pith. sign in

arxiv: 2312.16376 · v3 · submitted 2023-12-27 · ⚛️ physics.flu-dyn · physics.app-ph· physics.comp-ph

Acoustics-based Active Control of Unsteady Flow Dynamics using Reinforcement Learning Driven Synthetic Jets

Siddharth Rout , Khai Phan , Chao-An Lin This is my paper

Pith reviewed 2026-05-24 05:30 UTC · model grok-4.3

classification ⚛️ physics.flu-dyn physics.app-phphysics.comp-ph
keywords acoustic feedbackdeep reinforcement learningsynthetic jetscylinder wakedrag reductionnoise reductionflow control
0
0 comments X

The pith

Acoustic signals from a downstream hydrophone array alone can train a reinforcement learning agent to actuate synthetic jets and reduce both drag and noise in cylinder wakes.

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

The paper establishes that acoustic measurements alone can drive closed-loop control of unsteady cylinder wakes. A deep reinforcement learning agent uses signals from a downstream hydrophone array to decide when and how to activate synthetic jets on the cylinder. This yields up to 9.5 percent less radiated noise and 23.8 percent less drag at Reynolds number 100. Readers would care if this non-intrusive sensing approach proves reliable, since it avoids placing sensors inside the flow.

Core claim

The authors establish that acoustic observables serve as the sole state input for a DRL policy that learns to modulate synthetic jet actuation on the cylinder surface. By exploiting the acoustic signatures of vortex shedding, the policy suppresses coherent wake structures, yielding reductions of up to 9.5% in radiated noise and 23.8% in drag for the DFG 2D benchmark at Re=100. This demonstrates a direct mapping from far-field acoustic emissions to near-field flow control without any velocity or pressure field measurements.

What carries the argument

A deep reinforcement learning agent that takes time-series acoustic pressure data from a downstream hydrophone array as input and outputs commands for synthetic jet actuators placed on the cylinder.

Load-bearing premise

The downstream acoustic pressure signals contain enough information about the near-field vortex dynamics to allow the DRL policy to learn effective actuation without velocity data.

What would settle it

If a policy trained solely on acoustic signals produces jet actuation that leaves measured drag and noise levels unchanged from the baseline uncontrolled flow, the sufficiency claim would be falsified.

Figures

Figures reproduced from arXiv: 2312.16376 by Chao-An Lin, Khai Phan, Siddharth Rout.

Figure 12
Figure 12. Figure 12: A more important observation is the reduction in time [PITH_FULL_IMAGE:figures/full_fig_p008_12.png] view at source ↗
read the original abstract

Flow generated noise are caused shear flows and, hence, they can be used as feedback to control the flow. Existing flow control uses state variables like velocity, pressure, or vorticity, none use acoustic observables as the primary control signal. It is tough to model a classical control algorithm using sound level but data-driven approaches are not as do not have to explicitly model the physics. We present an acoustics-driven framework for active control of unsteady wake dynamics behind a circular cylinder, in which sound is used as the primary feedback signal for flow regulation. The approach integrates deep reinforcement learning (DRL) with synthetic jet actuation, using acoustic measurements acquired from a downstream hydrophone array to inform control decisions in real time. Unlike conventional flow control strategies that rely on velocity or pressure field sensing, the proposed method establishes a direct link between far-field acoustic emissions and near-field actuation. Within this formulation, the DRL agent learns control policies that exploit acoustic signatures of vortex shedding to modulate synthetic jet actuation on the cylinder surface. The resulting control suppresses coherent wake structures and mitigates flow-induced disturbances. Quantitative results show reductions of up to 9.5\% in radiated noise and 23.8\% in drag under the tested conditions, accompanied by a marked attenuation of wake oscillations, for a DFG 2D benchmark flow with Reynolds number 100. These findings demonstrate that acoustic sensing alone can provide sufficient information for effective closed-loop flow control and highlight its potential as a non-intrusive feedback modality for coupled aerodynamic and aeroacoustic optimization in bluff-body flows. The codes for the algorithm can be found here: https://github.com/Siddharth-Rout/FlowControlDRL.

