Pith. sign in

REVIEW 4 major objections 7 minor 39 references

Gray-Box Computed Torque Control for Differential-Drive Mobile Robot Tracking

T0 review · 4 major / 7 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read A gray-box computed torque controller learns a differential-drive robot's dynamics and tracking gains in 11 short episodes, using a reinforcement learning agent to tune only eight physical parameters instead of a neural-network policy.

desk verdict A neat structured-policy idea for DDMR control, but the stability guarantee is only formal under exact parameter knowledge and the experiments lack numbers; worth review after major revision. read the letter →

arxiv 2509.00571 v1 pith:S7I5UJF7 submitted 2025-08-30 cs.RO cs.SYeess.SY

Arman Javan Sekhavat Pishkhani (University of Tehran , Tehran , Iran) This is my paper · ORCID
classification cs.ROcs.SYeess.SY
keywords differential-drivemobilerobotcomputedtorquecontrolgray-boxreinforcementlearningTD3trajectorytrackingcriticaldampingMuJoCo
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

This paper proposes a middle path between model-based computed torque control and model-free deep reinforcement learning for differential-drive mobile robots. The controller keeps the structure of computed torque—a model-based law that can cancel the robot's nonlinear dynamics—but leaves eight physical parameters to be learned by the TD3 reinforcement learning algorithm instead of requiring them to be known in advance. The paper claims this gray-box design can identify the system and tune the controller within 11 episodes of five seconds each, and that hard constraints on the learned parameters keep them physically plausible while the chosen pole configuration guarantees a stable, critically damped tracking error. In simulation, the learned controller tracks three test trajectories better than a Lyapunov-based kinematic controller and nearly as well as a computed torque controller given exact parameters. If the claim holds, it means accurate dynamic tracking can be obtained with very little interaction data and without a prior parameter-identification step.

What carries the argument

The central object is the gray-box policy: a computed torque control law whose eight tunable parameters (σ1..σ4, cV, cD, α, β) replace the actor network of TD3. The physics is encoded in M(θ) and C(ω) derived from the DDMR equations of motion, and in the friction terms τ_V and τ_D; the learning is encoded in a tanh reparameterization that keeps σ and friction parameters inside physically plausible ranges while leaving the optimization unconstrained. The pole-placement formulas kp = 3(α²+ε)², ki = (α²+ε)³, kd = 3(α²+ε), and their β counterparts, enforce the equal-pole configuration that Lemma 1 shows gives the fastest non-oscillatory response. Substitution of the control law into the plant dy

What would settle it

Run the trained GCTC on a real differential-drive robot or in a higher-fidelity simulator with actuator delay, then intentionally perturb one learned parameter (for example, cV) by increasing amounts and record the tracking-error response; the guarantee would be disproved if small perturbations produce overshoot, oscillation, or instability that the error equation (52) does not predict.

Watch

Extended reading notes

Core claim

The central claim is that replacing the neural-network policy of a reinforcement learning agent with a reparameterized computed torque controller—the gray-box GCTC—lets a single learning process identify the robot's dynamic parameters and synthesize a high-performance tracking controller. The control law takes the form u = M(θ)(p̈_d + Kp e + Ki ∫e + Kd ė) + C(ω) + viscous and Coulomb friction terms, where M, C, and the friction coefficients are expressed through four positive lumped parameters σ1..σ4 and two friction coefficients cV, cD. Only these plus two pole-placement parameters α, β are learned. Because the controller structure cancels the modeled nonlinearities, the tracking error obey

Load-bearing premise

The guarantee of stability and critical damping depends on the learned parameters exactly reproducing the true dynamic and friction coefficients; if they do not, the linear error equation that justifies the guarantee does not hold, and the paper does not analyze this mismatch.

Editorial extensions

If this is right

  • A single short training budget of 11 episodes of 5 seconds on one sinusoid yields a controller that transfers to qualitatively different trajectories, suggesting structured parameter search is far more sample-efficient than black-box policy search.
  • The exact-parameter CTC is treated as an upper bound; GCTC approaches it, implying that dynamic model knowledge can be recovered online during control rather than through a separate identification experiment.
  • Because the trained controller is just the computed torque law with fixed coefficients, it can run at high frequency on low-cost hardware in real time.
  • The pole-placement construction gives closed-loop tracking error dynamics that are independent of the trajectory shape, as long as the learned parameters are correct.
  • Adding integral action and friction compensation reduces steady-state error and improves tracking of high-velocity trajectories compared to kinematic control.
  • The learning process simultaneously identifies the plant and designs the controller, so no prior identification step is needed.
  • The learned controller can be deployed on a low-cost microcontroller because the resulting policy has low computational complexity, unlike a neural-network policy.

