Pith. sign in

REVIEW 3 major objections 5 minor 34 references

Actuator dead-zones can be cancelled by one fixed data projection, with no plant model and no dead-zone parameters.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-31 16:32 UTC pith:76H2VSTA

load-bearing objection Clean architectural fix for dead-zone blindness in velocity-form SPC; theory has one real gap on how the proxy Hankel is filled, but hardware is real and the paper deserves referees. the 3 major comments →

arxiv 2607.28142 v1 pith:76H2VSTA submitted 2026-07-30 eess.SY cs.SY

Data-Driven Dead-Zone Compensation via Projection in Predictive Control Setting

Mahmood Mazare , Hossein Ramezani This is my paper
classification eess.SY cs.SY
keywords data-driven controldead-zone compensationsubspace predictive controlbehavioral systems theorydisturbance estimationoffset-free trackingHankel projection
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Actuator dead-zones leave motion systems stuck with a steady-state offset: inside a band of commanded effort the plant simply does not move. Velocity-form data-driven predictive controllers hide that signature by integrating around the mismatch, so they never see the drop-off they are compensating. This paper shows that a second, absolute subspace predictor—built from the same offline trajectories but without integral action—turns a persistent mismatch into a proportional steady-state residual. Embedding that residual as a proxy inside a behavioral Hankel matrix reduces mismatch estimation to one offline-computed orthogonal projection, evaluated online as a single matrix-vector product with no dynamic observer, no probing signal, and no run-time residual calculation. Plugged into a subspace predictive controller, the loop stays recursively feasible and practically input-to-state stable, and recovers exact offset-free tracking once the dead-band crossing settles. Real-time tests on a lightly damped multi-DOF torsion plant and a high-precision power amplifier confirm the result across physical dead-bands.

Core claim

The steady-state error of an absolute subspace predictor maps the unknown actuator mismatch through a fixed, data-identified gain. Embedding that residual as a proxy in a behavioral Hankel matrix lets a single offline orthogonal projection extract the mismatch online, so a velocity-form subspace predictive controller can shift its baseline effort across the dead-band and recover asymptotic offset-free tracking without modeling the plant or parameterizing the dead-zone.

What carries the argument

The fixed observer matrix L_d: the orthogonal projection of the rolling input-output window onto the left null space of a Hankel matrix that already contains the absolute-predictor mismatch proxy. It collapses prediction, residual generation, and mismatch extraction into one matrix-vector product.

Load-bearing premise

Over the short observer window the physical mismatch must look roughly constant, and the offline secondary data must already contain natural dead-zone crossings so the Hankel matrix spans the mismatched trajectories.

What would settle it

Hold the torsion plant near the dead-band edge, inject a load disturbance that makes the effective mismatch jump faster than the observer horizon, and check whether tracking error returns to zero after the window fills with the new value; persistent offset after that window would refute the claim.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Unknown actuator dead-zones can be rejected without identifying a nonlinear inverse or tuning a dynamic disturbance observer.
  • Once control increments settle and the mismatch becomes constant, tracking becomes asymptotically offset-free, not merely ultimately bounded.
  • The same projection handles both mechanical static-friction dead-bands and electronic dead-time dead-bands.
  • Online cost collapses to one matrix-vector product with no injected test signals and no run-time prediction-error loop.

Where Pith is reading between the lines

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

  • The same absolute-residual projection should apply to other slowly varying actuator faults that leave a DC signature, such as bias drift or soft saturation edges.
  • Observer horizon length trades noise averaging against the constant-mismatch assumption; a multi-scale or adaptive horizon could widen the regime where the bound holds.
  • When there are fewer outputs than inputs, only the output-observable component of multi-channel mismatch is recovered, so simultaneous multi-actuator dead-bands may need extra sensors or sequential excitation.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes a model-free dead-zone compensator for subspace predictive control. Alongside a velocity-form (incremental) predictor used for offset-free SPC, an absolute subspace predictor is identified from the same nominal data; its steady-state residual is shown to equal Φd under LTI dynamics and constant mismatch (Eqs. 10–16). That residual is mapped to a proxy d_proxy=Φ†e_spc, embedded as a block in a secondary behavioral Hankel matrix H, and the online mismatch estimate is reduced to one fixed left-nullspace projection L_d (Eq. 22) with no dynamic observer or probing. The integrated loop is claimed to be recursively feasible and practically ISS (Lemmas 1–2, Theorems 1–2), recovering asymptotic offset-free tracking once dead-band traversal settles. Real-time validation on a Quanser multi-DOF torsion plant (±0.18 V dead-band) and a simulated high-precision power amplifier are reported.