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 paper introduces an acoustics-based active flow control framework for the unsteady wake behind a circular cylinder at Re=100. A DRL agent uses far-field acoustic pressure signals from a downstream hydrophone array as its sole observation to modulate synthetic-jet actuation on the cylinder surface, claiming reductions of up to 9.5% in radiated noise and 23.8% in drag together with attenuation of wake oscillations. The work positions acoustic sensing as a sufficient, non-intrusive alternative to direct velocity or pressure measurements for closed-loop suppression of vortex shedding.

Significance. If substantiated, the result would demonstrate that far-field acoustics can close the loop for DRL-based bluff-body control, reducing sensor intrusiveness in aeroacoustic applications. The public GitHub link for the algorithm code is a clear reproducibility asset. However, the absence of any observability analysis, DRL implementation details, or statistical characterization of the quoted percentages currently prevents assessment of whether the claimed performance is robust or merely incidental.

major comments (3)
  1. [Abstract] Abstract (and presumably the Results section): the central performance claims (9.5 % noise reduction, 23.8 % drag reduction) are stated without any information on DRL architecture, reward function, training convergence criteria, mesh resolution, or statistical variability across runs. These omissions make the quantitative results unverifiable and load-bearing for the claim that acoustic sensing alone suffices.
  2. [Method / Results] No section presents an observability analysis, transfer-function check, or ablation that isolates the acoustic channel from possible hydrodynamic contamination or phase loss in the downstream array. Without such verification the mapping from far-field hydrophone signals to near-wake vortex phase and amplitude remains an untested assumption.
  3. [Results] The manuscript supplies no baseline comparisons (e.g., open-loop actuation, velocity-based DRL, or classical feedback) nor any ablation removing the acoustic sensor to quantify its specific contribution. This weakens the assertion that the reported gains arise specifically from acoustic feedback.
minor comments (2)
  1. [Abstract] Abstract contains grammatical errors (e.g., 'Flow generated noise are caused shear flows') that should be corrected for clarity.
  2. [Method] Notation for the hydrophone array geometry, synthetic-jet placement, and acoustic propagation model is not introduced before the results are quoted.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive comments, which highlight opportunities to improve clarity and verifiability. We address each major point below and will revise the manuscript accordingly where appropriate.

read point-by-point responses

  1. Referee: [Abstract] Abstract (and presumably the Results section): the central performance claims (9.5 % noise reduction, 23.8 % drag reduction) are stated without any information on DRL architecture, reward function, training convergence criteria, mesh resolution, or statistical variability across runs. These omissions make the quantitative results unverifiable and load-bearing for the claim that acoustic sensing alone suffices.

    Authors: We agree that additional implementation details are needed for verifiability. The full manuscript contains a Methods section with the DRL setup (PPO algorithm, network architecture, reward based on noise and drag penalties), but these were not summarized in the abstract or Results. In revision we will add a concise paragraph to the abstract and expand the Results section with: network sizes, reward function equation, convergence criteria (e.g., reward plateau over 500 episodes), mesh resolution (standard 2D cylinder benchmark at Re=100), and statistical variability (mean and standard deviation of reductions across 5 independent training runs). The public GitHub code already allows reproduction of these quantities. revision: yes

  2. Referee: [Method / Results] No section presents an observability analysis, transfer-function check, or ablation that isolates the acoustic channel from possible hydrodynamic contamination or phase loss in the downstream array. Without such verification the mapping from far-field hydrophone signals to near-wake vortex phase and amplitude remains an untested assumption.

