pith. sign in

arxiv: 2502.08170 · v2 · submitted 2025-02-12 · 🪐 quant-ph · cs.SY· eess.SY

Learning-Based Design of LQG Controllers in Quantum Coherent Feedback

Chunxiang Song , Yanan Liu , Guofeng Zhang , Huadong Mo , Daoyi Dong This is my paper

Pith reviewed 2026-05-23 03:28 UTC · model grok-4.3

classification 🪐 quant-ph cs.SYeess.SY
keywords differential evolutionLQG controlquantum coherent feedbackphysical realizabilityquantum optical systemcontroller designperformance optimization
0
0 comments X

The pith

A differential evolution algorithm with quantum-specific modules designs LQG controllers that reach lower performance indices while satisfying physical realizability.

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

The paper develops a differential evolution algorithm modified for LQG controller design in quantum coherent feedback. Four added modules handle the physical constraints that quantum systems place on realizable controllers. The method is demonstrated on a quantum optical system, where it produces three controllers with lower LQG costs than earlier designs. A reader would care because the approach automates the search for controllers that can actually be built in quantum hardware. The work centers on balancing optimization performance against the requirement that the controller must obey quantum physical laws.

Core claim

By incorporating relaxed feasibility rules, a scheduled penalty function, adaptive search range adjustment, and bet-and-run initialization into the differential evolution framework, the algorithm finds LQG controller parameters for quantum systems that deliver lower performance indices than existing approaches while ensuring the controllers meet physical realizability constraints.

What carries the argument

Differential evolution algorithm augmented with relaxed feasibility rules, scheduled penalty function, adaptive search range adjustment, and bet-and-run initialization to enforce quantum physical realizability.

If this is right

  • Three controllers with different plant configurations are obtained for the tested quantum optical system.
  • The designs achieve lower LQG performance indices than prior methods.
  • All produced controllers satisfy physical realizability constraints required for practical implementation.
  • The same algorithmic structure applies to performance optimization of other linear quantum systems under realizability constraints.

Where Pith is reading between the lines

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

  • The method may generalize to controller design problems in other quantum platforms beyond the optical case examined.
  • It could reduce reliance on manual parameter tuning when scaling to higher-dimensional quantum systems.
  • Combining the search strategy with gradient-based local refinement might further lower the achieved indices.
  • Experimental implementation of the resulting controllers on actual quantum hardware would test whether the simulated performance gains hold in practice.

Load-bearing premise

The four added modules for handling physical realizability in the search process do not exclude better controller solutions that a different method might find.

What would settle it

A controller obtained by another optimization technique on the same quantum optical system that achieves a strictly lower LQG index and still satisfies the physical realizability conditions would falsify the superiority claim.

Figures

Figures reproduced from arXiv: 2502.08170 by Chunxiang Song, Daoyi Dong, Guofeng Zhang, Huadong Mo, Yanan Liu.

Figure 1
Figure 1. Figure 1: The workflow of the proposed algorithm. It starts with a “bet and run” initialization to select a promising solution for further refinement. In each [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The penalty factor ϱp increases with the step number p. In the early stages, only a small penalty is applied to infeasible solutions, with an emphasis on improving system performance. In the later stages, the penalty for violating constraints increases rapidly, aiming to find feasible solutions near the region of solutions with good performance. For the p round, the DE algorithm runs to completion, retaini… view at source ↗
Figure 3
Figure 3. Figure 3: Search Range. (a) Fixed search range; (b) Adaptive search range. [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Applications (a) Indirect Coupling: The plant and controller interact indirectly through a shared external medium, such as a quantum field, without [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Enhanced Differential Evolution for Indirect Coupling Model in [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
read the original abstract

In this paper, we propose a differential evolution (DE) algorithm specifically tailored for the design of Linear-Quadratic-Gaussian (LQG) controllers in quantum systems. Building upon the foundational DE framework, the algorithm incorporates specialized modules, including relaxed feasibility rules, a scheduled penalty function, adaptive search range adjustment, and the ``bet-and-run'' initialization strategy. These enhancements improve the algorithm's exploration and exploitation capabilities while addressing the unique physical realizability requirements of quantum systems. The proposed method is applied to a quantum optical system, where three distinct controllers with varying configurations relative to the plant are designed. The resulting controllers demonstrate superior performance, achieving lower LQG performance indices compared to existing approaches. Additionally, the algorithm ensures that the designs comply with physical realizability constraints, guaranteeing compatibility with practical quantum platforms. The proposed approach holds significant potential for application to other linear quantum systems in performance optimization tasks subject to physically feasible constraints.