Significance. If the construction is sound, the contribution is practically useful: dead-zone compensation without plant model, dead-zone parameterization, injected excitation, or run-time residual filtering—only a single offline-computed matrix-vector product. Pairing an absolute “DC sensor” predictor with a velocity-form SPC controller is a clean architectural idea, and the hardware result on a lightly damped sixth-order torsion system strengthens the case beyond pure simulation. The stability argument follows standard robust-MPC/ISS patterns once a bounded estimation error is granted. The main value is therefore methodological and experimental rather than a new stability theory; the load-bearing novelty sits in the residual-as-proxy → fixed projection pipeline.

major comments (3)
  1. [§4.2–5.1, Lemma 1] §4.2–4.3 and §5.1 (Eqs. 16–19, 22; Lemma 1): The only rigorous identity linking residual to physical mismatch is the steady-state map e_ss=Φd under constant u_ss, y_ss, d. Section 4.3 nevertheless evaluates d_proxy(k)=Φ†e_spc(k) at every sample of the secondary trajectory—including dead-band crossings—and stacks the resulting D_p into H. During transients the one-step absolute residual also contains free response and state error, so those columns are not pure constant-d behaviors. Lemma 1’s claim that w_true∈col(H) (hence ‖d̃‖≤ε_d linear in σ_d) therefore needs an extra, currently unstated justification: either that transient proxies remain accurate enough not to distort the column space, or that quasi-steady dead-zone samples dominate the span of H. Without that, the disturbance bound feeding Theorems 1–2 is not secured by the given derivation. Please either (i) restrict the secondary d
  2. [Assumption 1, Remark 1, §7.2] Assumption 1 and Remark 1 vs. experimental regimes: The observer treats d as locally constant over N_o, with ε_d∝σ_d. On the torsion plant the dead-band is essentially static friction/thresholding (piecewise constant once settled), which fits; on the amplifier, dead-time effects and the forced drop-off interval can change on a timescale comparable to N_o/f_s=30 μs. The paper should state clearly for which class of mismatches (static dead-zone vs. fast blanking/dead-time) the rate condition and the secondary-data excitation requirement are expected to hold, and what fails if the secondary trajectory never dwells inside the dead-band long enough to populate H with accurate proxies.
  3. [Theorem 2, §7] §6 / Theorem 2: Offset-free recovery is argued by σ_d→0 once Δu→0 and the command stops traversing the dead-band, so ε_d→0 and integral action finishes the job. This is plausible for a static dead-zone after settling, but it is not automatic under persistent reference switching (square wave in §7.1) or repeated forced drop-offs (§7.2): each crossing re-excites σ_d>0. Please make the “once traversal settles” regime precise (e.g., constant or slowly varying references after a finite number of crossings) and, if possible, report a quantitative estimation-error or residual metric alongside the tracking plots so that the practical size of ε_d is visible.
minor comments (5)
  1. [§4.2, §7.2] When p<m (amplifier, §4.2 last paragraph), Φ† recovers only the observable component of d. State explicitly in the problem formulation and in §7.2 that single-channel excitation is assumed, and that simultaneous multi-channel dead-bands are out of scope.
  2. [§6.1, Eq. (23b)] P_d in the SPC equality (23b) is said to be “obtained directly from the identified predictor blocks” but is never written. Give the block expression (analogous to Φ) so the QP is reproducible.
  3. [§3.1–3.2] Offline identification assumes a disturbance-free dataset (d≡0). A short remark on sensitivity to mild offline mismatch or noise, and on the role of Tikhonov λ, would help practitioners.
  4. [§7 figures] Figures 1–3 are conceptual and helpful; ensure axis labels/units on the experimental Figs. 5 and 7 are readable in print and that the ±0.18 V dead-band is marked on the u_cmd plot.
  5. Minor notation: u_eff vs u_eff, and consistent use of cmd/spc/tr subscripts; also arXiv ID year 2026 looks like a placeholder—verify metadata.

