Pith. sign in

REVIEW 5 major objections 5 minor 88 references

Real Time Control of Tandem-Wing Experimental Platform Using Concerto Reinforcement Learning

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

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

Pith's one-line read A reinforcement-learning controller can run above 2500 Hz on ordinary CPUs while still updating online, and it cuts tracking error by 18.3% to 60.7% when paired with PID, Adaptive PID, or MRAC.

desk verdict Useful CPU timing result buried under an unsupported finite-time convergence proof; the paper overclaims as written. read the letter →

arxiv 2502.10429 v1 pith:NCHK4JOW submitted 2025-02-08 cs.LG cs.AIcs.ROcs.SYeess.SY

Zhang Minghao , Yang Xiaojun , Wang Zhihe , Wang Liang This is my paper · ORCID
classification cs.LGcs.AIcs.ROcs.SYeess.SY
keywords concertoreinforcementlearningreal-timecontrolonlineweightupdatesflapping-wingaircrafttandem-wingplatformfinite-timeLyapunovconvergenceedgeinference
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper proposes CRL2RT, a reinforcement-learning control algorithm designed to run online at high frequency on a dragonfly-inspired, direct-drive tandem-wing experimental platform. It claims CRL2RT sustains a control frequency above 2500 Hz on ordinary CPUs while still updating network weights from incoming data, and that when paired with PID, Adaptive PID, or MRAC it lowers tracking error by 18.3% to 60.7% under 40 Hz and 60 Hz flapping conditions. The core argument is that safety and finite-time convergence can be guaranteed by alternating classical and learned actions and by composing past policy updates according to a Lyapunov condition, so real-time RL need not sacrifice online learning.

What carries the argument

The two load-bearing mechanisms are the Time-Interleaved module and the Rule-Based Policy Composer. The Time-Interleaved module alternates classical-control steps with RL steps so the classical controller's bounded error-reduction capability keeps the combined error Lipschitz continuous and bounds worst-case deviation. The Rule-Based Policy Composer, using a gradient-domain Laplace transform estimate of Q-values from reward histories without an edge-side critic, adds posterior-screened policy weights at the start of each descent segment so that the Lyapunov function satisfies $dQ/dt \le -c Q^{\alpha}$, yielding finite-time convergence. On the edge, inference is restructured as direct numerical array operations, removing tensorization and de-tensorization overhead.

What would settle it

Run the trained CRL2RT policy on the physical DDTWEP with its six-component balance under the same 40 Hz and 60 Hz flapping commands and compare tracking error with the semi-physical simulation. If the 18.3% to 60.7% improvement over the classical controllers does not appear, or if the slowest control frequency drops below 2000 Hz once weight updates and data transmission are included, the paper's central claim is falsified.

Watch

Extended reading notes

Core claim

The central claim is that real-time reinforcement learning does not have to trade away online weight updates: by running weight updates on the cloud side, running direct matrix-operation inference on the edge side, interleaving classical and learned controllers, and injecting policy-composer weights that satisfy a finite-time Lyapunov condition, CRL2RT reaches a slowest control frequency around 2534 Hz on a general-purpose CPU while still updating from experience. The paper further claims the same framework works with three different classical controllers, improving tracking performance by 18.3% to 60.7% in the 40 Hz and 60 Hz configurations, with finite-time convergence and safety provided by the two theoretical modules.

Load-bearing premise

All measured frequencies and tracking gains come from a semi-physical simulation whose eight-degree-of-freedom tandem-wing model and realism validation are taken from earlier work cited in the paper; if that model does not reproduce the real unsteady aerodynamic loads on the actual platform, the numbers will not transfer to hardware.

Editorial extensions

If this is right

  • A 2000 Hz motion-control requirement can be met by an RL controller that updates its weights online on commodity CPUs, without FPGA or accelerator hardware.
  • The same learned framework can be layered onto PID, Adaptive PID, or MRAC baselines, upgrading existing classical controllers rather than replacing them.
  • Under 40 Hz and 60 Hz flapping conditions, tracking error falls by 18.3% to 60.7%; at 20 Hz with MRAC the improvement is not significant because MRAC alone already performs well.
  • Removing tensor operations from edge inference cuts per-step execution time by more than a factor of three, making I/O the dominant cost.
  • The Lipschitz and finite-time Lyapunov conditions imply that, during training, worst-case tracking error stays bounded on every two-step interleaved cycle.

Reading between the lines

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

  • A natural extension is to test the architecture on other high-rate mechatronic systems, such as quadruped joints or precision motion stages; that would show whether the 2500 Hz result is a property of the controller structure or of this particular eight-DOF platform.
  • The 20 Hz MRAC case suggests that when the classical partner already performs well the exploration signal weakens; an altered exploration schedule might recover gains in that regime, but this is not examined in the paper.
  • Because the same simulation model is used for training and evaluation, a physical flight or tethered hardware trial is the decisive next step, and the platform's own six-component balance makes that trial directly feasible.
  • The gradient-domain Laplace estimator obtains Q-values from reward histories without an edge-side critic, a trick that may carry over to other latency-sensitive RL deployments.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 5 minor

Summary. The paper proposes CRL2RT, a reinforcement learning framework for real-time control of the Direct-Drive Tandem-Wing Experimental Platform (DDTWEP). The algorithm combines classical controllers (PID, Adaptive PID, MRAC) with an RL controller via a time-interleaved module and a rule-based policy composer, with cloud-side weight updates and edge-side inference. The authors claim a control frequency above 2500 Hz on standard CPUs, finite-time convergence and safety guarantees, and tracking improvements of 18.3% to 60.7% when combined with the three classical controllers, based on experiments in a semi-physical simulation.

Significance. If the claims were fully supported, the paper would represent a useful engineering contribution to real-time RL control, particularly the latency-focused decomposition of edge/cloud computation and the attempt to integrate RL with classical controllers at high frequency. The timing breakdown in Table 7 is informative. However, the theoretical guarantees for finite-time convergence and safety are not established by the derivations as written, and the headline performance claim is contradicted by the paper's own results in the MRAC 20 Hz configurations. The reliance on an unvalidated simulation model further limits the current significance.