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 a differential evolution algorithm augmented with four specialized modules (relaxed feasibility rules, scheduled penalty function, adaptive search range adjustment, and bet-and-run initialization) for designing LQG controllers in quantum coherent feedback systems. These enhancements are intended to improve exploration while enforcing physical realizability constraints. The method is applied to a quantum optical system to synthesize three controllers with different plant configurations; the authors claim these achieve lower LQG performance indices than prior approaches while remaining physically realizable.

Significance. If the performance and realizability claims are rigorously verified, the work would supply a practical, domain-adapted optimizer for quantum coherent control, addressing a recurring bottleneck in translating LQG designs to experimental platforms. The explicit incorporation of quantum-specific constraints into an evolutionary framework is a constructive contribution that could extend to other linear quantum systems.

major comments (2)
  1. [Abstract and §4] Abstract and §4 (numerical example): the central claim that the synthesized controllers achieve lower LQG performance indices than existing approaches is stated without tabulated index values, baseline comparisons, error bars, or statistical tests. This prevents assessment of whether the reported improvement is meaningful or reproducible.
  2. [§3] §3 (algorithm): the four custom modules are presented as necessary to navigate the realizable set, yet no ablation experiments (disabling one module at a time), no head-to-head comparison against standard constrained DE or other optimizers (e.g., interior-point or genetic algorithms with exact feasibility checks), and no post-optimization verification that the final solutions satisfy the physical realizability conditions exactly (beyond the penalty term) are provided. These omissions leave open the possibility that the modules either admit non-realizable points or exclude lower-index feasible designs.
minor comments (2)
  1. [§2] Notation for the LQG cost functional and the physical realizability constraints should be introduced with explicit equations early in §2 to improve readability for readers outside the immediate subfield.
  2. [§4] Figure captions for the quantum optical system diagrams should explicitly label which blocks correspond to the plant, controller, and coherent feedback paths.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments. We address each major point below and commit to revisions that strengthen the empirical validation of our claims.

read point-by-point responses

  1. Referee: [Abstract and §4] Abstract and §4 (numerical example): the central claim that the synthesized controllers achieve lower LQG performance indices than existing approaches is stated without tabulated index values, baseline comparisons, error bars, or statistical tests. This prevents assessment of whether the reported improvement is meaningful or reproducible.

    Authors: We agree that the current presentation lacks the quantitative detail needed for rigorous assessment. In the revised manuscript we will insert a dedicated comparison table in §4 that reports the exact LQG performance indices for all three controller configurations, the corresponding baseline values from the referenced prior methods, means and standard deviations obtained from 20 independent runs of the optimizer, and the results of paired statistical tests (t-tests with p-values) against each baseline. These additions will make the performance advantage both transparent and reproducible. revision: yes

  2. Referee: [§3] §3 (algorithm): the four custom modules are presented as necessary to navigate the realizable set, yet no ablation experiments (disabling one module at a time), no head-to-head comparison against standard constrained DE or other optimizers (e.g., interior-point or genetic algorithms with exact feasibility checks), and no post-optimization verification that the final solutions satisfy the physical realizability conditions exactly (beyond the penalty term) are provided. These omissions leave open the possibility that the modules either admit non-realizable points or exclude lower-index feasible designs.

    Authors: We accept that ablation studies and external benchmarks are required to substantiate the necessity of the four modules. The revised version will include (i) four ablation runs in which each module is disabled individually while keeping the others active, (ii) direct comparisons against a standard constrained differential-evolution implementation and against a genetic algorithm that enforces feasibility via exact projection, and (iii) an explicit post-optimization verification step that recomputes the physical-realizability equalities on the final parameter vectors and reports the residual norms. These experiments will be presented in an expanded §3 and will directly address concerns about possible admission of non-realizable points or exclusion of superior feasible designs. revision: yes

Circularity Check

0 steps flagged

No significant circularity; optimization procedure applied to external objective

full rationale

The paper presents a differential evolution algorithm augmented with custom modules (relaxed feasibility rules, scheduled penalty, adaptive search adjustment, bet-and-run) to minimize an LQG performance index for quantum controllers subject to physical realizability constraints. The central results are numerical outcomes from applying this optimizer to a specific quantum optical system and comparing the achieved indices against previously published controller designs. No derivation reduces a claimed prediction to a fitted parameter by construction, no self-citation chain supplies the uniqueness or correctness of the result, and the performance metric is defined externally to the algorithm. The method is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The approach builds directly on the standard differential evolution framework; no free parameters, invented entities, or non-standard axioms are mentioned in the abstract.

