path optimization is performed using the proximal policy optimization algorithm with stochastic exploration and a stage-guided reward mechanism. A unified mapping formulation then translates the optimized path into motion trajectories for the bending die, enabling direct fabrication on a six-axis free-bending machine. Experimental results demonstrate that FPRO consistently generates collision-free, manufacturable paths with smoother geometric profiles compared to Cartesian-based methods. It also achieves faster convergence and superior performance in terminal alignment, path length, obstacle avoidance, and manufacturability compared to state-of-the-art reinforcement learning baselines. Real-world validation confirms the close geometric correspondence between the manufactured pipe and its digital design, validating the practical feasibility of FPRO.

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

2 major / 2 minor

Summary. The manuscript proposes the Frenet-based pipe routing optimization (FPRO) framework, which integrates manufacturability knowledge into a reinforcement learning approach for free-form pipe routing in aeroengines. The method formulates the routing task as a boundary-value problem in the Frenet frame, represents paths via curvature and torsion profiles generated by cubic Hermite interpolation, enforces hard bounds on these quantities derived from a six-axis free-bending machine, optimizes with PPO augmented by a stage-guided reward, and provides a mapping from the optimized path to die motion trajectories for direct fabrication. The central claims are that FPRO produces collision-free, smoother, and more manufacturable paths than Cartesian-based methods and state-of-the-art RL baselines, with faster convergence and superior terminal alignment, path length, obstacle avoidance, and manufacturability metrics, plus real-world validation demonstrating close geometric correspondence between the digital design and the physically manufactured pipe.

Significance. If the quantitative claims are substantiated, the work offers a concrete advance in design-for-manufacturing by embedding machine-specific constraints directly into the RL loop rather than post-processing. The Frenet-frame representation and stage-guided reward address geometric and sequential aspects of pipe routing that are relevant to aeroengine applications. The attempt at real-world fabrication validation is a positive step toward demonstrating practical utility, though its evidentiary weight depends on the details of the comparison data.

major comments (2)
  1. [Abstract and Experimental Results] Abstract and Experimental Results section: the claims of superior performance in terminal alignment, path length, obstacle avoidance, manufacturability, and faster convergence, together with real-world validation, are stated without any numerical values, baseline specifications, error bars, number of trials, or statistical tests. This absence prevents assessment of effect sizes and reproducibility of the reported advantages.
  2. [Manufacturing Knowledge Integration] Manufacturing Knowledge Integration (description of constraint embedding): the framework treats manufacturability as fully captured by hard bounds on curvature and torsion ranges from the six-axis machine. No analysis or experimental evidence is provided to rule out unmodeled effects such as springback, wall thinning, or die-contact dynamics that could produce fabrication failures even when curvature and torsion remain inside the stated limits; this assumption is load-bearing for the claim that the optimized paths are directly fabricable without further iteration.
minor comments (2)
  1. [Methods] The cubic Hermite interpolation formula for the curvature and torsion profiles should be written explicitly as an equation, with clear definitions of the control points and boundary conditions.
  2. [Figures] Figure captions comparing FPRO and baseline paths should include quantitative annotations (e.g., maximum curvature, total length, or collision clearance) rather than relying solely on visual inspection.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed feedback on our manuscript. We have carefully reviewed the major comments and provide point-by-point responses below. Revisions have been made to improve the clarity and substantiation of our claims while honestly acknowledging the scope of our current manufacturability modeling.

read point-by-point responses

  1. Referee: [Abstract and Experimental Results] Abstract and Experimental Results section: the claims of superior performance in terminal alignment, path length, obstacle avoidance, manufacturability, and faster convergence, together with real-world validation, are stated without any numerical values, baseline specifications, error bars, number of trials, or statistical tests. This absence prevents assessment of effect sizes and reproducibility of the reported advantages.

    Authors: We agree that explicit numerical values, variability measures, and statistical details would strengthen the abstract and facilitate assessment of the reported advantages. Although the Experimental Results section contains comparative tables and figures with quantitative metrics, we have revised the abstract to include representative performance values (such as mean improvements in path length, terminal alignment error, and convergence iterations) drawn from our experiments, along with indications of the number of trials performed. We have also added error bars to relevant plots and included details on baseline configurations and statistical comparisons in the Experimental Results section of the revised manuscript. revision: yes

  2. Referee: [Manufacturing Knowledge Integration] Manufacturing Knowledge Integration (description of constraint embedding): the framework treats manufacturability as fully captured by hard bounds on curvature and torsion ranges from the six-axis machine. No analysis or experimental evidence is provided to rule out unmodeled effects such as springback, wall thinning, or die-contact dynamics that could produce fabrication failures even when curvature and torsion remain inside the stated limits; this assumption is load-bearing for the claim that the optimized paths are directly fabricable without further iteration.