major comments (5)
  1. [Section IV.B, Eqs. (30)-(32)] The step from Eq. (30) to Eq. (31) is not a valid algebraic simplification: Eq. (30) contains the terms Q_i(0), r_i(0), gamma K_{i+1}, and Q_{i+1}(0), all of which are discarded without justification. Consequently, the Q-value estimator Q_i(theta) = r_i(theta)/(1-gamma_i) in Eq. (32) and the composer update in Eq. (34) are not derived from the Bellman equation. Since the finite-time convergence proof in Section V.C.2 relies directly on this Q-estimator, the theoretical foundation of the central claim is unsupported.
  2. [Section V.C.2, Eqs. (59)-(74)] The finite-time Lyapunov proof is circular and incomplete. Equation (59) is stated as a relation that the policy composer must satisfy, but the existence of theta_CP and the sequence m_g is never demonstrated, and no constraints on m_g are given. Equations (68)-(69) assert the sign conditions dQ/dS < 0 and dQ/dtheta < 0 without proof, and Eq. (73) is imposed rather than derived. The final argument that an exponentially growing product dominates a polynomially growing Q^alpha does not establish the pointwise inequality in Eq. (63) that is required for finite-time Lyapunov convergence. These gaps are load-bearing for the claim that CRL2RT 'ensures finite-time convergence.'
  3. [Section V.B, Eqs. (50)-(53)] The safety guarantee of the time-interleaved module depends on a finite upper bound P_E,rl,max on the RL policy's ability to increase the tracking error. The paper never establishes such a bound for the actual policy class used in the experiments. Without this bound, the Lipschitz inequality in Eq. (51) is conditional, and the claim that the algorithm 'ensures safety' through the time-interleaved module is not supported.
  4. [Abstract and Table 11] The abstract states that CRL2RT 'enhances tracking performance by 18.3% to 60.7%' when integrated with PID, Adaptive PID, and MRAC, but Table 11 reports errors of -172.0% and -115.3% for the CRL2RT(MRAC) configuration in the two 20 Hz conditions. The paper acknowledges this degradation in Section VI.D, but the unqualified abstract claim and the contribution statement in Section II are not accurate over the full set of reported experiments.
  5. [Section VI.B.1 and Table 1] The experimental results are generated entirely in a semi-physical simulation, and the subsystem models are delegated to the authors' prior work [21] and to an unreviewed preprint [62]. No physical hardware experiments are reported, and no independent validation of the 8-DOF model is provided. The timing measurements in Table 7 appear to be single trials without error bars or code release, so the headline 2534 Hz control frequency is not established as a robust, reproducible result.
minor comments (5)
  1. [General] Section numbering is inconsistent: the text refers to 'Section 3.2' and 'Section 4.3' when the relevant sections are actually IV.B and V.C, and Section IV is titled 'Primarily,' which appears to be a typographical error.
  2. [Eqs. (22)-(25)] The gradient-domain Laplace transform is not rigorously defined: Eq. (22) is dimensionally unclear, and the exponent in Eqs. (24)-(25) uses both s and the newly introduced g_d without a clear relation between these variables.
  3. [Table 7] The three column headers in Table 7 are ambiguous, and the corresponding frameworks are not defined precisely in the text; the table also appears to report a single measurement per item with no statistics, which should be stated explicitly.
  4. [Figures 10-12] The plots are labeled 'Rewards,' but the reward function in Eq. (17) is a sum of tracking errors that the algorithm minimizes; the y-axis label and the direction of improvement should be clarified to avoid confusion.
  5. [Editorial] There are several typographical errors, including Eq. (2) repeating T_M,2 and omitting T_M,4, Eq. (10) using inconsistent subscripts (J_MZZ, C), and reference [56] being incomplete.

Circularity Check

4 steps flagged · score 8.0 of 10

Finite-time convergence is assumed in Eq. (59) and Eq. (73), the Q-estimator is a reward rescaling, and the experimental platform's fidelity rests on same-team citations; the central theoretical guarantee reduces to its own construction.

  1. self definitional [Section IV.B (Eqs. (30)-(32))]
    "g ∙ K_i(g_d) − Q_i(0) = g ∙ R_i(g) − r_i(0) + γ ∙ K_{i+1}(g_d) − Q_{i+1}(0) (30) Simplifying this yields: K_i(g) = R_i(g_d)/1 − γ (31) ... To mitigate the error in estimating Q_i(θ) based on r_i(θ): Q_i(θ) = r_i(θ)/(1 − γ_i) (32)"

    Equation (31) is obtained from Eq. (30) by discarding the terms Q_i(0), r_i(0), and γ K_{i+1}(g), so it is not a consequence of the Bellman relation (28)-(29). Eq. (32) then fixes Q_i(θ) as r_i(θ)/(1-γ_i), i.e., the Q-value is defined to be a rescaled reward. Because the policy update (43)-(44) and the Lyapunov argument use this Q, the 'estimate' is the paper's own construction rather than an independently derived quantity, and any convergence conclusion drawn from it is self-referential.

  2. self definitional [Section IV.C.2 (Eqs. (59)-(61))]
    "Q_g̿(Δθ)/Δθ = (1 + m_g) ∙ Q_{g−1}̿(Δθ)/Δθ (59) ... The primary function of the Policy Composer is to achieve the following posterior-screened expression: dθ_CP,g,j(t)/dt = Q_g̿(Δθ)/Δθ = ∏(1+m_g)^g_0 ∙ Q_0̿(Δθ)/Δθ (61)"

    Eq. (59) is not derived; it is the 'posterior acceptance condition' that defines how θ_CP is chosen, and it asserts that Q_g/Δθ is (1+m_g) times Q_{g-1}/Δθ. This is exactly the exponentially fast decrease of the Lyapunov candidate Q that the proof is supposed to establish. The product formula in Eq. (60) and the composer derivative in Eq. (61) merely restate this assumption, so the claimed finite-time convergence is put in by hand rather than obtained from the learning dynamics.

2 more flagged steps
  1. self definitional [Section IV.C.2 (Eqs. (63), (72)-(74))]
    "To satisfy the finite-time Lyapunov convergence condition: dQ/dθ ∙ ∏(1+m_g)^g_0 ∙ Q_0̿(Δθ)/Δθ ≤ −c∗ ∙ Q(t)^α (73) c∗ = c/[(dQ/dw)^2 ∙ α_w ∙ γ + 1] (74) From equation (73), it is evident that ∏(1+m_g)^g_0 grows exponentially, while Q(t)^α grows polynomially. Thus, the finite-time Lyapunov convergence condition is satisfied."

    Eq. (73) is imposed as a condition to be satisfied, not derived from the update rules; the definition of c* in Eq. (74) only renames the constant. The concluding 'it is evident' argument observes that the assumed inequality is consistent with the desired exponential/polynomial comparison, but it does not show that the policy update produces dQ/dt ≤ -c Q^α. With m_g unconstrained, Eq. (59)-(73) can always be made true by choosing m_g large, so the finite-time convergence result reduces to the assumption.

  2. self citation load bearing [Section III.B, Table 1 and Section VI.B.1; refs [21], [62]]
    "A simulation system has been established to provide the necessary training data for facilitating reinforcement learning. The realism of this simulation system has been validated in existing literature [62]. Table 1: System Properties Provide parameters such as wing inertia. See literature for calculation method[21]."

    The paper's performance experiments are all run in the semi-physical simulation; the realism of that simulation is the load-bearing premise for transferring the 2534 Hz and 18.3-60.7% results to DDTWEP. That premise is supported only by [62], a same-team SSRN preprint, while every subsystem model in Table 1 is delegated to [21], also by the same team. No independent hardware validation or externally verified model is provided in this paper, so the experimental claims rest on a self-citation chain.

full rationale

The real-time engineering contribution (matrix-operation inference, load balancing, and the measured 2534 Hz edge speed) is largely self-contained and non-circular; the timing tables do not depend on the theory. However, the paper's headline guarantees of finite-time convergence and safety are not independent derivations. The Q-estimator of Eq. (32) is defined from rewards, the composer condition Eq. (59) assumes exponential decrease of Q, and the Lyapunov inequality Eq. (73) is imposed rather than verified. The finite-time convergence theorem therefore reduces to the paper's own definitions. The simulation fidelity behind the experimental transfer is also delegated to same-team citations [21], [62]. These are load-bearing, not minor, so the circularity score is 8.