    Authors: We acknowledge that an explicit observability study would strengthen the claim. The work relies on the established aeroacoustic link (Curle’s analogy) that far-field pressure fluctuations are generated by near-wake vortex shedding and unsteady forces; the DRL agent learns this mapping empirically from data. In revision we will add a new subsection with (i) cross-correlation between hydrophone signals and lift coefficient time series and (ii) a brief discussion of phase delay based on acoustic propagation distance. A full transfer-function derivation or hydrodynamic-contamination ablation is not feasible within the current 2D acoustic formulation but can be noted as a limitation and direction for future 3D work. revision: partial

  3. Referee: [Results] The manuscript supplies no baseline comparisons (e.g., open-loop actuation, velocity-based DRL, or classical feedback) nor any ablation removing the acoustic sensor to quantify its specific contribution. This weakens the assertion that the reported gains arise specifically from acoustic feedback.

    Authors: The current results compare the DRL-controlled case against the uncontrolled (no-actuation) baseline, which already quantifies the net benefit. To address the request we will add open-loop synthetic-jet cases (constant blowing and periodic actuation at the shedding frequency) in the revised Results section. A velocity-based DRL comparison requires an entirely different sensor configuration and is outside the acoustics-focused scope of the paper; we will explicitly discuss this limitation. Removing the acoustic sensor reverts to open-loop or no control, which the added baselines will cover. revision: yes

Circularity Check

0 steps flagged

No circularity: data-driven RL results are not reduced to inputs by construction

full rationale

The paper describes a DRL agent trained on acoustic pressure signals to control synthetic jets, with performance metrics (noise and drag reductions) obtained from forward simulations of the closed-loop system. No equations, fitted parameters, or self-citations are presented that define the reported outcomes as equivalent to the training data by construction. The approach is self-contained as an empirical demonstration of a learned policy; the central claim of acoustic sufficiency is tested via simulation outcomes rather than assumed or renamed from prior fits.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

The abstract contains no explicit free parameters, axioms, or invented entities; the approach rests on standard DRL training and CFD discretization whose details are not supplied.

pith-pipeline@v0.9.0 · 5845 in / 1262 out tokens · 25708 ms · 2026-05-24T05:30:02.701512+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

56 extracted references · 56 canonical work pages · 2 internal anchors

  1. [1]

    Activecontrolofflow-inducedcavityoscillations,

    Cattafesta, L. N., Song, Q., Williams, D. R., Rowley, C. W., and Alvi, F. S., 2008,“Activecontrolofflow-inducedcavityoscillations,”ProgressinAerospace Sciences, 44(7), pp. 479–502

    work page 2008

  2. [2]

    Flow feedback control based on variable area cavitating venturi and its application in hybrid rocket motors,

    Tan,G.,Tian,H.,Gu,X.,Meng,X.,Wei,T.,Zhang,Y.,andCai,G.,2023,“Flow feedback control based on variable area cavitating venturi and its application in hybrid rocket motors,” Acta Astronautica,211, pp. 238–248

    work page 2023

  3. [3]

    Joslin, R. D. and Miller, D. N., 2009,Fundamentals and applications of Modern Flow Control, American Inst. of Aeronautics and Astronautics

    work page 2009

  4. [4]

    and Feng, L., 2018,Introduction, Cambridge Aerospace Series, Cam- bridge University Press, p

    Wang, J. and Feng, L., 2018,Introduction, Cambridge Aerospace Series, Cam- bridge University Press, p. 1–22

    work page 2018

  5. [5]

    Biomimeticflowcontroltechniques for aerospace applications: a comprehensive review,

    Rose,B.,NatarajanS,G.,andVt,G.,2021,“Biomimeticflowcontroltechniques for aerospace applications: a comprehensive review,” Reviews in Environmental Science and Bio/Technology,20

    work page 2021

  6. [6]

    Aerial and aquatic biological and bioinspired flow control strategies,

    Othman, A. K., Zekry, D. A., Saro-Cortes, V., Lee, K. J. P., and Wissa, A. A., 2023, “Aerial and aquatic biological and bioinspired flow control strategies,” Communications Engineering,2(1), 30

    work page 2023

  7. [7]

    Modification of a circular cylinder wake with synthetic jet: Vortex shedding modes and mechanism,