    Authors: We thank the referee for highlighting this important aspect of our modeling assumptions. The FPRO framework integrates manufacturability by enforcing hard bounds on curvature and torsion that are directly derived from the specifications of the six-axis free-bending machine, enabling a direct mapping to die motion trajectories. While dedicated isolation experiments on effects such as springback or wall thinning are not included in the current study, the real-world fabrication validation shows successful production of pipes with close geometric match to the optimized designs and no reported failures under the tested conditions. We acknowledge that this empirical support does not comprehensively exclude all unmodeled process effects. In the revised manuscript we have added an explicit discussion of these limitations in the concluding section and outlined directions for future incorporation of higher-fidelity manufacturing simulations. revision: partial

Circularity Check

0 steps flagged

Minor self-citation not load-bearing; derivation uses standard RL with independent constraints and experimental validation

full rationale

The FPRO framework formulates pipe routing as a boundary-value problem in the Frenet frame, represents paths via curvature/torsion profiles generated by cubic Hermite interpolation, enforces manufacturability as hard bounds on those profiles, and optimizes with off-the-shelf PPO plus a custom stage-guided reward. Reported performance gains (collision-free paths, terminal alignment, path length, etc.) and real-world geometric correspondence are empirical outcomes of this optimization under the stated constraints, not quantities defined by or fitted to the same evaluation data. No equations reduce metrics to tautological inputs, and the central manufacturability claim rests on external machine-derived bounds plus physical validation rather than self-referential definitions or self-citation chains.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 0 invented entities

The abstract supplies limited technical detail; the framework rests on geometric representation assumptions and domain-derived bounds whose precise numerical values or fitting procedures are not stated.

free parameters (2)
  • curvature and torsion permissible ranges
    Bounds derived from manufacturing knowledge; exact values or selection procedure not provided in abstract.
  • stage-guided reward weights
    Hyperparameters controlling the multi-stage reward mechanism; not quantified in abstract.
axioms (1)
  • domain assumption Pipe paths can be represented accurately and completely by curvature and torsion profiles generated via cubic Hermite interpolation in the Frenet frame.
    Invoked to reformulate the routing problem as a boundary-value problem.

pith-pipeline@v0.9.0 · 5822 in / 1538 out tokens · 59539 ms · 2026-05-21T06:17:09.077201+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

41 extracted references · 41 canonical work pages · 1 internal anchor

  1. [1]

    Boschetto, L

    A. Boschetto, L. Bottini, Design for manufacturing of surfaces to improve accuracy in Fused Deposition Modeling, Rob. Comput. Integr. Manuf. 37 (2016) 103-114. https://doi.org/10.1016/j.rcim.2015.07.005

    work page doi:10.1016/j.rcim.2015.07.005 2016

  2. [2]

    Kuo, S.H

    T.-C. Kuo, S.H. Huang, H. -C. Zhang, Design for manufacture and design for ‘X’: concepts, applications, and perspectives, Comput. Ind. Eng. 41 (2001) 241-260. https://doi.org/10.1016/S0360-8352(01)00045-6

    work page doi:10.1016/s0360-8352(01)00045-6 2001

  3. [3]

    Z. Wang, X. Yan, S. Melkote, D. Rosen, McGAN: Generating manufacturable designs by embedding manufacturing rules into conditional generative adversarial network, Adv. Eng. Inf. 64 (2025) 103074. https://doi.org/10.1016/j.aei.2024.103074

    work page doi:10.1016/j.aei.2024.103074 2025

  4. [4]

    Yin, L.D

    Y.H. Yin, L.D. Xu, Z.M. Bi, H. Chen, C. Zhou, A Novel Human-Machine Collaborative Interface for Aero-Engine Pipe Routing, IEEE Trans. Ind. Inf. 9 (2013) 2187-2199. https://doi.org/10.1109/tii.2013.2257805

    work page doi:10.1109/tii.2013.2257805 2013

  5. [5]

    C. Wang, Z. Wang, S. Zhang, Y. Xiang, Z. Li, J. Tan, Self -Learning-Based Optimization for Free -form Pipe Routing in Aeroengine with Dynamic Design Environment, (2025). https://doi.org/10.48550/arXiv.2504.03669

    work page doi:10.48550/arxiv.2504.03669 2025

  6. [6]