Circularity Check

1 steps flagged

No load-bearing circularity: proxy-to-projection chain is a data construction with a steady-state identity extended off-equilibrium, not a result forced by its own inputs.

specific steps
  1. other [§4.2–4.3, Eqs. 16–19; Lemma 1]
    "e_ss = [P^{(1)}_{1,a}(1_{T_ini} ⊗ I_m)] d. ... d_proxy(k) = Φ† e_spc(k). ... For each discrete time step k ∈ [T_ini+1, T_tr], we compute the prediction error e_spc(k) via (9) and map it to the mismatch proxy d_proxy(k) via (18). ... By Willems' Fundamental Lemma, any valid trajectory of the LTI system must reside in the column space of the offline Hankel matrix H. Thus, w_true(k) ∈ col(H)."

    Not strong circularity: the only rigorous residual-to-mismatch identity is the steady-state map (16), yet §4.3 fills D_p (and thus H) with d_proxy at every sample of a dynamic secondary trajectory. Lemma 1 then treats true constant-d trajectories as lying in that column space. This extends a steady-state definition off-equilibrium without a separate transient error bound; it is a validity gap in the construction of H, not a case where the claimed estimate or tracking result equals its inputs by algebra. Online L_d[u_ini; y_ini] and the hardware offset-free recovery remain independent of that definitional stretch.

full rationale

The derivation identifies absolute and incremental predictors from a disturbance-free offline dataset (Eqs. 5, 7), derives the steady-state map e_ss = Φ d under constant u_ss, y_ss, d (Eq. 16), builds a secondary Hankel H from operational proxies d_proxy = Φ† e_spc (Eqs. 18–19), and reduces online estimation to the fixed left-nullspace map L_d (Eq. 22). Recursive feasibility and practical ISS (Theorems 1–2) then follow from a bounded-rate assumption on d and a Minkowski RPI terminal set—standard MPC arguments, not tautologies. Closed-loop claims are checked on hardware (Quanser torsion, ±0.18 V dead-band) and a power-amplifier simulation against an external baseline (Lazar et al.), so outcomes are not forced by the fit. There are no self-citations by Mazare/Ramezani, no uniqueness theorem imported from the authors, and no fitted parameter renamed as a prediction of the same quantity. The only mild definitional tension is that H is populated with d_proxy computed from the steady-state identity at every time index of dynamic trajectories; Lemma 1’s claim w_true ∈ col(H) therefore leans on an unstated extra regularity of those columns. That is an assumption/correctness gap (whether transient residuals still span the constant-d behaviors), not circularity: the online estimate is still a nontrivial map from new (u_ini, y_ini) data, and offset-free recovery is attributed to velocity-form integral action once σ_d → 0, which is independent content. Score 1 only for that mild embedding stretch; central claims do not reduce by construction.

Axiom & Free-Parameter Ledger

5 free parameters · 7 axioms · 2 invented entities

The claim rests on behavioral LTI trajectory span (Willems), standard MPC terminal-set/ISS machinery, and several domain modeling choices: mismatch-free offline PE data, slowly varying d over N_o, existence of a secondary trajectory that already hits the dead-zone, and p≥m or single-channel excitation for Φ†. Free parameters are the usual SPC/regularization/horizon/weight knobs plus the secondary-data construction. No new physical entity is postulated; ‘DDESO’ and d_proxy are algorithmic constructs.