    Feng, L.-H. and Wang, J.-J., 2014, “Modification of a circular cylinder wake with synthetic jet: Vortex shedding modes and mechanism,” European Journal of Mechanics - B/Fluids,43, pp. 14–32

    work page 2014

  8. [8]

    Optimum-wavelength forcing of a bluff body wake,

    Bhattacharya, S. and Gregory, J. W., 2018, “Optimum-wavelength forcing of a bluff body wake,” Physics of Fluids,30(1), p. 015101

    work page 2018

  9. [9]

    Phase-averaged characteristics of flow around a circular cylinder under acoustic excitation control,

    Fujisawa, N., Takeda, G., and Ike, N., 2004, “Phase-averaged characteristics of flow around a circular cylinder under acoustic excitation control,” Journal of Fluids and Structures,19(2), pp. 159–170

    work page 2004

  10. [10]

    C., Andino, M

    Lin, J. C., Andino, M. Y., Alexander, M. G., Whalen, E. A., Spoor, M. A., Tran, J. T., and Wygnanski, I. J.,An Overview of Active Flow Control Enhanced Vertical Tail Technology Development

    work page

  11. [11]

    Boeing Finish Tests of 757 Vertical Tail with Advanced Technology,

    2023, “Boeing Finish Tests of 757 Vertical Tail with Advanced Technology,” https://www.nasa.gov/aeronautics/ nasa-boeing-finish-tests-of-757-vertical-tail-with-advanced-technology/

    work page 2023

  12. [12]

    A new mathematical explanation of what trig- gered the catastrophic torsional mode of the Tacoma Narrows Bridge,

    Arioli, G. and Gazzola, F., 2015, “A new mathematical explanation of what trig- gered the catastrophic torsional mode of the Tacoma Narrows Bridge,” Applied Mathematical Modelling,39(2), pp. 901–912

    work page 2015

  13. [13]

    VibrationControlina101-StoreyBuilding Using a Tuned Mass Damper,

    Tuan,A.Y.andShang,G.Q.,2014,“VibrationControlina101-StoreyBuilding Using a Tuned Mass Damper,” Journal of Applied Science and Engineering,17, pp. 141–156. 8 / PREPRINT FOR REVIEW Fig. 9 a) Resulting SPL of Test case 1 ([±1.50, ±2.25, ±2.75]), b) Resulting SPL of Test case 2 ([±2.00, ±2.50, ±3.00])

    work page 2014

  14. [14]

    Experimental study on wind-induced vibration and aerodynamic mitigation measures of a building over 800 meters,

    Gu, M., Su, L., Quan, Y., Huang, J., and Fu, G., 2022, “Experimental study on wind-induced vibration and aerodynamic mitigation measures of a building over 800 meters,” Journal of Building Engineering,46, p. 103681

    work page 2022

  15. [15]

    PASSIVE CONTROL OF VIV WITH DRAG REDUCTION,

    Owen, J., Bearman, P., and Szewczyk, A., 2001, “PASSIVE CONTROL OF VIV WITH DRAG REDUCTION,” Journal of Fluids and Structures,15(3), pp. 597–605

    work page 2001

  16. [16]

    Suppressing vortex-induced vibrations via passive means,

    Baek, H. and Karniadakis, G., 2009, “Suppressing vortex-induced vibrations via passive means,” Journal of Fluids and Structures,25(5), pp. 848–866

    work page 2009

  17. [17]

    Wake stabilization mechanism of low-drag suppression devices for vortex-induced vibration,

    Law, Y. Z. and Jaiman, R., 2017, “Wake stabilization mechanism of low-drag suppression devices for vortex-induced vibration,” Journal of Fluids and Struc- tures, 70, pp. 428–449

    work page 2017

  18. [18]

    Distributedforcingofflowoveracircularcylinder,

    Kim,J.andChoi,H.,2005,“Distributedforcingofflowoveracircularcylinder,” Physics of Fluids,17(3), p. 033103

    work page 2005

  19. [19]