Assumptions & free parameters 6 free parameters · 6 assumptions · 3 invented entities

The central claim rests on a simulation model validated only by the authors' own prior preprints, an unproved gradient-domain Laplace transform, imposed convergence conditions, and six design parameters whose values are not reported. The paper contributes integration and deployment engineering, but the theoretical machinery does not add independent support.

free parameters (6)
  • m_g = not specified
    Eq (59) uses m_g to impose the rate at which the policy composer accelerates Q-decrease each dynamic descent phase. No formula or value is given; it is selected based on observed performance.
  • beta = not specified
    The noise term theta_noise(x, theta_k) is scaled by beta in [0,1] in Section V.C.3. Beta tunes composer perturbation strength and is not fixed in the paper.
  • GDS length L and DDP count N = not specified
    Section V.C.1 segments single-life into every L time step gradient descent segments and every N segments dynamic descent phases. L and N are required to run the policy composer and are never given values.
  • State history M and lookahead H = M=7, H=3
    Eq (76) sets the state as M past observations and actions plus H future errors. M=7 and H=3 are chosen without ablation or justification.
  • Reward scale lambda = not specified
    Eq (17) defines the reward as lambda times the sum of absolute tracking errors, with lambda used to scale the reward into [0,1] but no value reported.
  • Time-interleaved period = k mod 2 = 0
    Eqs (35)-(39) alternate classical and RL actions every other step. This duty cycle is a design choice that is not varied or justified.
assumptions (6)
  • domain assumption The 8-DOF simulation model (Eqs 3-15) and its subsystem models faithfully represent DDTWEP's unsteady tandem-wing loads.
    Every experimental result is generated in the semi-physical simulation of Fig 9. Table 1 delegates each subsystem model to the authors' prior work [21], and Section VI.B.1 cites [62] for realism; [62] is the authors' own preprint, not an independent validation.
  • ad hoc to paper Gradient-domain Laplace transform identities (Eqs 22-25) are valid and admit the simplification K_i(g)=R_i(g)/(1-gamma) in Eq (31).
    The transform is defined with theta = integral of g from 0 to infinity, a dimensionally inconsistent expression, and is invoked to replace Bellman's equation with Q_i(theta)=r_i(theta)/(1-gamma_i). No proof of these properties is given.
  • ad hoc to paper The discount factor satisfies gamma_i approximately equal to gamma_bar over the single life (Assumption 1).
    Assumption 1 in Section IV.B asserts that gamma_i can be replaced by its mean because existing research indicates the influence becomes negligible; the cited references [70,71] are about sample complexity, not about this approximation.
  • domain assumption The error dynamics obey E(t+Delta_t)=E(t)+P_rl Delta_t and E(t+2Delta_t)=E(t+Delta_t)-P_class Delta_t, with P_class constant and P_rl bounded.
    Used in Section V.B to derive the Lipschitz inequality. The linear-in-time error model and constant Time-Interleaved Capability are asserted rather than derived from the platform dynamics.
  • ad hoc to paper The signs dQ/dS < 0, dQ/dtheta < 0, and [r_k+(gamma-1)Q] tending to 0 hold throughout training.
    These sign conditions, Eqs (68)-(71), are needed so that dQ/dt in Eq (72) is negative enough to satisfy the finite-time Lyapunov convergence condition. They are stated without proof or experimental verification.
  • ad hoc to paper Q_bar_g(Delta_theta)/Delta_theta = (1+m_g) Q_bar_{g-1}(Delta_theta)/Delta_theta can be enforced by choosing theta_CP, and the chosen m_g makes the product grow exponentially.
    Eq (59) is the defining rule for the policy composer, not a consequence of learning dynamics. The subsequent finite-time proof uses this imposed identity as its main engine, which makes the convergence argument circular.
invented entities (3)
  • Gradient-domain Laplace transform and its complex variable g_d = S_P + i D_R
    purpose: Provides the Q-value estimation rule Q_i approximately r_i/(1-gamma_i) and the Delta_Q update formula used by the Rule-Based Policy Composer.
    The transform is introduced in Section IV.B with no external benchmark or formal justification; its only role is to produce the reward-to-Q mapping the controller needs.
  • Rule-Based Policy Composer weight theta_CP(t)
    purpose: Adds posterior-screened policy weight at each gradient descent segment to accelerate convergence according to Eq (61).
    theta_CP is defined by the imposed identity Eq (59); no independent measurement or falsifiable prediction is attached to it.
  • Time-Interleaved Capability P_class and P_RL,max
    purpose: Used in Section V.B to prove Lipschitz continuity and safety constraints.
    No procedure is given to measure these values on the platform; the proof assumes their existence and boundedness.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Real Time Control of Tandem-Wing Experimental Platform Using Concerto Reinforcement Learning." pith.science (2026). https://pith.science/paper/NCHK4JOW

@misc{pith2026250210429,
  author       = {Pith},
  title        = {Pith review of: Real Time Control of Tandem-Wing Experimental Platform Using Concerto Reinforcement Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NCHK4JOW}},
  note         = {Machine review of arXiv:2502.10429}
}
read the original abstract

This paper introduces the CRL2RT algorithm, an advanced reinforcement learning method aimed at improving the real-time control performance of the Direct-Drive Tandem-Wing Experimental Platform (DDTWEP). Inspired by dragonfly flight, DDTWEP's tandem wing structure causes nonlinear and unsteady aerodynamic interactions, leading to complex load behaviors during pitch, roll, and yaw maneuvers. These complexities challenge stable motion control at high frequencies (2000 Hz). To overcome these issues, we developed the CRL2RT algorithm, which combines classical control elements with reinforcement learning-based controllers using a time-interleaved architecture and a rule-based policy composer. This integration ensures finite-time convergence and single-life adaptability. Experimental results under various conditions, including different flapping frequencies and yaw disturbances, show that CRL2RT achieves a control frequency surpassing 2500 Hz on standard CPUs. Additionally, when integrated with classical controllers like PID, Adaptive PID, and Model Reference Adaptive Control (MRAC), CRL2RT enhances tracking performance by 18.3% to 60.7%. These findings demonstrate CRL2RT's broad applicability and superior performance in complex real-time control scenarios, validating its effectiveness in overcoming existing control strategy limitations and advancing robust, efficient real-time control for biomimetic aerial vehicles.

Figures

Figures reproduced from arXiv: 2502.10429 by the authors.