    C. Liu, L. Wu, G.X. Li, W.S. Xiao, L.P. Tan, D.P. Xu, J.J. Guo, AI-based 3D pipe automation layout with enhanced ant colony optimization algorithm, Autom. Constr. 167 (2024) 22. https://doi.org/10.1016/j.autcon.2024.105689

    work page doi:10.1016/j.autcon.2024.105689 2024

  7. [7]

    C. Liu, L. Wu, G.X. Li, H. Zhang, W.S. Xiao, D.P. Xu, J.J. Guo, W.T. Li, Improved multi-search strategy A* algorithm to solve three-dimensional pipe routing design, Expert Syst. Appl. 240 (2024) 26. https://doi.org/10.1016/j.eswa.2023.122313

    work page doi:10.1016/j.eswa.2023.122313 2024

  8. [8]

    Lin, X.-y

    Y. Lin, X.-y. Bian, Z.-r. Dong, A discrete hybrid algorithm based on Differential Evolution and Cuckoo Search for optimizing the layout of ship pipe route, Ocean Eng. 261 (2022) 112164. https://doi.org/10.1016/j.oceaneng.2022.112164

    work page doi:10.1016/j.oceaneng.2022.112164 2022

  9. [9]

    Y. Kim, K. Lee, B. Nam, Y. Han, Application of reinforcement learning based on curriculum learning for the pipe auto-routing of ships, J. Comput. Des. Eng. 10 (2023) 318-328. https://doi.org/10.1093/jcde/qwad001

    work page doi:10.1093/jcde/qwad001 2023

  10. [10]

    Y. Lin, Q. Zhang, A multi -objective cooperative particle swarm optimization based on hybrid dimensions fo r ship pipe route design, Ocean Eng. 280 (2023) 114772. https://doi.org/10.1016/j.oceaneng.2023.114772

    work page doi:10.1016/j.oceaneng.2023.114772 2023

  11. [11]

    Zhang, Y

    Q. Zhang, Y. Lin, Integrating multi -agent reinforcement learning and 3D A* search for facility layout problem considering connector-assembly, J. Intell. Manuf. 35 (2024) 3393-3418. https://doi.org/10.1007/s10845-023-02209-x

    work page doi:10.1007/s10845-023-02209-x 2024

  12. [12]

    C. Wang, Z. Wang, S. Zhang, Y. Lin, Y. Xiang, L. Wang, J. Tan, Cross-forming-process transfer enabled graph neural networks for accurate axial -forming prediction in metal tu be bending, Appl. Soft Comput. 175 (2025) 113114. https://doi.org/10.1016/j.asoc.2025.113114

    work page doi:10.1016/j.asoc.2025.113114 2025

  13. [13]

    J. Li, Z. Wang, S. Zhang, Y. Lin, L. Jiang, J. Tan, Task incremental learning -driven Digital-Twin predictive modeling for customized metal forming product ma nufacturing process, Rob. Comput. Integr. Manuf. 85 (2024) 102647. https://doi.org/10.1016/j.rcim.2023.102647

    work page doi:10.1016/j.rcim.2023.102647 2024

  14. [14]

    C. Sun, Z. Wang, S. Zhang, X. Liu, L. Wang, J. Tan, Toward axial accuracy prediction and optimization of metal tube bending forming: A novel GRU-integrated Pb -NSGA-III optimization framework, Eng. Appl. Artif. Intell. 114 (2022) 105193. https://doi.org/10.1016/j.engappai.2022.105193

    work page doi:10.1016/j.engappai.2022.105193 2022

  15. [15]

    Z. Wang, J. Li, Y. Yuan, S. Zhang, W. Hu, J. Ma, J. Tan, Digital -twin-enabled online wrinkling monitoring of metal tube bending manufacturing: A multi-fidelity approach using forward-convolution-GAN, Appl. Soft Comput. 171 (2025) 112684. https://doi.org/10.1016/j.asoc.2024.112684

    work page doi:10.1016/j.asoc.2024.112684 2025

  16. [16]

    Z. Wang, Y. Xiang, S. Zhang, X. Liu, J. Ma, J. Tan, L. Wang, Physics -informed springback prediction of 3D aircraft tubes with six-axis free-bending manufacturing, Aerosp. Sci. Technol. 147 (2024) 109022. https://doi.org/10.1016/j.ast.2024.109022

    work page doi:10.1016/j.ast.2024.109022 2024

  17. [17]

    J. Li, Z. Wang, S. Zhang, J. Ji, Y. Xiang, D. Wang, J. Tan, Digital-Twin virtual model real-time construction via spatio-temporal cascade reconstruction for full-field plastic deformation monitoring in metal tube bending manufacturing, Rob. Comput. Integr. Manuf. 91 (2025) 102860. https://doi.org/10.1016/j.rcim.2024.102860

    work page doi:10.1016/j.rcim.2024.102860 2025

  18. [18]