Reading between the lines

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

  • The stability and critical-damping guarantee is exact only in the nominal case: under parameter mismatch the error equation gains extra terms, so the practical claim is conditional on learned parameters closely matching the true ones; the paper does not quantify the mismatch region.
  • A direct validation would be to compare the learned σi, cV, cD against independently identified physical parameters; strong agreement would confirm the gray-box premise, while disagreement would indicate the performance comes from something other than exact cancellation.
  • The same gray-box treatment could be applied to other mechanical systems with known computed-torque structure, such as manipulators, whenever the unknown parameters enter the control law in a structured way.
  • The Lemma 1 argument relies on linear time-invariant error dynamics; if unmodeled effects such as wheel slip or actuator dynamics become significant, the guaranteed critical damping may not hold on real hardware.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 7 minor

Summary. The paper proposes Gray-box Computed Torque Control (GCTC) for differential-drive mobile robot tracking. The method replaces the neural-network policy of a TD3 reinforcement-learning agent with a structured computed-torque controller whose physical parameters (mass/inertia-like terms, viscous and Coulomb friction coefficients) and feedback gains are learned. The controller is parameterized so that the learned gains enforce a triple-pole, critically damped error response when substituted into the nominal dynamics. The authors report training on a single sinusoidal trajectory for only 11 episodes of 5 s and evaluating on three other trajectories in MuJoCo, comparing with a CTC using exact parameters and a kinematic controller. The central claims are that GCTC finds near-optimal controller parameters with very few episodes and that the added constraints guarantee a stable, critically damped closed-loop response.

Significance. If the central claims were fully supported, the paper would make a useful contribution: it gives a compact analytic CTC parameterization for a DDMR with friction, embeds physically motivated box constraints via a tanh parameterization, and demonstrates that a structured policy can be learned much more sample-efficiently than a black-box neural policy. The algebraic derivation of the nominal error dynamics is a strength, and the use of a gray-box policy in TD3 is a sensible direction. However, the principal stability guarantee is derived under an exact-parameter cancellation assumption that is not satisfied by the learned controller, and the experimental section does not report the quantitative data needed to verify the claimed performance or sample efficiency. The significance of the paper therefore rests on an unproven and, as stated, incorrect guarantee.

