pith. sign in

arxiv: 2603.03965 · v2 · submitted 2026-03-04 · 📡 eess.SY · cs.SY

Adaptive Modular Geometric Control of Robotic Manipulators

Mahdi Hejrati , Amir Hossein Barjini , Gokhan Alcan , Jouni Mattila This is my paper

Pith reviewed 2026-05-15 16:54 UTC · model grok-4.3

classification 📡 eess.SY cs.SY
keywords adaptive controlgeometric controlrobotic manipulatorsmodular controlparametric uncertaintyexponential stabilityredundant robots
0
0 comments X

The pith

Decomposing manipulator dynamics into modules enables local geometric controllers with a single adaptation gain that cut RMS position error by at least 12.2 percent.

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

This paper introduces an adaptive modular geometric control approach for robotic manipulators by breaking the overall system dynamics into separate modules. Local geometric controllers are then designed for each module, with a single adaptation gain handling uncertainties throughout the system to keep estimates consistent and stable. Exponential stability is shown when the model is perfect, and simulations on a redundant arm confirm at least 12.2 percent lower RMS position errors than competing methods while using similar control effort. The framework addresses parametric uncertainties without drift in estimates.

Core claim

The adaptive modular geometric control framework decomposes the manipulator dynamics into individual modules to design local geometric control laws at the module level. A geometric adaptation law with only one adaptation gain is incorporated to handle parametric uncertainties, ensuring physically consistent and drift-free estimates. Exponential stability is established for the nominal case, and simulations demonstrate a reduction in RMS position error of at least 12.2% compared to state-of-the-art controllers under nearly identical control effort.

What carries the argument

The adaptive modular geometric control framework, which decomposes overall dynamics into modules for independent local geometric control laws augmented by a single-gain adaptation mechanism.

If this is right

  • Exponential stability holds in the nominal case without adaptation.
  • RMS position error reduces by at least 12.2 percent with nearly identical control effort.
  • Parametric uncertainties are compensated while preserving high tracking performance.
  • A single adaptation gain suffices for the entire system with physically consistent estimates.

Where Pith is reading between the lines

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

  • The modular decomposition could simplify controller design for higher-degree-of-freedom robots by adding modules incrementally.
  • Similar local geometric laws with minimal adaptation might extend to other coupled mechanical systems such as vehicle suspensions.
  • The single-gain structure reduces tuning effort, which could aid real-time deployment on embedded hardware.

Load-bearing premise

Manipulator dynamics can be decomposed into independent modules where local geometric control laws can be designed separately while still guaranteeing global stability and performance.

What would settle it

A hardware test on a redundant manipulator showing no RMS position error reduction near 12 percent or loss of stability under parameter variations would disprove the performance and stability claims.

Figures

Figures reproduced from arXiv: 2603.03965 by Amir Hossein Barjini, Gokhan Alcan, Jouni Mattila, Mahdi Hejrati.