Figure 1
Figure 1. Illustration of system integration of the DDD [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 7
Figure 7. Actor network The critic network adopts a seven-layer MLP structure, as shown in [PITH_FULL_IMAGE:figures/full_fig_p017_7.png] view at source ↗
Figure 8
Figure 8. Critic network F. Custom Edge Inference Framework for Real-Time Control Existing studies typically employ PyTorch for the policy inference process in reinforcement learning[75, 78]. However, the real-time control at 2000 Hz emphasizes I/O overhead and the additional time required for converting received numpy array data to tensors. To address these challenges, the neural network inference process has been restructur… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

88 extracted references · 60 canonical work pages

  1. [21]

    ConcertoRL: A Reinforcement Learning Approach for Finite-Time Single-Life Enhanced Control and its Application to Direct-Drive Tandem-Wing Experiment Platforms,

    Zhang, M., Song, B., Chen, C., Lang, X., and Wang, L. "ConcertoRL: A Reinforcement Learning Approach for Finite-Time Single-Life Enhanced Control and its Application to Direct-Drive Tandem-Wing Experiment Platforms," Applied Intelligence, 2024, p. in press

    2024

  2. [62]

    Development of Minimal Biorobotic Stealth Distance and Its Application in the Design of Direct-Drive Dragonfly-Inspired Aircraft

    ZHANG, M., Song, B., Yang, X., Wang, L., and Lang, X. "Development of Minimal Biorobotic Stealth Distance and its Application in the Design of Direct-Drive Dragonfly-Inspired Aircraft," Available at SSRN 4930951. doi: https://doi.org/10.48550/arXiv.2410.15559

    work page Pith review arXiv doi:10.48550/arxiv.2410.15559

  3. [1]

    COLIBRI: A hovering flapping twin -wing robot,

    Roshanbin, A., Altartouri, H., Kará sek, M., and Preumont, A. "COLIBRI: A hovering flapping twin -wing robot," International Journal of Micro Air Vehicles Vol. 9, No. 4, 2017, pp. 270-282. doi: 10.1177/1756829317695563

    work page doi:10.1177/1756829317695563 2017

  4. [2]

    Dove: A biomimetic flapping -wing micro air vehicle,

    Yang, W., Wang, L., and Song, B. "Dove: A biomimetic flapping -wing micro air vehicle," International Journal of Micro Air Vehicles Vol. 10, No. 1, 2018, pp. 70-84. doi: https://doi.org/10.1177/1756829317734837

    work page doi:10.1177/1756829317734837 2018

  5. [3]

    Design and Experimental Verification of a Roll Control Strategy for Large Wingspan Flapping -Wing Aerial Vehicle,

    Meng, R., Song, B., Xuan, J., Yang, X., and Xue, D. "Design and Experimental Verification of a Roll Control Strategy for Large Wingspan Flapping -Wing Aerial Vehicle," Journal of Bionic Engineering, 2024, pp. 1 - 18. doi: https://doi.org/10.1007/s42235-024-00532-4

    work page doi:10.1007/s42235-024-00532-4 2024

  6. [4]

    A numerical study on the aerodynamic effects of dynamic twisting on forward flight flapping wings,

    Dong, Y., Song, B., Yang, W., and Xue, D. "A numerical study on the aerodynamic effects of dynamic twisting on forward flight flapping wings," Bioinspiration & Biomimetics, 2024. doi: 10.1088/1748-3190/ad253b

    work page doi:10.1088/1748-3190/ad253b 2024

  7. [5]

    Effects of dynamical spanwise retraction and stretch on flapping-wing forward flights,

    Kang, L., Bifeng, S., Ang, C., Zhihe, W., Dong, X., and Wenqing, Y. "Effects of dynamical spanwise retraction and stretch on flapping-wing forward flights," Chinese Journal of Aeronautics, 2024. doi: https://doi.org/10.1016/j.cja.2024.01.006

    work page doi:10.1016/j.cja.2024.01.006 2024

  8. [6]

    Experimental study on the effect of increased downstroke duration for an FWAV with morphing-coupled wing flapping configuration,

    Chen, A., Song, B., Wang, Z., Liu, K., Xue, D., and Yang, X. "Experimental study on the effect of increased downstroke duration for an FWAV with morphing-coupled wing flapping configuration," Journal of Bionic Engineering Vol. 21, No. 1, 2024, pp. 192-208. doi: https://doi.org/10.1007/s42235-023-00443-w

    work page doi:10.1007/s42235-023-00443-w 2024

Show all 88 references
  1. [7]

    Design and Verification of a Large -Scaled Flapping-Wing Aircraft Named “Cloud Owl

    Meng, R., Song, B., Xuan, J., and Yang, X. "Design and Verification of a Large -Scaled Flapping-Wing Aircraft Named “Cloud Owl”," Applied Sciences Vol. 13, No. 9, 2023, p. 5667. doi: https://doi.org/10.3390/app13095667

    2023 doi

  2. [8]

    HiFly -Dragon: A Dragonfly Inspired Flapping Flying Robot with Modified, Resonant, Direct-Driven Flapping Mechanisms,

    Ma, H., Gong, P., Tian, Y., Wu, Q., Pan, M., Yin, H., Liu, Y., and Chen, C. "HiFly -Dragon: A Dragonfly Inspired Flapping Flying Robot with Modified, Resonant, Direct-Driven Flapping Mechanisms," Drones Vol. 8, No. 4, 2024, p. 126. doi: https://doi.org/10.3390/drones8040126

    2024 doi

  3. [9]

    Research on the Energy Balance of Long Endurance Hover -capable Direct -drive Dragonfly -like Aircraft Affected by Design Parameters,

    Zhang M, Song B, Lang X, Yang X, Sun Z, Xue D, Xuan J, and C., Y. "Research on the Energy Balance of Long Endurance Hover -capable Direct -drive Dragonfly -like Aircraft Affected by Design Parameters," APISAT-2022. Niigata City, 2022

    2022

  4. [10]

    Three-dimensional sweeping motion effects on hovering dragonflies,

    Zhu, Z., Song, B., Xue, D., Yang, W., Lang, X., and Yang, X. "Three-dimensional sweeping motion effects on hovering dragonflies," Aerospace Science and Technology Vol. 127, 2022, p. 107701. doi: https://doi.org/10.1016/j.ast.2022.107701 26

    2022

  5. [11]

    Tandem -wing interactions on aerodynamic performance inspired by dragonfly hovering,

    Peng, L., Zheng, M., Pan, T., Su, G., and Li, Q. "Tandem -wing interactions on aerodynamic performance inspired by dragonfly hovering," Royal Society open science Vol. 8, No. 8, 2021, p. 202275. doi: https://doi.org/10.1098/rsos.202275

    2021 doi

  6. [12]

    Festo BionicOpter

    "Festo BionicOpter." 2021

    2021

  7. [13]

    Quad-thopter: Tailless flapping wing robot with four pairs of wings,

    De Wagter, C., Kará sek, M., and de Croon, G. "Quad-thopter: Tailless flapping wing robot with four pairs of wings," International Journal of Micro Air Vehicles Vol. 10, No. 3, 2018, pp. 244-253. doi: 10.1177/1756829318794972

    2018 doi

  8. [14]

    A tailless aerial robotic flapper reveals that flies use torque coupling in rapid banked turns,