    Z. Y an, J. Yan, Y. Wu, S. Cai, H. Wang, A novel reinforcement learning based tuna swarm optimization algorithm for autonomous underwater vehicle path planning, Math. Comput. Simul. 209 (2023) 55 -86. https://doi.org/10.1016/j.matcom.2023.02.003

    work page doi:10.1016/j.matcom.2023.02.003 2023

  19. [19]

    H. Feng, Q. Hu, Z. Zhao, X. Feng, Smooth path planning under maximum curvature constraints for autonomous underwater vehicles based on rapidly -exploring random tree star with B -spline curves, Eng. Appl. Artif. Intell. 133 (2024) 108583. https://doi.org/10.1016/j.engappai.2024.108583

    work page doi:10.1016/j.engappai.2024.108583 2024

  20. [20]

    G. Hu, B. Du, G. Wei, HG -SMA: hierarchical guided slime mould algorithm for smooth path planning, Artif. Intell. Rev. 56 (2023) 9267-9327. https://doi.org/10.1007/s10462-023-10398-3

    work page doi:10.1007/s10462-023-10398-3 2023

  21. [21]

    B. Song, Z. Wang, L. Zou, An improved PSO al gorithm for smooth path planning of mobile robots using continuous high - degree Bezier curve, Appl. Soft Comput. 100 (2021) 106960. https://doi.org/10.1016/j.asoc.2020.106960

    work page doi:10.1016/j.asoc.2020.106960 2021

  22. [22]

    X. Wu, K. Hirota, Z. Jia, Y. Ji, K. Zhao, Y. Dai, Ameliorated equilibrium op timizer with application in smooth path planning oriented unmanned ground vehicle, Knowl.-Based Syst. 260 (2023) 110148. https://doi.org/10.1016/j.knosys.2022.110148

    work page doi:10.1016/j.knosys.2022.110148 2023

  23. [23]

    Upadhyay, A

    S. Upadhyay, A. Ratnoo, Continuous-Curvature Path Planning With Obstacle Avoidance Using Four Parameter Logistic Curves, IEEE Rob. Autom. Lett. 1 (2016) 609-616. https://doi.org/10.1109/LRA.2016.2521165

    work page doi:10.1109/lra.2016.2521165 2016

  24. [24]

    Mazaheri, S

    H. Mazaheri, S. Goli, A. Nourollah, A survey of 3D Space Path -Planning Methods and Algorithms, ACM Comput. Surv. 57 (2025) 32. https://doi.org/10.1145/3673896

    work page doi:10.1145/3673896 2025

  25. [25]

    Kobayashi, Differential Geometry of Curves and Surfaces, 1 ed., Springer Singapore, 2019

    S. Kobayashi, Differential Geometry of Curves and Surfaces, 1 ed., Springer Singapore, 2019

    work page 2019

  26. [26]

    Muzzini, N

    F. Muzzini, N. Capodieci, F. Ramanzin, P. Burgio, GPU implementation of the Frenet Path Planner for embedded autonomous systems: A case study in the F1tenth scenario, J. Syst. Archit. 154 (2024) 103239. https://doi.org/10.1016/j.sysarc.2024.103239

    work page doi:10.1016/j.sysarc.2024.103239 2024

  27. [27]

    Y. Sun, D. Ren, S. Lian, S. Fu, X. Teng, M. Fan, Robust Path Planner for Autonomous Vehicles on Roads With Large Curvature, IEEE Rob. Autom. Lett. 7 (2022) 2503-2510. https://doi.org/10.1109/LRA.2022.3143294

    work page doi:10.1109/lra.2022.3143294 2022

  28. [28]

    Proximal Policy Optimization Algorithms

    J. Schulman, F. Wolski, P. Dhariwal, A. Radford, O. Klimov, Proximal policy optimization algorithms, (2017). https://doi.org/10.48550/arXiv.1707.06347

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1707.06347 2017

  29. [29]

    J. Yang, J. Huo, M. Xi, J. He, Z. Li, H.H. Song, A Time-Saving Path Planning Scheme for Autonomous Underwater Vehicles With Complex Underwater Conditions, IEEE Internet Things J. 10 (2023) 1001 -1013. https://doi.org/10.1109/JIOT.2022.3205685

    work page doi:10.1109/jiot.2022.3205685 2023

  30. [30]