axioms (1)
  • domain assumption The foundational differential evolution framework can be extended with domain-specific modules for quantum constraints.
    The paper states it builds upon the foundational DE framework.

pith-pipeline@v0.9.0 · 5700 in / 1084 out tokens · 21068 ms · 2026-05-23T03:28:56.063718+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

44 extracted references · 44 canonical work pages · 1 internal anchor

  1. [1]

    C. C. Gerry and P. L. Knight, Introductory quantum optics. Cambridge Univ. Press, 2023

    work page 2023

  2. [2]

    H ∞ control of linear quantum stochastic systems,

    M. R. James, H. I. Nurdin, and I. R. Petersen, “ H ∞ control of linear quantum stochastic systems,”IEEE Trans. Autom. Control, vol. 53, no. 8, pp. 1787–1803, 2008

    work page 2008

  3. [3]

    Quantum control theory and applications: a survey,

    D. Dong and I. R. Petersen, “Quantum control theory and applications: a survey,” IET Control Theory Appl. , vol. 4, no. 12, pp. 2651–2671, 2010

    work page 2010

  4. [4]

    Fault-tolerant coherent H ∞ control for linear quantum systems,

    Y . Liu, D. Dong, I. R. Petersen, Q. Gao, S. X. Ding, S. Yokoyama, and H. Yonezawa, “Fault-tolerant coherent H ∞ control for linear quantum systems,” IEEE Trans. Autom. Control, vol. 67, no. 10, pp. 5087–5101, 2021

    work page 2021

  5. [5]

    Quantum estimation, control and learning: opportunities and challenges,

    D. Dong and I. R. Petersen, “Quantum estimation, control and learning: opportunities and challenges,” Annu. Rev. Control., vol. 54, pp. 243–251, 2022

    work page 2022

  6. [6]

    Continuous quantum error correction via quantum feedback control,

    C. Ahn, A. C. Doherty, and A. J. Landahl, “Continuous quantum error correction via quantum feedback control,” Phys. Rev. A, vol. 65, no. 4, p. 042301, 2002

    work page 2002

  7. [7]

    Rapid measurement of quantum systems using feedback control,

    J. Combes, H. M. Wiseman, and K. Jacobs, “Rapid measurement of quantum systems using feedback control,” Phys. Rev. Lett. , vol. 100, no. 16, p. 160503, 2008

    work page 2008

  8. [8]

    Lyapunov-based feedback preparation of GHZ entanglement of N-qubit systems,

    Y . Liu, S. Kuang, and S. Cong, “Lyapunov-based feedback preparation of GHZ entanglement of N-qubit systems,” IEEE Trans. Cybern., vol. 47, no. 11, pp. 3827–3839, 2016

    work page 2016

  9. [9]

    Rapid feedback stabilization of quantum systems with application to preparation of multiqubit entangled states,

    S. Kuang, G. Li, Y . Liu, X. Sun, and S. Cong, “Rapid feedback stabilization of quantum systems with application to preparation of multiqubit entangled states,” IEEE Trans. Cybern. , vol. 52, no. 10, pp. 11 213–11 225, 2021

    work page 2021

  10. [10]

    Fast State Stabilization using Deep Reinforcement Learning for Measurement-based Quantum Feedback Control

    C. Song, Y . Liu, D. Dong, and H. Yonezawa, “Fast state stabilization using deep reinforcement learning for measurement-based quantum feedback control,” arXiv preprint arXiv:2408.11328 , 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  11. [11]

    Auxiliary task-based deep reinforcement learning for quantum control,

    S. Zhou, H. Ma, S. Kuang, and D. Dong, “Auxiliary task-based deep reinforcement learning for quantum control,” IEEE Trans. Cybern. , vol. 55, no. 2, pp. 712–725, 2025

    work page 2025

  12. [12]

    Advantages of coherent feedback for cooling quantum oscillators,

    R. Hamerly and H. Mabuchi, “Advantages of coherent feedback for cooling quantum oscillators,” Phys. Rev. Lett. , vol. 109, no. 17, p. 173602, 2012

    work page 2012

  13. [13]

    Coherent feedback that beats all measurement-based feedback protocols,

    K. Jacobs, X. Wang, and H. M. Wiseman, “Coherent feedback that beats all measurement-based feedback protocols,” New J. Phys., vol. 16, no. 7, p. 073036, 2014

    work page 2014

  14. [14]