major comments (4)
  1. [Sections 3.2 and 3.3, Eqs. (48) and (52)] The derivation of the error dynamics (52) requires that the matrices M, C and friction coefficients c_V, c_D in the control law exactly match the true plant parameters. In GCTC these are trainable and are never shown to coincide with the plant. With estimates M_hat, C_hat, cV_hat, cD_hat, substituting (48) into (43) gives a persistent disturbance on the right-hand side, e.g. terms such as (M_hat - M) p_dd, (C_hat - C), and (cV_hat - cV) Lambda p_dot. Equation (52) therefore does not hold, the asymptotic convergence argument is void, and the critical-damping conclusion based on (54) is unsupported. The paper provides no bound on the mismatch disturbance, no Lyapunov/ISS analysis, and no experimental evidence that the learned parameters converge to the true values. The abstract and Section 5 claim that the constraints 'guarantee a stable and critically damped closed-loop time response'; th
  2. [Section 4, Figures 2-4] The experimental section contains no numerical tracking-error values, no learning curves, no control-effort data, and no quantitative comparison with the baselines. The claims that GCTC 'outperforms' the raw CTC with exact parameters and the kinematic controller are not supported by the figures alone, which show only trajectories without axes labels, legends, or error statistics. This is especially problematic because the exact-parameter CTC should, by Eq. (52), produce zero tracking error; any comparison that claims GCTC is better requires careful numerical reporting. Without such data, the central performance claim is unverifiable.
  3. [Lemma 1, Eqs. (59)-(60)] The proof of Lemma 1 is not correct as written. In Eq. (59) the term |F(t)| need not remain bounded away from zero for large t: if the coefficient of the monomial corresponding to the largest eigenvalue alpha_tilde is zero for the particular initial condition, then |F(t)| decays to zero rather than satisfying |F(t)| >= M1 > 0. The argument therefore does not establish the claimed fastest non-oscillatory property for all initial conditions. Since this lemma is the basis for the triple-pole placement in Eqs. (65)-(67), the critical-damping rationale needs to be either rigorously repaired or replaced by a direct and correct argument.
  4. [Section 4 and Algorithm 1] The paper does not report the reward weight matrices H_e and H_u, the parameter ranges (centers and radii), TD3 hyperparameters, initial parameter values, episode termination thresholds, torque limits, or the number of random seeds. The sample-efficiency claim ('only a few short learning episodes') is central to the paper and cannot be assessed without these details. The absence of quantitative learning curves and seed statistics makes the reported 11-episode training result irreproducible.
minor comments (7)
  1. [Notation throughout] The notation is inconsistent in places: the policy is denoted mu_Pi and also Pi, and the parameter vector Pi is used both for the policy parameters and for the matrix in Eq. (39). Please unify the notation.
  2. [Section 3.2, Eq. (54)] Taking vector powers elementwise in Eq. (54) is nonstandard and should be stated more explicitly, or the scalar equations for the two axes should be separated.
  3. [Section 3.2, Eq. (65)] The phrase 'fastest non-oscillatory time-response' is defined only through Lemma 1, whose proof has a gap (see Major Comment 3). Consider rephrasing the pole-placement criterion as a design choice supported by numerical evidence.
  4. [Section 2.3, Eq. (43)] The friction model in Eq. (45) uses sign(Lambda p_dot), but at p_dot = 0 the sign function is discontinuous; the paper does not discuss the well-posedness of the closed-loop dynamics at zero wheel velocity.
  5. [Figures 2-4] The trajectory plots lack axis labels and legends. It is impossible to read the scale of the tracking errors or distinguish the controllers clearly. A zoomed-in error plot would be much more informative.
  6. [References] Reference [5] is cited as 'Mobile Robots', which appears to be a book chapter; the full bibliographic information is missing. Please provide the complete citation.
  7. [Algorithm 1] The while condition uses ||e(t)|| but the infinity norm or a specific threshold is not defined. Please specify how the tracking-error threshold is measured.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the derivation chain is self-contained and the evaluation uses trajectories not used for fitting.

full rationale

The paper's central derivation is the computed-torque control law (48) substituted into the dynamic model (43) to obtain the error dynamics (52). This is a conditional algebraic result: it holds when the controller parameters M, C, cV, cD exactly equal the plant parameters. The stability and critical-damping conclusions follow from the characteristic equation (54) and the pole-placement choice (65)–(67). This is a standard control-design construction, not a fitted parameter renamed as a prediction. The training and evaluation are separated: parameters are learned on one sinusoidal trajectory, and the claims are tested on a different sinusoid, a circle, and a square, so there is no fitted-input-called-prediction pattern. There are no self-citations that are load-bearing, no imported uniqueness theorems, and no ansatz smuggled in via citation. The tanh reparameterization is a standard constraint-handling transformation. The main weakness is that the learned parameters are not guaranteed to match the true plant, so the nominal derivation (52) may not hold during actual deployment; however, this is a robustness/soundness limitation, not a circularity. Overall, the derivation chain is self-contained and not circular.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The paper introduces no new physical entities. The key free parameters are the learned controller coefficients and the manually set reward weights/range bounds. The axiomatic load is mostly the standard DDMR dynamic model plus the questionable Lemma 1 and the exact-parameter match assumption for the stability proof.