    M. Xi, J. Yang, J. Wen, Z. Li, W. Lu, X. Gao, An Information-Assisted Deep Reinforcement Learning Path Planning Scheme for Dynamic and Unknown Underwater Environment, IEEE Tran s. Neural Networks Learn. Syst. (2023) 1 -12. https://doi.org/10.1109/TNNLS.2023.3332172

    work page doi:10.1109/tnnls.2023.3332172 2023

  31. [31]

    J. Hao, T. Yang, H. Tang, C. Bai, J. Liu, Z. Meng, P. Liu, Z. Wang, Exploration in Deep Reinforcement Learning: From Single- Agent to Multiagent Domain, IEEE Trans . Neural Networks Learn. Syst. 35 (2024) 8762 -8782. https://doi.org/10.1109/TNNLS.2023.3236361

    work page doi:10.1109/tnnls.2023.3236361 2024

  32. [32]

    Ladosz, L

    P. Ladosz, L. Weng, M. Kim, H. Oh, Exploration in deep reinforcement learning: A survey, Inf. Fusion. 85 (2022) 1 -22. https://doi.org/10.1016/j.inffus.2022.03.003

    work page doi:10.1016/j.inffus.2022.03.003 2022

  33. [33]

    Rahman, Y

    M.M. Rahman, Y. Xue, Robust policy optimization in deep reinforcement learning, (2022). https://doi.org/10.48550/arXiv.2212.07536

    work page doi:10.48550/arxiv.2212.07536 2022

  34. [34]

    Akiba, S

    T. Akiba, S. Sano, T. Yanase, T. Ohta, M. Koyama, Optuna: A Next-generation Hyperparameter Optimization Framework, in: Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, Association for Computing Machinery, Anchorage, AK, USA, 2019, pp. 2623–2631. https://doi.org/10.1145/3292500.3330701

    work page doi:10.1145/3292500.3330701 2019

  35. [35]

    Haarnoja, A

    T. Haarnoja, A. Zhou , P. Abbeel, S. Levine, Soft Actor -Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor, in: D. Jennifer, K. Andreas (Eds.) Proceedings of the 35th International Conference on Machine Learning, PMLR, Proceedings of Machine Learning Research, 2018, pp. 1861--1870

    work page 2018

  36. [36]

    Fujimoto, H

    S. Fujimoto, H. Hoof, D. Meger, Addressing Function Approximation Error in Actor -Critic Methods, in: D. Jennifer, K. Andreas (Eds.) Proceedings of the 35th International Conference on Machine Learning, PMLR, Proceedings of Machine Learning Research, 2018, pp. 1587--1596

    work page 2018

  37. [37]

    Cobbe, J

    K.W. Cobbe, J. Hilton, O. Klimov, J. Schulman, Phasic Policy Gradient, in: M. Marina, Z. Tong (Eds.) Proceedings of the 38th International Conference on Machine Learning, PMLR, Proceedings of Machine Learning Research, 2021, pp. 2020--2027

    work page 2021

  38. [38]

    Huang, R.F.J

    S. Huang, R.F.J. Dossa, C. Ye, J. Braga, D. Chakraborty, K. Mehta, J.G.M. Araú jo, CleanRL: high -quality single -file implementations of deep reinforcement learning algorithms, Journal of Machine L earning Research. 23 (2022) Article 274. https://dl.acm.org/doi/10.5555/3586589.3586863

    work page doi:10.5555/3586589.3586863 2022

  39. [39]

    L. Wang, Z. Wang, S. Zhang, Y. Lin, M. Fu, C. Sun, Spiral metal tube multi-roller bending springback prediction model based on dynamic deformation enhancement ana lysis, Int. J. Solids Struct. 254 -255 (2022) 111940. https://doi.org/10.1016/j.ijsolstr.2022.111940

    work page doi:10.1016/j.ijsolstr.2022.111940 2022

  40. [40]

    Y. Tao, A. Both, R.I. Silveira, K. Buchin, S. Sijben, R.S. Purves, P. Laube, D. Peng, K. Toohey, M. Duckham, A comparative analysis of trajectory simi larity measures, GISci. Remote Sens. 58 (2021) 643 -669. https://doi.org/10.1080/15481603.2021.1908927

    work page doi:10.1080/15481603.2021.1908927 2021

  41. [41]

    Z. Wang, C. Wang, S. Zhang, L. Qiu, Y. Lin, J. Tan, C. Sun, Towards high-accuracy axial springback: Mesh-based simulation of metal tube bending via geometry/process-integrated graph neural networks, Expert Syst. Appl. 255 (2024) 124577. https://doi.org/10.1016/j.eswa.2024.124577

    work page doi:10.1016/j.eswa.2024.124577 2024