Figure 1
Figure 1. Figure 1: Interconnected multi-rigid body system decomposed into modules of rigid body and joints. of adjacent bodies. By taking the time-derivative of (17), one can establish the analytic representation of body-fixed acceleration, A𝑖 = 𝐴𝑑 −1 𝑖−1,𝑖 A𝑖−1 − [𝜉𝑖 𝜃̇ 𝑖 , 𝐴𝑑 −1 𝑖−1,𝑖 V𝑖−1] + 𝜉𝑖 𝜃̈ 𝑖 . (18) The desired terms V 𝑑 𝑖 and A 𝑑 𝑖 are computed by replacing 𝜃𝑖 and 𝜃̇ 𝑖 with 𝜃 𝑑 𝑖 and 𝜃̇ 𝑑 𝑖 in (17) and (18), res… view at source ↗
Figure 2
Figure 2. Figure 2: Schematic illustration of the SPD manifold P(4). The true inertia 𝐿𝑖 , nominal model 𝐿0 𝑖 , and estimate 𝐿̂ 𝑖 are depicted by local ellipsoidal proxies associated with symmetric positive-definite matrices, while the curves represent Rieman￾nian geodesics between them. Theorem 3. Consider a complex robotic system decom￾posed into the subsystems, as shown in [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: 4R redundant heavy-duty manipulator scheme em￾ployed in simulations. The total mass is ≈ 9350 kg, reach ≈ 7 m tuned to achieve their best attainable performance under comparable control effort. For the comparison of the non￾adaptive controllers, the initial joint configuration is chosen as 𝜽(0) = [−10, 70, −90, −70]⊤deg, 𝜽̇ (0) = 𝟎, so as to induce a sufficiently large initial configuration error. This cho… view at source ↗
Figure 4
Figure 4. Figure 4: End-effector tracking performance under the con￾sidered controllers. (a)–(c) report the Cartesian position re￾sponses in the 𝑥-, 𝑦-, and 𝑧-directions, respectively, while (d) shows the corresponding end-effector orientation regulation. the corresponding end-effector orientation regulation. It is observed that both geometric controllers, namely MGC and GIC, outperform the Euclidean-space-based modular con￾t… view at source ↗
Figure 7
Figure 7. Figure 7: Time histories of the end-effector tracking errors. (a)–(c) correspond to the Cartesian position errors in the 𝑥-, 𝑦-, and 𝑧-directions, respectively, and (d) shows the orientation error [PITH_FULL_IMAGE:figures/full_fig_p012_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Joint control torques produced by the considered controllers. (a)–(d) correspond to joints 1–4, respectively. The results indicate that the proposed MGC yields smoother torque profiles while requiring control effort comparable to, or lower than, the benchmark methods. [12] M. C. Nah, J. Lachner, N. Hogan, Modular robot control with motor primitives, arXiv preprint arXiv:2505.10694 (2025). [13] W.-H. Zhu, V… view at source ↗
Figure 6
Figure 6. Figure 6: Snapshots of the manipulator motion under the considered non-adaptive controllers. [9] J. Seo, S. Yoo, J. Chang, H. An, H. Ryu, S. Lee, A. Kruthiventy, J. Choi, R. Horowitz, Se (3)-equivariant robot learning and control: a tutorial survey, International Journal of Control, Automation and Systems 23 (5) (2025) 1271–1306. [10] F. J. Abu-Dakka, V. Kyrki, Geometry-aware dynamic movement primitives, in: 2020 IE… view at source ↗
Figure 9
Figure 9. Figure 9: Comparison of the tracking performance of the MGC and AMGC schemes under −10% parametric uncertainty [PITH_FULL_IMAGE:figures/full_fig_p013_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Comparison of the tracking performance of the MGC and AMGC schemes under +10% parametric uncertainty. (a) (b) [PITH_FULL_IMAGE:figures/full_fig_p013_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Time histories of the estimated parameters under a) +10% parametric uncertainty and b) −10% parametric un￾certainty. Each curve represents the norm of the corresponding parameter estimate. [20] A. H. Barjini, M. Bahari, M. Hejrati, J. Mattila, Surrogate-enhanced modeling and adaptive modular control of all-electric heavy-duty robotic manipulators, arXiv preprint arXiv:2508.06313 (2025). [21] L. Ding, H. X… view at source ↗
read the original abstract

This paper proposes an adaptive modular geometric control framework for robotic manipulators. The proposed methodology decomposes the overall manipulator dynamics into individual modules, enabling the design of local geometric control laws at the module level. To address parametric uncertainties, geometric adaptation law is incorporated into the control structure, requiring only a single adaptation gain for the entire system while ensuring physically consistent and drift-free parameter estimates. Exponential stability of the proposed controller is established in the nominal case. Numerical simulations on a complex redundant robotic manipulator are conducted to evaluate the proposed approach against existing modular and geometric control methods. The results show that the proposed method reduces the RMS position error by at least 12.2% compared with state-of-the-art controllers under almost the same control effort. In addition, the adaptive extension demonstrates strong capability in compensating for parametric uncertainties and preserving high tracking performance.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper proposes an adaptive modular geometric control framework for robotic manipulators that decomposes the system dynamics into individual modules to enable independent local geometric control laws. A geometric adaptation law with a single adaptation gain is added to handle parametric uncertainties while ensuring physically consistent, drift-free estimates. Exponential stability is claimed for the nominal case, and simulations on a redundant manipulator show at least 12.2% reduction in RMS position error versus state-of-the-art modular and geometric controllers under comparable effort.

Significance. If the global stability guarantee holds under the modular decomposition, the framework would offer a practical advance for controlling complex redundant manipulators by permitting modular design with minimal tuning (one gain) and uncertainty compensation. The reported simulation improvement, if robustly supported, indicates measurable tracking gains without increased effort.

major comments (2)
  1. [Stability analysis] Stability analysis: The central claim of global exponential stability for the composed closed-loop system requires an explicit derivation showing that the sum of local Lyapunov functions remains negative definite after including inter-module coupling torques from the configuration-dependent inertia and Coriolis terms. The abstract states that local laws preserve global properties but provides no indication that residual cross terms were shown to be dominated or canceled.
  2. [Numerical simulations] Simulation results: The 12.2% RMS position error reduction is load-bearing for the performance claim, yet the manuscript must specify the exact baseline controllers (including their parameters), the redundant manipulator model, initial conditions, and whether the improvement is accompanied by error bars or statistical tests over repeated trials to confirm it is not an artifact of a single run.
minor comments (2)
  1. [Abstract] Abstract: The description of the geometric adaptation law should briefly indicate how the single gain enforces physical consistency and drift-free behavior, as this is central to the adaptive extension.
  2. [Methodology] Notation: Module indices and coupling terms should be defined consistently when transitioning from individual module dynamics to the global vector field.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments, which help clarify the presentation of our stability analysis and simulation results. We address each major comment below.