free parameters (6)
  • sigma_1, sigma_2, sigma_3, sigma_4 = Learned via TD3, not reported
    These four constants parameterize the inertia matrix M(theta) and are trained to maximize the reward; they are not derived from physical measurement.
  • c_V, c_D = Learned via TD3, not reported
    Viscous and Coulomb friction coefficients, treated as trainable parameters with guessed ranges.
  • alpha, beta = Learned via TD3, not reported
    Pole placement parameters in the characteristic polynomial, learned to set Kp, Ki, Kd.
  • epsilon = Chosen arbitrarily, value not reported
    Positive scalar to keep closed-loop poles strictly negative; described as non-trainable but chosen by hand.
  • Reward weights H_e, H_u = Not reported
    Positive semi-definite and positive definite matrices in the reward function, chosen by hand and not specified.
  • Parameter range centers and radii = Guessed, not reported
    Centers and radii of the tanh reparameterization (70)-(72), chosen manually to constrain learned values.
assumptions (4)
  • domain assumption DDMR dynamic model (14)-(23) with constants C_i
    The model is cited from prior work [12] and assumed accurate for the simulated robot.
  • domain assumption Nonholonomic constraint p_dot = Lambda_tilde Phi_dot (Eq. 7)
    Standard nonholonomic constraint for a differential-drive robot with no wheel slip.
  • ad hoc to paper Lemma 1: equal real poles give the fastest non-oscillatory response
    The lemma is proved in the paper, but the proof implicitly assumes nonzero initial conditions and that |F(t)| is eventually bounded below; it is not generally true and is used to justify the critical-damping parameterization.
  • ad hoc to paper Exact parameter match to obtain error equation (52)
    The closed-loop error equation and its stability are derived by substituting the control law into the dynamics assuming M, C, cV, cD exactly match the plant; this contradicts the paper's stated setup of unknown parameters.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Gray-Box Computed Torque Control for Differential-Drive Mobile Robot Tracking." pith.science (2026). https://pith.science/paper/S7I5UJF7

@misc{pith2026250900571,
  author       = {Pith},
  title        = {Pith review of: Gray-Box Computed Torque Control for Differential-Drive Mobile Robot Tracking},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/S7I5UJF7}},
  note         = {Machine review of arXiv:2509.00571}
}
read the original abstract

This study presents a learning-based nonlinear algorithm for tracking control of differential-drive mobile robots. The Computed Torque Method (CTM) suffers from inaccurate knowledge of system parameters, while Deep Reinforcement Learning (DRL) algorithms are known for sample inefficiency and weak stability guarantees. The proposed method replaces the black-box policy network of a DRL agent with a gray-box Computed Torque Controller (CTC) to improve sample efficiency and ensure closed-loop stability. This approach enables finding an optimal set of controller parameters for an arbitrary reward function using only a few short learning episodes. The Twin-Delayed Deep Deterministic Policy Gradient (TD3) algorithm is used for this purpose. Additionally, some controller parameters are constrained to lie within known value ranges, ensuring the RL agent learns physically plausible values. A technique is also applied to enforce a critically damped closed-loop time response. The controller's performance is evaluated on a differential-drive mobile robot simulated in the MuJoCo physics engine and compared against the raw CTC and a conventional kinematic controller.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

39 extracted references · 31 canonical work pages

  1. [1]

    A review of mobile robots: Concepts, methods, theoretical framework, and applications,

    F. Rubio, F. Valero, and C. Llopis-Albert, “A review of mobile robots: Concepts, methods, theoretical framework, and applications,” International Journal of Advanced Robotic Systems, vol. 16, no. 2, p. 1729881419839596, Mar. 2019, doi: 10.1177/1729881419839596

    work page doi:10.1177/1729881419839596 2019

  2. [2]

    A Comprehensive Study of Mobile Robot: History, Developments, Applications, and Future Research Perspectives,

    R. Raj and A. Kos, “A Comprehensive Study of Mobile Robot: History, Developments, Applications, and Future Research Perspectives,” Applied Sciences, vol. 12, no. 14, p. 6951, Jul. 2022, doi: 10.3390/app12146951

    work page doi:10.3390/app12146951 2022

  3. [3]

    Teleoperation of a mobile robot through the Internet,

    R. L. Alves, R. F. Vassallo, E. O. Freire, and T. F. Bastos -Filho, “Teleoperation of a mobile robot through the Internet,” in Proceedings of the 43rd IEEE Midwest Symposium on Circuits and Systems (Cat.No.CH37144), Lansing, MI, USA: IEEE, 2000, pp. 930–933. doi: 10.1109/MWSCAS.2000.952906

    work page arXiv

  4. [4]

    Design and Control for Differential Drive Mobile Robot,

    B. D. Hirpo and P. W. Zhongmin, “Design and Control for Differential Drive Mobile Robot,” International Journal of Engineering Research & Technology, vol. 6, no. 10, Oct. 2017, doi: 10.17577/IJERTV6IS100138

    work page doi:10.17577/ijertv6is100138 2017

  5. [5]

    Mobile Robots,

    “Mobile Robots,” in Robotics, London: Springer London, 2010, pp. 469 –521. doi: 10.1007/978 -1- 84628-642-1_11

    doi:10.1007/978 2010

  6. [6]