    Coherent quantum LQG control,

    H. I. Nurdin, M. R. James, and I. R. Petersen, “Coherent quantum LQG control,” Automatica, vol. 45, no. 8, pp. 1837–1846, 2009

    work page 2009

  15. [15]

    A quasi-separation principle and newton-like scheme for coherent quantum LQG control,

    I. G. Vladimirov and I. R. Petersen, “A quasi-separation principle and newton-like scheme for coherent quantum LQG control,” Syst. Control Lett., vol. 62, no. 7, pp. 550–559, 2013

    work page 2013

  16. [16]

    A numerical approach to optimal coherent quantum LQG controller design using gradient descent,

    A. K. Sichani, I. G. Vladimirov, and I. R. Petersen, “A numerical approach to optimal coherent quantum LQG controller design using gradient descent,” Automatica, vol. 85, pp. 314–326, 2017

    work page 2017

  17. [17]

    Direct and indirect couplings in coherent feedback control of linear quantum systems,

    G. Zhang and M. R. James, “Direct and indirect couplings in coherent feedback control of linear quantum systems,” IEEE Trans. Autom. Control, vol. 56, no. 7, pp. 1535–1550, 2010

    work page 2010

  18. [18]

    Coherent feedback control of linear quantum optical systems via squeezing and phase shift,

    G. Zhang, H. W. Joseph Lee, B. Huang, and H. Zhang, “Coherent feedback control of linear quantum optical systems via squeezing and phase shift,” SIAM J. Control Optim. , vol. 50, no. 4, pp. 2130–2150, 2012

    work page 2012

  19. [19]

    Control of quantum phenomena: past, present and future,

    C. Brif, R. Chakrabarti, and H. Rabitz, “Control of quantum phenomena: past, present and future,” New J. Phys., vol. 12, no. 7, p. 075008, 2010

    work page 2010

  20. [20]

    Teaching lasers to control molecules,

    R. S. Judson and H. Rabitz, “Teaching lasers to control molecules,” Phys. Rev. Lett., vol. 68, no. 10, p. 1500, 1992

    work page 1992

  21. [21]

    Dong and I

    D. Dong and I. R. Petersen, Learning and Robust Control in Quantum Technology. Springer Nat., 2023

    work page 2023

  22. [22]

    Learning- based quantum robust control: algorithm, applications, and experiments,

    D. Dong, X. Xing, H. Ma, C. Chen, Z. Liu, and H. Rabitz, “Learning- based quantum robust control: algorithm, applications, and experiments,” IEEE Trans. Cybern. , vol. 50, no. 8, pp. 3581–3593, 2020

    work page 2020

  23. [23]

    Sampling-based learning control for quantum systems with uncertain- ties,

    D. Dong, M. A. Mabrok, I. R. Petersen, B. Qi, C. Chen, and H. Rabitz, “Sampling-based learning control for quantum systems with uncertain- ties,” IEEE Trans. Control Syst. Technol., vol. 23, no. 6, pp. 2155–2166, 2015. 13

    work page 2015

  24. [24]

    Genetic algorithm optimization of laser pulses for molecular quantum state excitation,

    S. Sharma, H. Singh, and G. G. Balint-Kurti, “Genetic algorithm optimization of laser pulses for molecular quantum state excitation,” J. Chem. Phys. , vol. 132, no. 6, p. 064108, 2010

    work page 2010

  25. [25]

    Synthesis of linear coherent quantum control systems using a differential evolution algorithm,

    H. G. Harno and I. R. Petersen, “Synthesis of linear coherent quantum control systems using a differential evolution algorithm,” IEEE Trans. Autom. Control, vol. 60, no. 3, pp. 799–805, 2014

    work page 2014

  26. [26]

    Differential evolution–a simple and efficient heuristic for global optimization over continuous spaces,

    R. Storn and K. Price, “Differential evolution–a simple and efficient heuristic for global optimization over continuous spaces,” J. Glob. Optim., vol. 11, pp. 341–359, 1997

    work page 1997

  27. [27]

    On the usage of differential evolution for function optimiza- tion,

    R. Storn, “On the usage of differential evolution for function optimiza- tion,” in Proc. North Am. Fuzzy Inform. Process. Soc. IEEE, 1996, pp. 519–523

    work page 1996

  28. [28]

    Price, R

    K. Price, R. Storn, and J. Lampinen, Differential Evolution: a Practical Approach to Global Optimization . Springer, 2006

    work page 2006

  29. [29]

    Multiobjective differential evolution for feature selection in classification,

    P. Wang, B. Xue, J. Liang, and M. Zhang, “Multiobjective differential evolution for feature selection in classification,” IEEE Trans. Cybern. , vol. 53, no. 7, pp. 4579–4593, 2021

    work page 2021

  30. [30]