read point-by-point responses

  1. Referee: [Stability analysis] Stability analysis: The central claim of global exponential stability for the composed closed-loop system requires an explicit derivation showing that the sum of local Lyapunov functions remains negative definite after including inter-module coupling torques from the configuration-dependent inertia and Coriolis terms. The abstract states that local laws preserve global properties but provides no indication that residual cross terms were shown to be dominated or canceled.

    Authors: We acknowledge that the current manuscript states the global exponential stability result but does not provide a fully expanded derivation of how the sum of local Lyapunov functions accounts for the configuration-dependent coupling torques. In the revised manuscript we will add an explicit step-by-step derivation (new subsection in Section III) that shows the geometric control laws cancel the skew-symmetric Coriolis contributions and that any residual inertia-coupling terms are dominated by the negative-definite quadratic forms arising from the local gains, thereby preserving the negative-definiteness of the composite Lyapunov derivative. revision: yes

  2. Referee: [Numerical simulations] Simulation results: The 12.2% RMS position error reduction is load-bearing for the performance claim, yet the manuscript must specify the exact baseline controllers (including their parameters), the redundant manipulator model, initial conditions, and whether the improvement is accompanied by error bars or statistical tests over repeated trials to confirm it is not an artifact of a single run.

    Authors: We agree that the simulation section requires additional detail to make the performance claim reproducible. In the revision we will: (i) name the exact baseline controllers (the modular geometric controller of [citation] and the geometric PD controller of [citation]) together with their numerical gain values; (ii) provide the complete 7-DOF redundant manipulator dynamic parameters (link lengths, masses, and inertia tensors); (iii) list the initial joint positions and velocities used; and (iv) report mean RMS errors and standard deviations over 20 independent trials together with a paired t-test confirming statistical significance of the observed improvement. revision: yes

Circularity Check

0 steps flagged

No significant circularity; modular decomposition and stability claims rest on standard geometric control analysis rather than self-referential reduction.

full rationale

The derivation decomposes manipulator dynamics into modules to enable independent local geometric controllers, incorporates a single-gain adaptation law for parametric uncertainties, and establishes nominal exponential stability (presumably via Lyapunov analysis on the closed-loop system). Simulation results report a 12.2% RMS error reduction against baselines under comparable effort. No quoted equations or steps in the provided abstract reduce a claimed prediction or stability guarantee to a fitted parameter chosen from the evaluation data, nor does any load-bearing premise collapse to a self-citation whose content is unverified within the paper. The central claims remain independently derivable from standard geometric and adaptive control principles applied to the stated decomposition, with empirical validation supplied separately by the numerical experiments.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The framework rests on the domain assumption that manipulator dynamics admit a clean modular decomposition and on standard geometric control axioms; the single adaptation gain is the primary free parameter whose value is not derived from first principles.

free parameters (1)
  • single adaptation gain
    One scalar gain used for the entire system; its specific value is chosen to achieve the reported performance but is not derived analytically from the dynamics.
axioms (1)
  • domain assumption Manipulator dynamics can be decomposed into independent modules for local geometric control design
    Invoked to enable module-level controllers while preserving global properties.

pith-pipeline@v0.9.0 · 5446 in / 1214 out tokens · 36226 ms · 2026-05-15T16:54:38.024348+00:00 · methodology

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

29 extracted references · 29 canonical work pages

  1. [1]

    Bullo, R

    F. Bullo, R. M. Murray, Proportional derivative (pd) control on the euclidean group, Cdc technical report 95-010, California Institute of Technology, available electronically viahttps://www.cds.caltech. edu/~murray/preprints/cds95-010.pdf(1995)

    work page 1995

  2. [2]

    T. Lee, M. Leok, N. H. McClamroch, Geometric tracking control of a quadrotor uav on se (3), in: 49th IEEE conference on decision and control (CDC), IEEE, 2010, pp. 5420–5425

    work page 2010

  3. [3]

    J. Seo, N. P. S. Prakash, A. Rose, J. Choi, R. Horowitz, Geomet- ric impedance control on se (3) for robotic manipulators, IFAC- PapersOnLine 56 (2) (2023) 276–283

    work page 2023

  4. [4]

    J. Seo, N. P. Prakash, X. Zhang, C. Wang, J. Choi, M. Tomizuka, R. Horowitz, Contact-rich se (3)-equivariant robot manipulation task learning via geometric impedance control, IEEE Robotics and Au- tomation Letters 9 (2) (2023) 1508–1515

    work page 2023

  5. [5]