    Localization of the Closed -Loop Differential Drive Mobile Robot Using Wheel Odometry,

    G. Singh and V. Kumar, “Localization of the Closed -Loop Differential Drive Mobile Robot Using Wheel Odometry,” in Soft Computing: Theories and Applications, R. Kumar, A. K. Verma, T. K. Sharma, O. P. Verma, and S. Sharma, Eds., Singapore: Springer Nature, 2023, pp. 809–817. doi: 10.1007/978-981- 19-9858-4_69

    doi:10.1007/978-981- 2023

  7. [7]

    Pose Estimation of a Differential -Drive Mobile Robot Using EKF -Based Sensor Fusion,

    F.-R. Martínez -Algarín, I. Santos -Ruiz, F. -R. López -Estrada, and J. -E. Domínguez -Zenteno, “Pose Estimation of a Differential -Drive Mobile Robot Using EKF -Based Sensor Fusion,” in 2024 XXVI Robotics Mexican Congress (COMRob), Nov. 2024, pp. 111 –116. doi: 10.1109/COMRob64055.2024.10777435

    work page arXiv 2024

  8. [8]

    MPC -Based Path Tracking of a Differential -Drive Mobile Robot with Optimization for Improved Control Performance,

    D. A. Marrugo and J. L. Villa, “MPC -Based Path Tracking of a Differential -Drive Mobile Robot with Optimization for Improved Control Performance,” in Applied Computer Sciences in Engineering, J. C. Figueroa-García, G. Hernández, J. L. Villa Ramirez, and E. E. Gaona García, Eds., Cham: Springer Nature Switzerland, 2023, pp. 276–289. doi: 10.1007/978-3-031...

    work page doi:10.1007/978-3-031-46739-4_25 2023

Show all 39 references
  1. [9]

    Computed-Torque Control of a Wheeled Mobile Manipulator,

    B. Mohamed, M. Neila, and D. Tarak, “Computed-Torque Control of a Wheeled Mobile Manipulator,” Int J Robot Eng, vol. 3, no. 1, Jun. 2018, doi: 10.35840/2631-5106/4105

    2018 doi

  2. [10]

    Hierarchical Fuzzy Sliding -Mode Adaptive Control for the Trajectory Tracking of Differential-Driven Mobile Robots,

    H.-M. Wu and M. Karkoub, “Hierarchical Fuzzy Sliding -Mode Adaptive Control for the Trajectory Tracking of Differential-Driven Mobile Robots,” Int. J. Fuzzy Syst., vol. 21, no. 1, pp. 33 –49, Feb. 2019, doi: 10.1007/s40815-018-0531-2

    2019 doi

  3. [11]

    Trajectory tracking control of a nonholonomic mobile robot with differential drive,

    R. L. S. Sousa, M. D. do Nascimento Forte, F. G. Nogueira, and B. C. Torrico, “Trajectory tracking control of a nonholonomic mobile robot with differential drive,” in 2016 IEEE Biennial Congress of Argentina (ARGENCON), Jun. 2016, pp. 1–6. doi: 10.1109/ARGENCON.2016.7585356

    2016

  4. [12]

    Dynamic Modelling of Differential -Drive Mobile Robots using Lagrange and Newton-Euler Methodologies: A Unified Framework,

    R. D. Ahmad Abu Hatab , “Dynamic Modelling of Differential -Drive Mobile Robots using Lagrange and Newton-Euler Methodologies: A Unified Framework,” Adv Robot Autom, vol. 02, no. 02, 2013, doi: 10.4172/2168-9695.1000107

    2013

  5. [13]

    Resolved -acceleration control of mechanical manipulators,

    J. Luh, M. Walker, and R. Paul, “Resolved -acceleration control of mechanical manipulators,” IEEE Transactions on Automatic Control, vol. 25, no. 3, pp. 468 –474, Jun. 1980, doi: 10.1109/TAC.1980.1102367

    1980

  6. [14]

    J. J. Craig, Introduction to robotics: mechanics and control, Fourth edition. New York, NY: Pearson, 2018

    2018

  7. [15]

    Parametric Identification and Controller Design for a Differential-Drive Mobile Robot⁎,

    T. G. Alves, W. F. Lages, and R. V. B. Henriques, “Parametric Identification and Controller Design for a Differential-Drive Mobile Robot⁎,” IFAC-PapersOnLine, vol. 51, no. 15, pp. 437–442, Jan. 2018, doi: 10.1016/j.ifacol.2018.09.184