    Kará sek, M., Muijres, F. T., Wagter, C. D., Remes, B. D. W., and Croon, G. C. H. E. d. "A tailless aerial robotic flapper reveals that flies use torque coupling in rapid banked turns," Science Vol. 361, No. 6407, 2018, pp. 1089-1094. doi: 10.1126/science.aat0350

    2018 doi

  9. [15]

    Platform design and tethered flight of a motor-driven flapping-wing system,

    Hines, L., Colmenares, D., and Sitti, M. "Platform design and tethered flight of a motor-driven flapping-wing system," 2015 IEEE international conference on robotics and automation (ICRA) . IEEE, 2015, pp. 5838 - 5845

    2015

  10. [16]

    Grow your limits: Continuous improvement with real-world rl for robotic locomotion,

    Smith, L., Cao, Y., and Levine, S. "Grow your limits: Continuous improvement with real-world rl for robotic locomotion," arXiv preprint arXiv:2310.17634, 2023. doi: https://doi.org/10.48550/arXiv.2310.17634

    2023 doi arXiv

  11. [17]

    Acting is seeing: Navigating tight space using flapping wings,

    Tu, Z., Fei, F., Zhang, J., and Deng, X. "Acting is seeing: Navigating tight space using flapping wings," 2019 International Conference on Robotics and Automation (ICRA). IEEE, 2019, pp. 95-101

    2019

  12. [18]

    Crawl and fly: A bio-inspired robot utilizing unified actuation for hybrid aerial-terrestrial locomotion,

    Tu, Z., Hui, C., Liu, L., Zhou, Y., Romano, D. R., and Deng, X. "Crawl and fly: A bio-inspired robot utilizing unified actuation for hybrid aerial-terrestrial locomotion," IEEE Robotics and Automation Letters Vol. 6, No. 4, 2021, pp. 7549-7556. doi: 10.1109/LRA.2021.3099246

    2021

  13. [19]

    Flying With Damaged Wings: The Effect on Flight Capacity and Bio-Inspired Coping Strategies of a Flapping Wing Robot,

    Tu, Z., Fei, F., Liu, L., Zhou, Y., and Deng, X. "Flying With Damaged Wings: The Effect on Flight Capacity and Bio-Inspired Coping Strategies of a Flapping Wing Robot," IEEE Robotics Automation Letters Vol. 6, No. 2, 2021, pp. 2114-2121

    2021

  14. [20]

    Bio -inspired rapid escape and tight body flip on an at -scale flapping wing hummingbird robot via reinforcement learning,

    Tu, Z., Fei, F., and Deng, X. "Bio -inspired rapid escape and tight body flip on an at -scale flapping wing hummingbird robot via reinforcement learning," IEEE Transactions on Robotics Vol. 37, No. 5, 2021, pp. 1742-1751. doi: 10.1109/TRO.2021.3064882

    2021

  15. [22]

    Sensitivity Analysis of Wing Geometric and Kinematic Parameters for the Aerodynamic Performance of Hovering Flapping Wing,

    Lang, X., Song, B., Yang, W., Yang, X., and Xue, D. "Sensitivity Analysis of Wing Geometric and Kinematic Parameters for the Aerodynamic Performance of Hovering Flapping Wing," Aerospace Vol. 10, No. 1, 2023, p. 74. doi: https://doi.org/10.3390/aerospace10010074

    2023 doi

  16. [23]

    Flapping trajectory characteristics and attitude control approach of a flapping -wing robot with 2 -DOF parallel mechanism,

    Wang, L., Song, B., Sun, Z., and Yang, X. "Flapping trajectory characteristics and attitude control approach of a flapping -wing robot with 2 -DOF parallel mechanism," Aerospace Science and Technology , 2024, p. 109020. doi: https://doi.org/10.1016/j.ast.2024.109020

    2024

  17. [24]

    Design and verification of large -scaled flapping wings for high altitude environment,

    Zhu, Z., Song, B., and Xue, D. "Design and verification of large -scaled flapping wings for high altitude environment," Applied Sciences Vol. 12, No. 10, 2022, p. 5140. doi: https://doi.org/10.3390/app12105140

    2022 doi

  18. [25]

    Effect of wing membrane material on the aerodynamic performance of flexible flapping wing,

    Lang, X., Song, B., Yang, W., and Yang, X. "Effect of wing membrane material on the aerodynamic performance of flexible flapping wing," Applied Sciences Vol. 12, No. 9, 2022, p. 4501. doi: https://doi.org/10.3390/app12094501

    2022 doi

  19. [26]

    Phasing of dragonfly wings can improve aerodynamic efficiency by removing swirl,

    Usherwood, J. R., and Lehmann, F.-O. "Phasing of dragonfly wings can improve aerodynamic efficiency by removing swirl," Journal of The Royal Society Interface Vol. 5, No. 28, 2008, pp. 1303-1307. doi: https://doi.org/10.1098/rsif.2008.0124

    2008

  20. [27]

    Experimental study on forewing –hindwing phasing in hovering and forward flapping flight,

    Nagai, H., Fujita, K., and Murozono, M. "Experimental study on forewing –hindwing phasing in hovering and forward flapping flight," AIAA journal Vol. 57, No. 9, 2019, pp. 3779-3790. doi: https://doi.org/10.2514/1.J058335 27

    2019 doi

  21. [28]

    Robot fine-tuning made easy: Pre-training rewards and policies for autonomous real -world reinforcement learning,

    Yang, J., Mark, M. S., Vu, B., Sharma, A., Bohg, J., and Finn, C. "Robot fine-tuning made easy: Pre-training rewards and policies for autonomous real -world reinforcement learning," 2024 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2024, pp. 4804-4811

    2024

  22. [29]

    Fully online meta-learning without task boundaries,

    Rajasegaran, J., Finn, C., and Levine, S. "Fully online meta-learning without task boundaries," arXiv preprint arXiv:2202.00263, 2022. doi: https://doi.org/10.48550/arXiv.2202.00263

    2022 doi arXiv

  23. [30]

    How to train your robot with deep reinforcement learning: lessons we have learned,

    Ibarz, J., Tan, J., Finn, C., Kalakrishnan, M., Pastor, P., and Levine, S. "How to train your robot with deep reinforcement learning: lessons we have learned," The International Journal of Robotics Research Vol. 40, No. 4-5, 2021, pp. 698-721. doi: https://doi.org/10.1177/0278...