free parameters (5)
  • Tikhonov regularization λ
    Used in closed-form absolute and incremental predictor fits (Eqs. 5, 7); value not reported; conditions the data inverse.
  • Horizons T_ini, N, N_o and data lengths T_d, T_tr = Case1: N=20,T_ini=15,N_o=15; Case2: N=15,T_ini=12,N_o=12
    Chosen by designer; N_o trades noise vs constant-d assumption (Remark 1). Case 1: N=20, T_ini=15, N_o=15; Case 2: N=15, T_ini=12, N_o=12.
  • SPC weights Q, R and input/rate bounds = Torsion: Q=50, R=1e5; Amp: Q=10, R=0.1 I_2
    Hand-tuned quadratic costs and constraints shape closed-loop response and feasibility region (Eq. 23).
  • Terminal set Ω_f / local gain K_f
    Existence assumed for RPI construction (Lemma 2); concrete computation on hardware not detailed.
  • Secondary operational dataset for H (proxy-mapped)
    Must contain natural dead-zone traversals; how it is collected and how much excitation is used is only qualitatively described (§4.3).
axioms (7)
  • domain assumption Unknown plant is discrete-time LTI in effective input u_eff; nonlinear loss is lumped as additive mismatch d(k) (Eqs. 1–3).
    Section 2; standard linear embedding of actuator nonlinearity as input disturbance.
  • domain assumption Assumption 1: ‖d(k)−d(k−1)‖≤σ_d; approximately constant over observer window N_o.
    Invoked for regressor construction and Lemma 1 error bound; load-bearing for exact recovery when σ_d→0.
  • standard math Willems’ fundamental lemma: PE offline trajectories span finite behavior; composite w(d) lies in col(H).
    Cited [30]; used for predictors and projection residual=0 idealization (§§3,5).
  • domain assumption Offline identification data collected with d≡0; secondary tr data contains real mismatches without injected probes.
    §3.1 and §4.3; if violated, Φ and H encode the wrong behavior.
  • domain assumption Φ=P^(1)_{1,a}(1⊗I_m) full column rank when p≥m (or single-channel excitation when p<m) so Φ† recovers usable d_proxy.
    §4.2; explicitly weakened for the SISO-output amplifier.
  • standard math Existence of local stabilizing K_f and compact RPI terminal set Ω_f for bounded estimation-error set W.
    Lemma 2; standard set-theoretic MPC hypothesis, not constructed numerically in the paper.
  • ad hoc to paper E⊤PE invertible so L_d is well-defined.
    §5.1; required for unique least-squares projection solution; depends on data richness and selector E.
invented entities (2)
  • Absolute-predictor mismatch proxy d_proxy=Φ† e_spc independent evidence
    purpose: Replace parameterized dead-zone inverse / dynamic disturbance state with a static data-driven residual map.
    Defined in §4 from absolute SPC residual; algorithmic construct, not a new physical quantity.
  • Fixed projection observer matrix L_d (DDESO) independent evidence
    purpose: Online mismatch estimate as one matrix-vector product without run-time prediction-error dynamics.
    Derived in §5 as L_d=−(E⊤PE)^{−1}E⊤P S_uy; name ‘DDESO’ is branding for this static map.

pith-pipeline@v1.2.0-daily-grok45 · 24456 in / 4342 out tokens · 87206 ms · 2026-07-31T16:32:37.490723+00:00 · methodology

0 comments
read the original abstract

Actuator dead-zones are a common and troublesome nonlinearity in motion control: a band of commanded effort over which the plant does not respond, leaving a steady-state offset or a limit cycle. This paper proposes a data-driven architecture that compensates such mismatches without a model of the plant and without any parameterization of the dead-zone. The central idea is to identify, alongside the velocity-form predictor used for control, a second absolute subspace predictor. Because the absolute predictor carries no integral action, it behaves as a data-driven steady-state sensor, so a persistent actuator mismatch appears as a proportional prediction residual. Embedding this residual as a proxy in a behavioral Hankel matrix reduces the mismatch estimate to a single fixed orthogonal projection evaluated online, with no dynamic estimator, no injected probing signal, and no run-time prediction-error computation. Integrated into a subspace predictive controller, the framework is shown to be recursively feasible and practically input-to-state stable, and it recovers offset-free tracking once the dead-band traversal settles. The approach is validated in real time on a sixth-order, lightly damped Quanser multi-DOF torsion system, whose complex-conjugate poles give a lightly damped open-loop response, achieving offset-free tracking across a $\pm 0.18$\,V actuator dead-band. A second study on a high-precision power amplifier shows that the same architecture rejects dead-time-induced nonlinearities in fast-switching power electronics.

Figures

Figures reproduced from arXiv: 2607.28142 by Hossein Ramezani, Mahmood Mazare.