    2018 doi

  8. [16]

    Identification of Differential Drive Robot Dynamic Model Parameters,

    M. Siwek, J. Panasiuk, L. Baranowski, W. Kaczmarek, P. Prusaczyk, and S. Borys, “Identification of Differential Drive Robot Dynamic Model Parameters,” Materials, vol. 16, no. 2, p. 683, Jan. 2023, doi: 10.3390/ma16020683

    2023 doi

  9. [17]

    Constrained model predictive control: Stability and optimality,

    D. Q. Mayne, J. B. Rawlings, C. V. Rao, and P. O. M. Scokaert, “Constrained model predictive control: Stability and optimality,” Automatica, vol. 36, no. 6, pp. 789 –814, Jun. 2000, doi: 10.1016/S0005 - 1098(99)00214-9

    doi

  10. [18]

    Reinforcement Learning-Enhanced Active Disturbance Rejection Control for Mobile Robot Trajectory Tracking,

    H. Usama, M. Nasir, R. Fareh, J. Ghommam, S. Khadraoui, and M. Bettayeb, “Reinforcement Learning-Enhanced Active Disturbance Rejection Control for Mobile Robot Trajectory Tracking,” in 2025 IEEE 22nd International Multi-Conference on Systems, Signals & Devices (SSD), Feb. 2025...

    2025

  11. [19]

    Integration of Q -Learning and PID Controller for Mobile Robots Trajectory Tracking in Unknown Environments,

    A. Munaf and A. R. J. Almusawi, “Integration of Q -Learning and PID Controller for Mobile Robots Trajectory Tracking in Unknown Environments,” JESA, vol. 57, no. 4, pp. 1023 –1033, Aug. 2024, doi: 10.18280/jesa.570410

    2024 doi

  12. [20]

    Robust Optimal Tracking Control for Wheel Mobile Robot via Reinforcement Learning,

    N. T. Dang and D. T. Duong, “Robust Optimal Tracking Control for Wheel Mobile Robot via Reinforcement Learning,” Journal of Robotics and Control (JRC), vol. 6, no. 3, pp. 1502–1509, Jun. 2025, doi: 10.18196/jrc.v6i3.26659

    2025 doi

  13. [21]

    Playing Atari with Deep Reinforcement Learning,

    V. Mnih et al., “Playing Atari with Deep Reinforcement Learning,” Dec. 19, 2013, arXiv: arXiv:1312.5602. doi: 10.48550/arXiv.1312.5602

    2013 doi arXiv

  14. [22]

    Continuous control with deep reinforcement learning,

    T. P. Lillicrap et al., “Continuous control with deep reinforcement learning,” Jul. 05, 2019, arXiv: arXiv:1509.02971. doi: 10.48550/arXiv.1509.02971

    2019 doi arXiv

  15. [23]

    Addressing Function Approximation Error in Actor -Critic Methods,

    S. Fujimoto, H. van Hoof, and D. Meger, “Addressing Function Approximation Error in Actor -Critic Methods,” Oct. 22, 2018, arXiv: arXiv:1802.09477. doi: 10.48550/arXiv.1802.09477

    2018 doi arXiv

  16. [24]

    R. S. Sutton and A. G. Barto, Reinforcement learning: an introduction, Second edition. in Adaptive computation and machine learning series. Cambridge, Massachusetts: The MIT Press, 2018

    2018

  17. [25]

    Reinforcement learning for control: Performance, stability, and deep approximators,

    L. Buşoniu, T. De Bruin, D. Tolić, J. Kober, and I. Palunko, “Reinforcement learning for control: Performance, stability, and deep approximators,” Annual Reviews in Control, vol. 46, pp. 8–28, 2018, doi: 10.1016/j.arcontrol.2018.09.005

    2018 doi

  18. [26]

    Deep Grey-Box Modeling With Adaptive Data-Driven Models Toward Trustworthy Estimation of Theory -Driven Models,

    N. Takeishi and A. Kalousis, “Deep Grey-Box Modeling With Adaptive Data-Driven Models Toward Trustworthy Estimation of Theory -Driven Models,” Oct. 24, 2022, arXiv: arXiv:2210.13103. doi: 10.48550/arXiv.2210.13103

    2022 doi arXiv

  19. [27]