    2021 doi

  24. [31]

    End -to-end training of deep visuomotor policies,

    Levine, S., Finn, C., Darrell, T., and Abbeel, P. "End -to-end training of deep visuomotor policies," The Journal of Machine Learning Research Vol. 17, No. 1, 2016, pp. 1334-1373. doi: https://doi.org/10.48550/arXiv.1504.00702

    2016 doi arXiv

  25. [32]

    Aerodynamic performance of a free-flying dragonfly—A span- resolved investigation,

    Hefler, C., Noda, R., Qiu, H., and Shyy, W. "Aerodynamic performance of a free-flying dragonfly—A span- resolved investigation," Physics of Fluids Vol. 32, No. 4, 2020, p. 041903

    2020

  26. [33]

    Feedback Control for Directional Rendezvous Using Constant -Magnitude Low Thrust,

    Balakrishnan, S., and Gurfil, P. "Feedback Control for Directional Rendezvous Using Constant -Magnitude Low Thrust," Journal of Guidance, Control, and Dynamics, 2024, pp. 1-15. doi: https://doi.org/10.2514/1.G008096

    2024 doi

  27. [34]

    Adaptive Flying Assistance Controller Design to Suppress Nonlinear Pilot-Induced Oscillations,

    Jiao, J., Sun, L., Tan, W., Liu, X., and Shang, Z. "Adaptive Flying Assistance Controller Design to Suppress Nonlinear Pilot-Induced Oscillations," Journal of Guidance, Control, and Dynamics Vol. 47, No. 10, 2024, pp. 2133-2147. doi: https://doi.org/10.2514/1.G008241

    2024 doi

  28. [35]

    New Equilibria and Dynamic Structures Under Continuous Optimal Feedback Control,

    Tsuruta, A., Bando, M., Hokamoto, S., and Scheeres, D. J. "New Equilibria and Dynamic Structures Under Continuous Optimal Feedback Control," Journal of Guidance, Control, and Dynamics Vol. 47, No. 10, 2024, pp. 2029-2040. doi: https://doi.org/10.2514/1.G008270

    2024 doi

  29. [36]

    Boosted Incremental Nonlinear Dynamic Inversion for Flexible Airplane Gust Load Alleviation,

    Beyer, Y., Steen, M., and Hecker, P. "Boosted Incremental Nonlinear Dynamic Inversion for Flexible Airplane Gust Load Alleviation," Journal of Guidance, Control, and Dynamics, 2024, pp. 1-20. doi: https://doi.org/10.2514/1.G007984

    2024 doi

  30. [37]

    Instantaneous wing kinematics tracking and force control of a high - frequency flapping wing insect MAV,

    Zhang, J., Cheng, B., and Deng, X. "Instantaneous wing kinematics tracking and force control of a high - frequency flapping wing insect MAV," Journal of Micro-Bio Robotics Vol. 11, 2016, pp. 67-84. doi: https://doi.org/10.1007/s12213-015-0085-4

    2016 doi

  31. [38]

    Scalable deep reinforcement learning for vision -based robotic manipulation,

    Kalashnikov, D., Irpan, A., Pastor, P., Ibarz, J., Herzog, A., Jang, E., Quillen, D., Holly, E., Kalakrishnan, M., and Vanhoucke, V. "Scalable deep reinforcement learning for vision -based robotic manipulation," Conference on robot learning. PMLR, 2018, pp. 651-673

    2018

  32. [39]

    Learning hand-eye coordination for robotic grasping with deep learning and large-scale data collection,

    Levine, S., Pastor, P., Krizhevsky, A., Ibarz, J., and Quillen, D. "Learning hand-eye coordination for robotic grasping with deep learning and large-scale data collection," The International journal of robotics research Vol. 37, No. 4-5, 2018, pp. 421-436. doi: https://doi.org...

    2018 doi arXiv

  33. [40]

    Distributional soft actor -critic: Off-policy reinforcement learning for addressing value estimation errors,

    Duan, J., Guan, Y., Li, S. E., Ren, Y., Sun, Q., and Cheng, B. "Distributional soft actor -critic: Off-policy reinforcement learning for addressing value estimation errors," IEEE transactions on neural networks and learning systems Vol. 33, No. 11, 2021, pp. 6584-6598. doi: ht...

    2021

  34. [41]

    Centralized cooperation for connected and automated vehicles at intersections by proximal policy optimization,

    Guan, Y., Ren, Y., Li, S. E., Sun, Q., Luo, L., and Li, K. "Centralized cooperation for connected and automated vehicles at intersections by proximal policy optimization," IEEE Transactions on Vehicular Technology Vol. 69, No. 11, 2020, pp. 12597-12608. doi: https://doi.org/10...

    2020 doi arXiv

  35. [42]

    Twin -delayed deep deterministic policy gradient algorithm for the energy management of microgrids,

    Domí nguez-Barbero, D., Garcí a-Gonzá lez, J., and Sanz-Bobi, M. Á. "Twin -delayed deep deterministic policy gradient algorithm for the energy management of microgrids," Engineering Applications of Artificial Intelligence Vol. 125, 2023, p. 106693. doi: https://doi.org/10.1016...

    2023

  36. [43]

    Deep reinforcement learning: A brief survey,

    Arulkumaran, K., Deisenroth, M. P., Brundage, M., and Bharath, A. A. "Deep reinforcement learning: A brief survey," IEEE Signal Processing Magazine Vol. 34, No. 6, 2017, pp. 26-38. doi: 10.1109/MSP.2017.2743240 28

    2017

  37. [44]

    Optimal and autonomous control using reinforcement learning: A survey,

    Kiumarsi, B., Vamvoudakis, K. G., Modares, H., and Lewis, F. L. "Optimal and autonomous control using reinforcement learning: A survey," IEEE transactions on neural networks and learning systems Vol. 29, No. 6, 2017, pp. 2042-2062

    2017

  38. [45]

    Formation control scheme with reinforcement learning strategy for a group of multiple surface vehicles,

    Nguyen, K., Dang, V. T., Pham, D. D., and Dao, P. N. "Formation control scheme with reinforcement learning strategy for a group of multiple surface vehicles," International Journal of Robust and Nonlinear Control Vol. 34, No. 3, 2024, pp. 2252-2279. doi: https://doi.org/10.339...

    2024 doi

  39. [46]

    Reinforcement learning ‐based optimal trajectory tracking control of surface vessels under input saturations,

    Wei, Z., and Du, J. "Reinforcement learning ‐based optimal trajectory tracking control of surface vessels under input saturations," International Journal of Robust and Nonlinear Control Vol. 33, No. 6, 2023, pp. 3807-3825. doi: https://doi.org/10.1002/rnc.6597

    2023 doi

  40. [47]

    Real -Time Model -Free Deep Reinforcement Learning for Force Control of a Series Elastic Actuator,

    Sambhus, R., Gokce, A., Welch, S., Herron, C. W., and Leonessa, A. "Real -Time Model -Free Deep Reinforcement Learning for Force Control of a Series Elastic Actuator," 2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2023, pp. 5645-5652

    2023

  41. [48]

    Learning real-time dynamic responsive gap-traversing policy for quadrotors with safety-aware exploration,

    Chen, S., Li, Y., Lou, Y., Lin, K., and Wu, X. "Learning real-time dynamic responsive gap-traversing policy for quadrotors with safety-aware exploration," IEEE Transactions on Intelligent Vehicles Vol. 8, No. 3, 2022, pp. 2271-2284. doi: 10.1109/TIV.2022.3229723

    2022

  42. [49]

    A Real-Time and Optimal Hypersonic Entry Guidance Method Using Inverse Reinforcement Learning,

    Su, L., Wang, J., and Chen, H. "A Real-Time and Optimal Hypersonic Entry Guidance Method Using Inverse Reinforcement Learning," Aerospace Vol. 10, No. 11, 2023, p. 948. doi: https://doi.org/10.3390/aerospace10110948

    2023 doi

  43. [50]

    Realizing a deep reinforcement learning agent for real-time quantum feedback,

    Reuer, K., Landgraf, J., Fösel, T., O’Sullivan, J., Beltrán, L., Akin, A., Norris, G. J., Remm, A., Kerschbaum, M., and Besse, J.-C. "Realizing a deep reinforcement learning agent for real-time quantum feedback," Nature Communications Vol. 14, No. 1, 2023, p. 7138. doi: https:...