Figure 1
Figure 1. Figure 1: Geometric comparison of steady-state mappings. (a) The absolute predictor captures the dead-zone as a non-zero prediction residual 𝑒𝑠𝑠 proportional to the mismatch 𝑑. (b) The incremental predictor’s variables (Δ𝑢, Δ𝑦) wash out the constant offset, causing the nominal and mismatched trajectory spaces to overlap and destroying the dead-zone signature. col(𝐻) Origin 𝑤0 𝑃 𝑤0 𝐸𝑑̂(𝑘) 𝑤(𝑑̂) = 𝑤0 + 𝐸𝑑̂(𝑘) [PITH_F… view at source ↗
Figure 2
Figure 2. Figure 2: Geometric interpretation of the Data-Driven Extended State Observer (DDESO). The measured trajectory 𝑤0 lies outside the physical data subspace col(𝐻). The observer extracts the optimal mismatch proxy 𝑑̂(𝑘) such that the composite vector 𝑤(𝑑̂) is orthogonally projected back onto the valid trajectory space. M. Mazare and H. Ramezani: Preprint submitted to Elsevier Page 17 of 16 [PITH_FULL_IMAGE:figures/ful… view at source ↗
Figure 3
Figure 3. Figure 3: Set-theoretic validation of the observer-driven control loop. Without dead-zone estimation, the controller must absorb the maximal physical deadband profile (𝑑max), dictating an immensely conservative robust terminal set Ωuncomp plagued by steady-state offsets. Via the proposed algebraic projection, the tracking dynamics shrink down to the nominal space expanded only by the minimal residual estimation bou… view at source ↗
Figure 4
Figure 4. Figure 4: The Quanser Multi-DOF Torsion cyber-physical testbed. The system features a servo base flexibly coupled to multiple inertial loads, presenting a highly oscillatory open-loop response compounded by static Coulomb friction and actuator dead-bands. M. Mazare and H. Ramezani: Preprint submitted to Elsevier Page 18 of 16 [PITH_FULL_IMAGE:figures/full_fig_p018_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Real-time experimental results on the Quanser Multi-DOF Torsion system. (Top) Tracking performance, showing the steady-state offset during the uncompensated phase (0–20 s) and the correction after the proposed DDESO is engaged (20–60 s). (Bottom) The control signal 𝑢cmd responding to the estimated dead-zone proxy and crossing the ±0.18V dead￾band. 𝑉𝑏𝑢𝑠 𝑆1 𝑆2 𝑉𝑏𝑢𝑠 𝑆3 𝑆4 𝑖 𝐿 𝐿𝑝 𝐿 𝑖𝐿𝑛 𝑖 𝑅𝑚 𝑜 𝐿𝑚 𝑅 𝑣𝐶𝑝 𝐶 𝑖𝐶𝑝 𝑅 … view at source ↗
Figure 6
Figure 6. Figure 6: Schematic of the industrial dual-stage current amplifier, adapted from [34]. Each half-bridge leg (𝑆1 , 𝑆2 and 𝑆3 , 𝑆4 ) drives a series inductor 𝐿; the output current 𝑖 𝑜 flows through the load 𝑅𝑚–𝐿𝑚 , with 𝑅–𝐶 snubber branches at each node. M. Mazare and H. Ramezani: Preprint submitted to Elsevier Page 19 of 16 [PITH_FULL_IMAGE:figures/full_fig_p019_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Closed-loop performance of the baseline iSPC [34] and the proposed DDESO-MPC on the high-precision power amplifier (simulation). The system tracks a 5A reference under an unmodeled dead-zone applied between 𝑡 = 6 × 10−4 s and 10 × 10−4 s. The DDESO reduces the transient offset seen in the baseline controller by shifting the baseline control effort. M. Mazare and H. Ramezani: Preprint submitted to Elsevier … view at source ↗

discussion (0)

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

Reference graph

Works this paper leans on

34 extracted references

  1. [1]

    Y. Liu, Y. Zhang, L. Liu, H. Li, H. Liu, Adaptive dynamic surface control for a class of dead-zone nonlinear systems via output feedback, IEEE Access (2021)

    2021

  2. [2]