    Differential evolution: A survey of the state-of-the-art,

    S. Das and P. N. Suganthan, “Differential evolution: A survey of the state-of-the-art,” IEEE Trans. Evol. Comput. , vol. 15, no. 1, pp. 4–31, 2010

    work page 2010

  31. [31]

    Differential evolution: A survey of theo- retical analyses,

    K. R. Opara and J. Arabas, “Differential evolution: A survey of theo- retical analyses,” Swarm Evol. Comput. , vol. 44, pp. 546–558, 2019

    work page 2019

  32. [32]

    Constraint-handling in nature- inspired numerical optimization: past, present and future,

    E. Mezura-Montes and C. A. C. Coello, “Constraint-handling in nature- inspired numerical optimization: past, present and future,” Swarm Evol. Comput., vol. 1, no. 4, pp. 173–194, 2011

    work page 2011

  33. [33]

    Constraint handling techniques for metaheuristics: a state-of-the-art review and new variants,

    N. D. Lagaros, M. Kournoutos, N. A. Kallioras, and A. N. Nordas, “Constraint handling techniques for metaheuristics: a state-of-the-art review and new variants,” Optim. Eng., vol. 24, no. 4, pp. 2251–2298, 2023

    work page 2023

  34. [34]

    An efficient constraint handling method for genetic algorithms,

    K. Deb, “An efficient constraint handling method for genetic algorithms,” Comput. Methods Appl. Mech. Eng. , vol. 186, no. 2-4, pp. 311–338, 2000

    work page 2000

  35. [35]

    Incorporating objective function information into the feasibility rule for constrained evolutionary optimization,

    Y . Wang, B. Wang, H. Li, and G. G. Yen, “Incorporating objective function information into the feasibility rule for constrained evolutionary optimization,” IEEE Trans. Cybern. , vol. 46, no. 12, pp. 2938–2952, 2015

    work page 2015

  36. [36]

    Michalewicz, Genetic algorithms+ data structures= evolution pro- grams

    Z. Michalewicz, Genetic algorithms+ data structures= evolution pro- grams. Springer, 2013

    work page 2013

  37. [37]

    A fuzzy rule-based penalty function approach for constrained evolutionary optimization,

    C. Saha, S. Das, K. Pal, and S. Mukherjee, “A fuzzy rule-based penalty function approach for constrained evolutionary optimization,” IEEE Trans. Cybern., vol. 46, no. 12, pp. 2953–2965, 2014

    work page 2014

  38. [38]

    Optimization of laminate stacking sequence for buckling load maximization by genetic algorithm,

    R. L. Riche and R. T. Haftka, “Optimization of laminate stacking sequence for buckling load maximization by genetic algorithm,” AIAA journal, vol. 31, no. 5, pp. 951–956, 1993

    work page 1993

  39. [39]

    On the use of non-stationary penalty functions to solve nonlinear constrained optimization problems with GA’s,

    J. A. Joines and C. R. Houck, “On the use of non-stationary penalty functions to solve nonlinear constrained optimization problems with GA’s,” inProc. IEEE Conf. Evol. Comput. IEEE, 1994, pp. 579–584

    work page 1994

  40. [40]

    K. A. D. Jong, Evolutionary Computation: A Unified Approach . MIT Press, 2006

    work page 2006

  41. [41]

    Unconventional initialization meth- ods for differential evolution,

    M. Ali, M. Pant, and A. Abraham, “Unconventional initialization meth- ods for differential evolution,” Appl. Math. Comput., vol. 219, no. 9, pp. 4474–4494, 2013

    work page 2013

  42. [42]

    Exploiting erraticism in search,

    M. Fischetti and M. Monaci, “Exploiting erraticism in search,” Oper. Res., vol. 62, no. 1, pp. 114–122, 2014

    work page 2014

  43. [43]

    Feedback control of quantum systems using continuous state estimation,

    A. C. Doherty and K. Jacobs, “Feedback control of quantum systems using continuous state estimation,” Phys. Rev. A, vol. 60, no. 4, p. 2700, 1999

    work page 1999

  44. [44]

    Singular perturbation of quantum stochas- tic differential equations with coupling through an oscillator mode,

    J. Gough and R. Van Handel, “Singular perturbation of quantum stochas- tic differential equations with coupling through an oscillator mode,” J. Stat. Phys., vol. 127, no. 3, pp. 575–607, 2007

    work page 2007