    2023 doi

  44. [51]

    Online Corrections to Neural Policy Guidance for Pinpoint Powered Descent,

    Cho, N., Shin, H. -S., Tsourdos, A., and Amato, D. "Online Corrections to Neural Policy Guidance for Pinpoint Powered Descent," Journal of Guidance, Control, and Dynamics Vol. 47, No. 5, 2024, pp. 945-963. doi: https://doi.org/10.2514/1.G007234

    2024 doi

  45. [52]

    (More) efficient reinforcement learning via posterior sampling,

    Osband, I., Russo, D., and Van Roy, B. "(More) efficient reinforcement learning via posterior sampling," Advances in Neural Information Processing Systems Vol. 26, 2013

    2013

  46. [53]

    Neuroscience -inspired artificial intelligence,

    Hassabis, D., Kumaran, D., Summerfield, C., and Botvinick, M. "Neuroscience -inspired artificial intelligence," Neuron Vol. 95, No. 2, 2017, pp. 245-258. doi: http://dx.doi.org/10.1016/j.neuron.2017.06.011

    2017 doi

  47. [54]

    Evolutionary implications of neural circuit structure and function,

    Tierney, A. "Evolutionary implications of neural circuit structure and function," Behavioural processes Vol. 35, No. 1-3, 1995, pp. 173-182. doi: 10.1016/0376-6357(95)00041-0

    1995 doi

  48. [55]

    Neural adaptation in the generation of rhythmic behavior,

    Pearson, K. "Neural adaptation in the generation of rhythmic behavior," Annual review of physiology Vol. 62, No. 1, 2000, pp. 723-753. doi: 10.1146/annurev.physiol.62.1.723

    2000 doi

  49. [56]

    Learning contact -rich manipulation skills with guided policy search,

    Levine, S., Wagener, N., and Abbeel, P. "Learning contact -rich manipulation skills with guided policy search," arXiv preprint arXiv:.05611, 2015

    2015

  50. [57]

    Reinforcement Learning Control of Hypersonic Vehicles and Performance Evaluations,

    Wallace, B. A., and Si, J. "Reinforcement Learning Control of Hypersonic Vehicles and Performance Evaluations," Journal of Guidance, Control, and Dynamics, 2024, pp. 1-14. doi: https://doi.org/10.2514/1.G008225

    2024 doi

  51. [58]

    Deep L 1 Stochastic Optimal Control Policies for Planetary Soft Landing,

    Pereira, M. A., Duarte, C. A., Theodorou, E. A., and Exarchos, I. "Deep L 1 Stochastic Optimal Control Policies for Planetary Soft Landing," Journal of Guidance, Control, and Dynamics Vol. 47, No. 4, 2024, pp. 651-665. doi: https://doi.org/10.2514/1.G007132

    2024 doi

  52. [59]

    Autonomous navigation of mobile robots in unknown environments using off -policy reinforcement learning with curriculum learning,

    Yin, Y., Chen, Z., Liu, G., Yin, J., and Guo, J. "Autonomous navigation of mobile robots in unknown environments using off -policy reinforcement learning with curriculum learning," Expert Systems with Applications Vol. 247, 2024, p. 123202. doi: https://doi.org/10.1016/j.eswa....

    2024

  53. [60]

    Adaptive policy learning for data -driven powertrain control with eco-driving,

    Kerbel, L., Ayalew, B., and Ivanco, A. "Adaptive policy learning for data -driven powertrain control with eco-driving," Engineering Applications of Artificial Intelligence Vol. 124, 2023, p. 106489. doi: https://doi.org/10.1016/j.engappai.2023.106489

    2023

  54. [61]

    Toward biomorphic robotics: A review on swimming central pattern generators,

    Tsybina, Y. A., Gordleeva, S. Y., Zharinov, A., Kastalskiy, I., Ermolaeva, A., Hramov, A., and Kazantsev, V. "Toward biomorphic robotics: A review on swimming central pattern generators," Chaos, Solitons & Fractals Vol. 165, 2022, p. 112864. doi: https://doi.org/10.1016/j.chao...

    2022

  55. [63]

    Application of an augmented Lagrangian approach to multibody systems with equality motion constraints,

    Potosakis, N., Paraskevopoulos, E., and Natsiavas, S. "Application of an augmented Lagrangian approach to multibody systems with equality motion constraints," Nonlinear Dynamics Vol. 99, No. 1, 2020, pp. 753 - 776. doi: https://doi.org/10.1007/s11071-019-05059-6

    2020 doi

  56. [64]

    Hamiltonian formulation with reduced variables for flexible multibody systems under linear constraints: Theory and experiment,

    Dong, S., Otsuka, K., and Makihara, K. "Hamiltonian formulation with reduced variables for flexible multibody systems under linear constraints: Theory and experiment," Journal of Sound and Vibration Vol. 547, 2023, p. 117535. doi: https://doi.org/10.1016/j.jsv.2022.117535

    2023

  57. [65]

    Adapting Rapid Motor Adaptation for Bipedal Robots,

    Kumar, A., Li, Z., Zeng, J., Pathak, D., Sreenath, K., and Malik, J. "Adapting Rapid Motor Adaptation for Bipedal Robots," 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). 2022, pp. 1161-1168

    2022

  58. [66]

    Rigid -body kinematics versus flapping kinematics of a flapping wing micro air vehicle,

    Caetano, J., Weehuizen, M., De Visser, C., De Croon, G., and Mulder, M. "Rigid -body kinematics versus flapping kinematics of a flapping wing micro air vehicle," Journal of Guidance, Control, and Dynamics Vol. 38, No. 12, 2015, pp. 2257-2269. doi: https://doi.org/10.2514/1.G000923

    2015 doi

  59. [67]

    Prescribed -time convergence with input constraints: A control Lyapunov function based approach,

    Garg, K., Arabi, E., and Panagou, D. "Prescribed -time convergence with input constraints: A control Lyapunov function based approach," 2020 American Control Conference (ACC). IEEE, 2020, pp. 962-967

    2020

  60. [68]

    Deterministic policy gradient algorithms,

    Silver, D., Lever, G., Heess, N., Degris, T., Wierstra, D., and Riedmiller, M. "Deterministic policy gradient algorithms," International conference on machine learning. Pmlr, 2014, pp. 387-395

    2014

  61. [69]

    S., and Barto, A

    Sutton, R. S., and Barto, A. G. Reinforcement learning: An introduction: MIT press, 2018

    2018

  62. [70]