    Zhang, Q

    Y. Zhang, Q. Yan, J. Cai, X. Wu, Adaptive iterative learning control for tank gun servo systems with input deadzone, IEEE Access (2020)

    2020

  3. [3]

    Q. Shou, H. Xia, Q. Yan, J. Cai, Y. Ma, Z. Bao, Robust learning control for a class of uncertain linear motor systems, J. Phys.: Conf. Ser. (2020)

    2020

  4. [4]

    L. Li, Z. Lin, Y. Jiang, C. Yu, J. Yao, Valve deadzone/backlash compensation for lifting motion control of hydraulic manipulators, Machines (2021)

    2021

  5. [5]

    X. Wang, Y. Ma, Q. Yan, Y. Zhang, X. Guan, Robust ilc of nonlinearly parametric time-delay systems with input deadzone, J. Phys.: Conf. Ser. (2021)

    2021

  6. [6]

    Zhang, Q

    H. Zhang, Q. Yan, J. Cai, S. Gao, Y. Zhang, Initial-rectification neuro-adaptive iterative learning control for robot manipulators with input deadzone and nonzero initial errors, IEEE Access (2023)

    2023

  7. [7]

    Y. Yu, L. Wang, Adaptive fault-tolerant finite-time flight-path angle control for aircraft systems with unknown deadzone and actuator faults, IEEE Access (2024)

    2024

  8. [8]

    L. Cao, S. Liu, L. Xu, An intelligent fault-tolerant control method for a flexible-link manipulator with an uncertain dead-zone and intermittent actuator faults, Mathematics (2024)

    2024

  9. [9]

    X. Su, Z. Liu, Direct adaptive compensation control of mechanical systems with unknown actuator failures and dead-zone nonlinearities (2018)

    2018

  10. [10]

    W. M. Bessa, M. Dutra, E. Kreuzer, An adaptive fuzzy dead-zone compensation scheme and its application to electro-hydraulic systems (2010)

    2010

  11. [11]

    W. M. Bessa, M. Dutra, E. Kreuzer, An adaptive fuzzy dead-zone compensation scheme for nonlinear systems, arXiv preprint (2022)

    2022

  12. [12]

    W. M. Bessa, An adaptive fuzzy sliding mode controller for nonlinear systems with non-symmetric dead-zone and its application to an electro- hydraulic system, arXiv preprint (2022)

    2022

  13. [13]

    D. Jung, J. W. Jeon, Synchronous control of 2-d.o.f master-slave manipulators using actuators with asymmetric nonlinear dead-zone characteristics, IEEE Access (2022)

    2022

  14. [14]

    Chan-Zheng, P

    C. Chan-Zheng, P. Borja, J. Scherpen, Dead-zone compensation via passivity-based control for a class of mechanical systems, IFAC- PapersOnLine (2022)

    2022

  15. [15]

    X. Feng, C. Wang, Adaptive tracking control for a class of uncertain mimo nonlinear systems with input constraints, J. Intell. Robot. Syst. (2025). M. Mazare and H. Ramezani: Preprint submitted to Elsevier Page 15 of 16 Data-Driven Dead-Zone Compensation

    2025

  16. [16]

    A. C. T. de Cequeira, M. Vale, D. G. V . Fonseca, F. M. U. Araújo, A. Maitelli, Estimation and compensation of dead-zone inherent to the actuators of industrial processes, in: Int. Conf. Informatics in Control, Automation and Robotics, 2010

    2010

  17. [17]

    Sunar, et al., Improved pole-placement control with feed-forward dead zone compensation for position tracking of electro-pneumatic actuator system, Elektrika (2021)

    N. Sunar, et al., Improved pole-placement control with feed-forward dead zone compensation for position tracking of electro-pneumatic actuator system, Elektrika (2021)

    2021

  18. [18]

    Wang, et al., Active fault-tolerant control for the dual-valve hydraulic system with unknown dead-zone, ISA Trans

    T. Wang, et al., Active fault-tolerant control for the dual-valve hydraulic system with unknown dead-zone, ISA Trans. (2023)

    2023

  19. [19]

    Jia, et al., A robust adaptive trajectory tracking algorithm using smc and machine learning for ffsgrs with actuator dead zones, Appl