    Elimination of Vortex Streets in Bluff-Body Flows,

    Dong, S., Triantafyllou, G. S., and Karniadakis, G. E., 2008, “Elimination of Vortex Streets in Bluff-Body Flows,” Phys. Rev. Lett.,100, p. 204501

    work page 2008

  20. [20]

    Active control of vortex- induced vibrations of a circular cylinder using windward-suction- leeward- blowing actuation,

    Wang, C., Tang, H., Yu, S. C. M., and Duan, F., 2016, “Active control of vortex- induced vibrations of a circular cylinder using windward-suction- leeward- blowing actuation,” Physics of Fluids,28(5), p. 053601

    work page 2016

  21. [21]

    Nonlinear optimal suppression of vortex shedding from a circular cylinder,

    Mao, X., Blackburn, H., and Sherwin, S., 2015, “Nonlinear optimal suppression of vortex shedding from a circular cylinder,” Journal of Fluid Mechanics,775, p. 241–265

    work page 2015

  22. [22]

    Feedbackcontrolofunstableflowandvortex- induced vibration using the eigensystem realization algorithm,

    Yao, W.andJaiman, R.K., 2017, “Feedbackcontrolofunstableflowandvortex- induced vibration using the eigensystem realization algorithm,” Journal of Fluid Mechanics, 827, p. 394–414

    work page 2017

  23. [23]

    Deep reinforcementlearningbasedsyntheticjetcontrolondisturbedflowoverairfoil,

    Yizhe, W., Mei, Y.-F., Aubry, N., Chen, Z., Wu, P., and Wu, W.-T., 2022, “Deep reinforcementlearningbasedsyntheticjetcontrolondisturbedflowoverairfoil,” Physics of Fluids,34, p. 033606

    work page 2022

  24. [24]

    Physics-constraineddeepreinforcementlearningforflowfielddenoising,

    Yousif, M. Z., Zhang, M., Yu, L., Yang, Y., Zhou, H., and Lim, H., 2023, “Physics-constraineddeepreinforcementlearningforflowfielddenoising,”Jour- nal of Fluid Mechanics,973, p. A12

    work page 2023

  25. [25]

    Reinforcement-learning-based control of confined cylinder wakes with stability analyses,

    Li, J. and Zhang, M., 2022, “Reinforcement-learning-based control of confined cylinder wakes with stability analyses,” Journal of Fluid Mechanics,932, p. A44

    work page 2022

  26. [26]

    Reinforcement learning for bluff body active flow control in experiments and simulations,

    Fan, D., Yang, L., Wang, Z., Triantafyllou, M. S., and Karniadakis, G. E., 2020, “Reinforcement learning for bluff body active flow control in experiments and simulations,” Proceedings of the National Academy of Sciences,117(42), pp. 26091–26098

    work page 2020

  27. [27]

    Ar- tificial neural networks trained through deep reinforcement learning discover control strategies for active flow control,

    Rabault, J., Kuchta, M., Jensen, A., Réglade, U., and Cerardi, N., 2019, “Ar- tificial neural networks trained through deep reinforcement learning discover control strategies for active flow control,” Journal of Fluid Mechanics,865, p. 281–302

    work page 2019

  28. [28]

    Adjoint-based machine learning for active flow control,

    Liu, X. and MacArt, J. F., 2023, “Adjoint-based machine learning for active flow control,” 2307.09980

    work page arXiv 2023

  29. [29]

    Deep Robot Sketching: AnapplicationofDeepQ-LearningNetworksforhuman-likesketch- ing,

    Fernandez-Fernandez, R., Victores, J. G., and Balaguer, C., 2023, “Deep Robot Sketching: AnapplicationofDeepQ-LearningNetworksforhuman-likesketch- ing,” Cognitive Systems Research,81, pp. 57–63

    work page 2023

  30. [30]

    Airfoil Shape Optimization using Deep Q- Network,

    Rout, S. and Lin, C.-A., 2022, “Airfoil Shape Optimization using Deep Q- Network,” 2211.17189

    work page arXiv 2022

  31. [31]

    Exciting a cavitating tip vortex with synthetic inflow turbulence: A CFD analysis of vortex kinematics, dynamics and sound generation,