    A tale of two -timescale reinforcement learning with the tightest finite-time bound,

    Dalal, G., Szorenyi, B., and Thoppe, G. "A tale of two -timescale reinforcement learning with the tightest finite-time bound," Proceedings of the AAAI Conference on Artificial Intelligence. Vol. 34, 2020, pp. 3701- 3708

    2020

  63. [71]

    Sample complexity bounds for two timescale value -based reinforcement learning algorithms,

    Xu, T., and Liang, Y. "Sample complexity bounds for two timescale value -based reinforcement learning algorithms," International Conference on Artificial Intelligence and Statistics. PMLR, 2021, pp. 811-819

    2021

  64. [72]

    Dynamic load balancing for real-time video encoding on heterogeneous CPU+ GPU systems,

    Momcilovic, S., Ilic, A., Roma, N., and Sousa, L. "Dynamic load balancing for real-time video encoding on heterogeneous CPU+ GPU systems," IEEE Transactions on Multimedia Vol. 16, No. 1, 2013, pp. 108-121. doi: 10.1109/TMM.2013.2284892

    2013

  65. [73]

    Unified stability analysis for Itô stochastic systems: From almost surely asymptotic to finite-time convergence,

    Luo, S., Deng, F., and Yu, X. "Unified stability analysis for Itô stochastic systems: From almost surely asymptotic to finite-time convergence," IEEE Transactions on Automatic Control Vol. 67, No. 1, 2021, pp. 406-412. doi: 10.1109/TAC.2021.3057990

    2021

  66. [74]

    Modular and hierarchically modular organization of brain networks,

    Meunier, D., Lambiotte, R., and Bullmore, E. T. "Modular and hierarchically modular organization of brain networks," Frontiers in neuroscience Vol. 4, 2010, p. 7572. doi: https://doi.org/10.3389/fnins.2010.00200

    2010 arXiv

  67. [75]

    Evolving Reservoirs for Meta Reinforcement Learning,

    Lé ger, C., Hamon, G., Nisioti, E., Hinaut, X., and Moulin -Frier, C. "Evolving Reservoirs for Meta Reinforcement Learning," International Conference on the Applications of Evolutionary Computation (Part of EvoStar). Springer, 2024, pp. 36-60

    2024

  68. [76]

    Mitigating Time -Delay in Nonlinear Dynamics Inversion for Multirotor Unmanned Aerial Vehicles,

    Jang, K., Bang, H., and Kim, Y. "Mitigating Time -Delay in Nonlinear Dynamics Inversion for Multirotor Unmanned Aerial Vehicles," Journal of Guidance, Control, and Dynamics, 2024, pp. 1-15. doi: https://doi.org/10.2514/1.G007567

    2024 doi

  69. [77]

    Vehicle yaw stability model predictive control strategy for dynamic and multi-objective requirements

    Wang, H., Wu, J., Chen, Z., He, R., and Li, H. "Vehicle yaw stability model predictive control strategy for dynamic and multi-objective requirements." SAE Technical Paper, 2024

    2024

  70. [78]

    On Robust Reinforcement Learning with Lipschitz - Bounded Policy Networks,

    Barbara, N. H., Wang, R., and Manchester, I. R. "On Robust Reinforcement Learning with Lipschitz - Bounded Policy Networks," arXiv preprint arXiv:2405.11432, 2024. 30 doi: https://doi.org/10.48550/arXiv.2405.11432

    2024 doi arXiv

  71. [79]

    Array programming with NumPy,

    Harris, C. R., Millman, K. J., Van Der Walt, S. J., Gommers, R., Virtanen, P., Cournapeau, D., Wieser, E., Taylor, J., Berg, S., and Smith, N. J. "Array programming with NumPy," Nature Vol. 585, No. 7825, 2020, pp. 357-362. doi: https://doi.org/10.1038/s41586-020-2649-2

    2020 doi

  72. [80]

    Flexible multibody impact simulations based on the isogeometric analysis approach,

    Rü ckwald, T., Held, A., and Seifried, R. "Flexible multibody impact simulations based on the isogeometric analysis approach," Multibody system dynamics Vol. 54, 2022, pp. 75-95. doi: https://doi.org/10.1007/s11044-021-09804-x

    2022 doi

  73. [81]

    Simulation and stability analysis of periodic flexible multibody systems,

    Han, S., and Bauchau, O. A. "Simulation and stability analysis of periodic flexible multibody systems," Multibody System Dynamics Vol. 50, No. 4, 2020, pp. 381-413. doi: https://doi.org/10.1007/s11044-020-09741-1

    2020 doi

  74. [82]

    Symbolic discovery of optimization algorithms,

    Chen, X., Liang, C., Huang, D., Real, E., Wang, K., Pham, H., Dong, X., Luong, T., Hsieh, C.-J., and Lu, Y. "Symbolic discovery of optimization algorithms," Advances in Neural Information Processing Systems Vol. 36, 2024

    2024

  75. [83]

    Combined MRAC for unknown MIMO LTI systems with parameter convergence,

    Roy, S. B., Bhasin, S., and Kar, I. N. "Combined MRAC for unknown MIMO LTI systems with parameter convergence," IEEE Transactions on Automatic Control Vol. 63, No. 1, 2017, pp. 283-290. doi: 10.1109/TAC.2017.2725955

    2017

  76. [84]

    An at -scale tailless flapping-wing hummingbird robot. I. Design, optimization, and experimental validation,

    Tu, Z., Fei, F., Zhang, J., and Deng, X. "An at -scale tailless flapping-wing hummingbird robot. I. Design, optimization, and experimental validation," IEEE Transactions on Robotics Vol. 36, No. 5, 2020, pp. 1511- 1525. doi: 10.1109/TRO.2020.2993217

    2020

  77. [85]

    Design optimization and system integration of robotic hummingbird,

    Zhang, J., Fei, F., Tu, Z., and Deng, X. "Design optimization and system integration of robotic hummingbird," 2017 IEEE International Conference on Robotics and Automation (ICRA). IEEE, Singapore, 2017, pp. 5422- 5428

    2017

  78. [86]

    Flappy hummingbird: An open source dynamic simulation of flapping wing robots and animals,

    Fei, F., Tu, Z., Yang, Y., Zhang, J., and Deng, X. "Flappy hummingbird: An open source dynamic simulation of flapping wing robots and animals," 2019 International Conference on Robotics and Automation (ICRA) . IEEE, 2019, pp. 9223-9229

    2019

  79. [87]

    A flight strategy for intelligent aerial vehicles learned from dragonfly,

    Hu, Z., and Deng, X. "A flight strategy for intelligent aerial vehicles learned from dragonfly," Aerial Vehicles. 2009, pp. 189-202

    2009

  80. [88]

    A survey of imitation learning: Algorithms, recent developments, and challenges,

    Zare, M., Kebria, P. M., Khosravi, A., and Nahavandi, S. "A survey of imitation learning: Algorithms, recent developments, and challenges," IEEE Transactions on Cybernetics, 2024. doi: 10.1109/TCYB.2024.3395626

    2024

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.