    L. Jia, et al., A robust adaptive trajectory tracking algorithm using smc and machine learning for ffsgrs with actuator dead zones, Appl. Sci. (2019)

    2019

  20. [20]

    Q. Zhou, S. Zhao, H. Li, R. Lu, C. Wu, Adaptive neural network tracking control for robotic manipulators with dead zone, IEEE Trans. Neural Netw. Learn. Syst. (2019)

    2019

  21. [21]

    L. Ma, M. Wang, Adaptive compensation tracking control for time-varying delay nonlinear systems with unknown actuator dead zone, Machines (2022)

    2022

  22. [22]

    Jang, Fuzzy logic deadzone compensation with feedback linearization of nonlinear systems, Appl

    J. Jang, Fuzzy logic deadzone compensation with feedback linearization of nonlinear systems, Appl. Math. (2019)

    2019

  23. [23]

    Chiang, Adaptive fuzzy tracking control for uncertain nonlinear time-delay systems with unknown dead-zone input (2013)

    C.-C. Chiang, Adaptive fuzzy tracking control for uncertain nonlinear time-delay systems with unknown dead-zone input (2013)

    2013

  24. [24]

    H. H. Tang, N. S. Ahmad, Fuzzy logic approach for controlling uncertain and nonlinear systems: a comprehensive review of applications and advances, Syst. Sci. & Control Eng. (2024)

    2024

  25. [25]

    Cheong, S

    J. Cheong, S. Han, J. Lee, Adaptive fuzzy dynamic surface sliding mode position control for a robot manipulator with friction and deadzone (2013)

    2013

  26. [26]

    Xiang, N

    W. Xiang, N. Li, Y. Sun, Fuzzy adaptive prescribed performance control for a class of uncertain nonlinear systems with unknown dead-zone inputs (2017)

    2017

  27. [27]

    L. Wang, D. Zhao, F. Liu, Q. Liu, Z. Zhang, Active disturbance rejection position synchronous control of dual-hydraulic actuators with unknown dead-zones, in: Italian National Conf. on Sensors, 2020

    2020

  28. [28]

    B. Ahi, M. Haeri, Linear active disturbance rejection control from the practical aspects, IEEE/ASME Trans. Mechatronics (2018)

    2018

  29. [29]

    M. M. Ebrahimi, M. Homaeinezhad, Numerical algorithm for nonlinearity compensation of hardly constrained actuation for trajectory tracking control of deadzone-included dynamic systems, ISA Trans. (2024)

    2024

  30. [30]

    J. C. Willems, P. Rapisarda, I. Markovsky, B. L. M. De Moor, A note on persistency of excitation, Syst. Control Lett. 54 (4) (2005) 325–329

    2005

  31. [31]

    Coulson, J

    J. Coulson, J. Lygeros, F. Dörfler, Data-enabled predictive control: In the shallows of the deepc, in: Proc. 18th Eur. Control Conf. (ECC), Naples, Italy, 2019, pp. 307–312

    2019

  32. [32]

    Favoreel, B

    W. Favoreel, B. De Moor, P. Van Overschee, Spc: Subspace predictive control, in: Proc. 14th IFAC World Congr., Beijing, China, 1999

    1999

  33. [33]

    De Persis, P

    C. De Persis, P. Tesi, Formulas for data-driven control: Stabilization, optimality, and robustness, IEEE Trans. Autom. Control 65 (3) (2020) 909–924

    2020

  34. [34]

    Lazar, P

    M. Lazar, P. C. N. Verheijen, Offset-free data-driven predictive control, in: 2022 IEEE 61st conference on decision and control (CDC), 2022, pp. 1099–1104. M. Mazare and H. Ramezani: Preprint submitted to Elsevier Page 16 of 16 Data-Driven Dead-Zone Compensation 𝑢𝑠𝑠 𝑦𝑠𝑠 (a) Absolute Mapping Nominal ( 𝑑 = 0) 𝑢𝑐𝑚𝑑 ̂ 𝑦𝑠𝑠 𝑦∗ 𝑠𝑠 𝑒𝑠𝑠 = Φ𝑑 Mismatch 𝑑 Δ𝑢 Δ𝑦 (b) In...

    2022