    Klapwijk, M., Lloyd, T., Vaz, G., van den Boogaard, M., and van Terwisga, T., 2022, “Exciting a cavitating tip vortex with synthetic inflow turbulence: A CFD analysis of vortex kinematics, dynamics and sound generation,” Ocean Engineering, 254, p. 111246

    work page 2022

  32. [32]

    Ueber eine besondere Art der Tonerregung,

    Strouhal, V., 1878, “Ueber eine besondere Art der Tonerregung,” Annalen der Physik,241(10), pp. 216–251

    work page

  33. [33]

    Wied. Ann. Vol. 53, N. 13-1894,

    Stefanini, A., 1895, “Wied. Ann. Vol. 53, N. 13-1894,” Il Nuovo Cimento, pp. 134–138

    work page

  34. [34]

    EarlyAdvancementsinTurbulence-GeneratedNoiseModelling: A Review,

    Rout,S.,2023,“EarlyAdvancementsinTurbulence-GeneratedNoiseModelling: A Review,”Boundary Layer Flows - Advances in Modelling and Simulation, D. A. Aprovitola and P. G. Pezzella, eds., IntechOpen, Rijeka

    work page 2023

  35. [35]

    On Sound Generated Aerodynamically. I. General The- ory,

    Lighthill, M. J., 1952, “On Sound Generated Aerodynamically. I. General The- ory,” Proceedings of the Royal Society of London Series A,211(1107), pp. 564–587

    work page 1952

  36. [36]

    and Hatakeyama, N., 2002, “Sound generation by a two-dimensional PREPRINT FOR REVIEW / 9 Fig

    Inoue, O. and Hatakeyama, N., 2002, “Sound generation by a two-dimensional PREPRINT FOR REVIEW / 9 Fig. 10 Stage-2 a) Drag coefficient of Test case 1, b) Drag coefficient of Test case 2 Fig. 11 Trend of lift coefficient during the process and notable reduction in the amplitude of oscillation circular cylinder in a uniform flow,” Journal of Fluid Mechanics...

    work page 2002

  37. [37]

    Analysis of non-uniform laminar flow past a circular cylinder on the flow and sound field evolution using direct numerical simulation approach,

    Kumar, N., Kumar, S., Arumuru, V., and Bhumkar, Y., 2023, “Analysis of non-uniform laminar flow past a circular cylinder on the flow and sound field evolution using direct numerical simulation approach,”

    work page 2023

  38. [38]

    Sound generated in laminar flow past a two-dimensional rectangular cylinder,

    Liow, Y., Tan, B., Thompson, M., and Hourigan, K., 2006, “Sound generated in laminar flow past a two-dimensional rectangular cylinder,” Journal of Sound and Vibration,295(1), pp. 407–427

    work page 2006

  39. [39]

    Computational aeroacoustics for low Mach number flows,

    Crighton, D., 1993, “Computational aeroacoustics for low Mach number flows,” Computational aeroacoustics, Springer, pp. 50–68

    work page 1993

  40. [40]

    Computa- tional fluid dynamics and aeroacoustics for low Mach number flow,

    Meister, A., Struckmeier, J., Meister, A., and Struckmeier, J., 2002, “Computa- tional fluid dynamics and aeroacoustics for low Mach number flow,” Hyperbolic Partial Differential Equations: Theory, Numerics and Applications, pp. 269– 320. 10 / PREPRINT FOR REVIEW Fig. 12 Time averaged velocity without control (red) and after control (blue) along the longi...

    work page 2002

  41. [41]

    An acoustic analogy applied to incompressible flow fields,

    Ask, J., Davidson, L., Enwald, H., and Larsson, J., 2003, “An acoustic analogy applied to incompressible flow fields,” https://api.semanticscholar.org/ CorpusID:117816245

    work page 2003

  42. [42]