    Residual Reinforcement Learning for Robot Control,

    T. Johannink et al., “Residual Reinforcement Learning for Robot Control,” Dec. 18, 2018, arXiv: arXiv:1812.03201. doi: 10.48550/arXiv.1812.03201

    2018 doi arXiv

  20. [28]

    PI and PID Controller Tuning with Deep Reinforcement Learning,

    K. B. Trujillo, J. G. Álvarez, and E. Cortés, “PI and PID Controller Tuning with Deep Reinforcement Learning,” in 2022 IEEE International Conference on Automation/XXV Congress of the Chilean Association of Automatic Control (ICA -ACCA), Oct. 2022, pp. 1 –6. doi: 10.1109/ICA - ...

    2022

  21. [29]

    Tuning of PID Controllers Using Reinforcement Learning for Nonlinear System Control,

    G. Bujgoi and D. Sendrescu, “Tuning of PID Controllers Using Reinforcement Learning for Nonlinear System Control,” Processes, vol. 13, no. 3, p. 735, Mar. 2025, doi: 10.3390/pr13030735

    2025 doi

  22. [30]

    MuJoCo: A physics engine for model -based control,

    E. Todorov, T. Erez, and Y. Tassa, “MuJoCo: A physics engine for model -based control,” in 2012 IEEE/RSJ International Conference on Intelligent Robots and Systems, Oct. 2012, pp. 5026 –5033. doi: 10.1109/IROS.2012.6386109

    2012

  23. [31]

    Compiling machine learning programs via high -level tracing

    “Compiling machine learning programs via high -level tracing.” Accessed: Aug. 29, 2025. [Online]. Available: https://research.google/pubs/compiling-machine-learning-programs-via-high-level-tracing/

    2025

  24. [32]

    Deterministic Policy Gradient Algorithms,

    D. Silver, G. Lever, N. Heess, T. Degris, D. Wierstra , and M. Riedmiller, “Deterministic Policy Gradient Algorithms,” in Proceedings of the 31st International Conference on Machine Learning, PMLR, Jan. 2014, pp. 387 –395. Accessed: Aug. 29, 2025. [Online]. Available: https://...

    2014

  25. [33]

    Kinematics, Localization and Control of Differential Drive Mobile Robot

    Sandeep Kumar Malu and Jharna Majumdar, “Kinematics, Localization and Control of Differential Drive Mobile Robot”, GJRE, vol. 14, no. H1, pp. 1–7, Jan. 2014

    2014

  26. [34]

    Path Planning and Trajectory Tracking Control for Two-Wheel Mobile Robot,

    I. A. Hassan, I. A. Abed, and W. A. Al-Hussaibi, “Path Planning and Trajectory Tracking Control for Two-Wheel Mobile Robot,” Journal of Robotics and Control (JRC), vol. 5, no. 1, pp. 1 –15, 2024, doi: 10.18196/jrc.v5i1.20489

    2024 doi

  27. [35]

    N. S. Nise, Control systems engineering, Seventh edition. Hoboken, NJ: Wiley, 2015

    2015

  28. [36]

    D. T. Greenwood, Advanced dynamics, Repr. with corr. Cambridge: Cambridge Univ. Press, 2006

    2006

  29. [37]

    Brown and A

    B. Brown and A. Zai, Deep reinforcement learning in action. Shelter Island, NY: Manning Publications Company, 2020

    2020

  30. [38]

    An algebraic and suboptimal solution of constrained model predictive control via tangent hyperbolic function,

    U. Dursun, F. Yıldız Taşçıkaraoğlu, and İ. Üstoğlu, “An algebraic and suboptimal solution of constrained model predictive control via tangent hyperbolic function,” Asian Journal of Control, vol. 23, no. 5, pp. 2420–2430, Sep. 2021, doi: 10.1002/asjc.2357

    2021 doi

  31. [39]

    Tip-Over Detection and Avoidance Algorithms as Stabilization Strategy for Small -Footprint and Lightweight Mobile Manipulators,

    A. Toledo Fuentes, F. Kempf, M. Kipfmüller, T. Bergmann, and M. J. Prieto, “Tip-Over Detection and Avoidance Algorithms as Stabilization Strategy for Small -Footprint and Lightweight Mobile Manipulators,” Machines, vol. 11, no. 1, p. 44, Dec. 2022, doi: 10.3390/machines11010044

    2022 doi

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.