    On Lighthill’s acoustic analogy for low Mach number flows,

    Layton, W. and Novotn`y, A., 2009, “On Lighthill’s acoustic analogy for low Mach number flows,”New Directions in Mathematical Fluid Mechanics: The Alexander V. Kazhikhov Memorial Volume, Springer, pp. 247–279

    work page 2009

  43. [43]

    Learning from delayed rewards,

    Watkins, C. J. C. H., 1989, “Learning from delayed rewards,”

    work page 1989

  44. [44]

    Q-learning,

    Watkins, C. J. C. H. and Dayan, P., 1992, “Q-learning,” Machine Learning,8(3), pp. 279–292

    work page 1992

  45. [45]

    Playing Atari with Deep Reinforcement Learning

    Mnih, V., Kavukcuoglu, K., Silver, D., Graves, A., Antonoglou, I., Wierstra, D., and Riedmiller, M., 2013, “Playing Atari with Deep Reinforcement Learning,” 1312.5602

    work page internal anchor Pith review Pith/arXiv arXiv 2013

  46. [46]

    Adam: A Method for Stochastic Optimization

    Kingma, D. P. and Ba, J., 2014, “Adam: A method for stochastic optimization,” arXiv preprint arXiv:1412.6980

    work page internal anchor Pith review Pith/arXiv arXiv 2014

  47. [47]

    DOLFIN: Automated Finite Element Com- puting,

    Logg, A. and Wells, G. N., 2010, “DOLFIN: Automated Finite Element Com- puting,” ACM Trans. Math. Softw.,37(2)

    work page 2010

  48. [48]

    N., and Hake, J., 2012,DOLFIN: a C++/Python finite element library, Springer Berlin Heidelberg, Berlin, Heidelberg, pp

    Logg, A., Wells, G. N., and Hake, J., 2012,DOLFIN: a C++/Python finite element library, Springer Berlin Heidelberg, Berlin, Heidelberg, pp. 173–225

    work page 2012

  49. [49]

    Langtangen, H. P. and Logg, A., 2016,A Gallery of Finite Element Solvers, Springer International Publishing, Cham, pp. 37–81

    work page 2016

  50. [50]

    Schäfer, M., Turek, S., Durst, F., Krause, E., and Rannacher, R., 1996,Bench- mark Computations of Laminar Flow Around a Cylinder, Vieweg+Teubner Ver- lag, Wiesbaden, pp. 547–566

    work page 1996

  51. [51]

    Reference values for drag and lift of a two-dimensional time- dependent flow around a cylinder,

    John, V., 2004, “Reference values for drag and lift of a two-dimensional time- dependent flow around a cylinder,” International Journal for Numerical Methods in Fluids,44(7), pp. 777–788

    work page 2004

  52. [52]

    Reinforcement Learning: A Survey,

    Kaelbling, L. P., Littman, M. L., and Moore, A. W., 1996, “Reinforcement Learning: A Survey,” J. Artif. Int. Res.,4(1), p. 237–285

    work page 1996

  53. [53]

    Optimal Adaptive Policies for Markov Decision Processes,

    Burnetas, A. N. and Katehakis, M. N., 1997, “Optimal Adaptive Policies for Markov Decision Processes,” Mathematics of Operations Research,22(1), pp. 222–255

    work page 1997

  54. [54]

    P., 2012,Introduction to aerospace materials, American Institute of Aeronautics and Astronautics

    Mouritz, A. P., 2012,Introduction to aerospace materials, American Institute of Aeronautics and Astronautics

    work page 2012

  55. [55]

    Deep learning,

    LeCun, Y., Bengio, Y., and Hinton, G., 2015, “Deep learning,” nature, 521(7553), p. 436

    work page 2015

  56. [56]

    Rayleigh, L. J. W. S., 1877,The Theory of Sound, Macmillan and Co, London. PREPRINT FOR REVIEW / 11 Fig. 14 Time averaged fluctuation without control (red) and after control (blue) along the longitudinal central line passing through the channel 12 / PREPRINT FOR REVIEW List of Figures 1 Interaction between DQN agent and Environment in Markov Decision Proc...

    work page