    T.Lee,F.C.Park,Ageometricalgorithmforrobustmultibodyinertial parameteridentification,IEEERoboticsandAutomationLetters3(3) (2018) 2455–2462

    work page 2018

  6. [6]

    T. Lee, P. M. Wensing, F. C. Park, Geometric robot dynamic iden- tification: A convex programming approach, IEEE Transactions on Robotics 36 (2) (2019) 348–365

    work page 2019

  7. [7]

    T. Lee, J. Kwon, F. C. Park, A natural adaptive control law for robot manipulators, in: 2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), IEEE, 2018, pp. 1–9

    work page 2018

  8. [8]

    Alcan, F

    G. Alcan, F. J. Abu-Dakka, V. Kyrki, Constrained trajectory opti- mization on matrix lie groups via lie-algebraic differential dynamic programming, Systems & Control Letters 204 (2025) 106220. First Author et al.:Preprint submitted to ElsevierPage 11 of 13 Short Title of the Article (a) VDC (b) GIC (c) MGC Figure 6:Snapshots of the manipulator motion unde...

    work page 2025

  9. [9]

    J. Seo, S. Yoo, J. Chang, H. An, H. Ryu, S. Lee, A. Kruthiventy, J. Choi, R. Horowitz, Se (3)-equivariant robot learning and control: a tutorial survey, International Journal of Control, Automation and Systems 23 (5) (2025) 1271–1306

    work page 2025

  10. [10]

    F. J. Abu-Dakka, V. Kyrki, Geometry-aware dynamic movement primitives, in: 2020 IEEE International Conference on Robotics and Automation (ICRA), IEEE, 2020, pp. 4421–4426

    work page 2020

  11. [11]

    Saveriano, F

    M. Saveriano, F. J. Abu-Dakka, V. Kyrki, Learning stable robotic skills on riemannian manifolds, Robotics and Autonomous Systems 169 (2023) 104510. Figure 7:Time histories of the end-effector tracking errors. (a)–(c) correspond to the Cartesian position errors in the𝑥-, 𝑦-, and𝑧-directions, respectively, and (d) shows the orientation error. Figure 8:Joint...

    work page 2023

  12. [12]

    M. C. Nah, J. Lachner, N. Hogan, Modular robot control with motor primitives, arXiv preprint arXiv:2505.10694 (2025)

    work page arXiv 2025

  13. [13]

    Zhu, Virtual decomposition control: toward hyper degrees of freedom robots, Vol

    W.-H. Zhu, Virtual decomposition control: toward hyper degrees of freedom robots, Vol. 60, Springer Science & Business Media, 2010

    work page 2010

  14. [14]

    Hejrati, J

    M. Hejrati, J. Mattila, Desired impedance allocation for robotic ma- nipulators, in: 2025 IEEE 64th Conference on Decision and Control (CDC), IEEE, 2025, pp. 2634–2641

    work page 2025

  15. [15]

    W.-H. Zhu, G. Vukovich, Virtual decomposition control for modular robotmanipulators,IFACProceedingsVolumes44(1)(2011)13486– 13491

    work page 2011

  16. [16]

    Humaloja, J

    J.-P. Humaloja, J. Koivumäki, L. Paunonen, J. Mattila, Decentralized observerdesignforvirtualdecompositioncontrol,IEEETransactions on Automatic Control 67 (5) (2021) 2529–2536

    work page 2021

  17. [17]

    Koivumäki, J.-P

    J. Koivumäki, J.-P. Humaloja, L. Paunonen, W.-H. Zhu, J. Mattila, Subsystem-based control with modularity for strict-feedback form nonlinear systems, IEEE Transactions on Automatic Control 68 (7) (2022) 4336–4343

    work page 2022

  18. [18]

    Hejrati, J

    M. Hejrati, J. Mattila, Orchestrated robust controller for precision control of heavy-duty hydraulic manipulators, IEEE Transactions on Automation Science and Engineering (2025)

    work page 2025

  19. [19]

    Hejrati, J

    M. Hejrati, J. Mattila, Impact-resilient orchestrated robust controller forheavy-dutyhydraulicmanipulators,IEEE/ASMETransactionson Mechatronics (2025). First Author et al.:Preprint submitted to ElsevierPage 12 of 13 Short Title of the Article Figure 9:Comparison of the tracking performance of the MGC and AMGC schemes under−10%parametric uncertainty. Figur...

    work page 2025

  20. [20]

    A. H. Barjini, M. Bahari, M. Hejrati, J. Mattila, Surrogate-enhanced modeling and adaptive modular control of all-electric heavy-duty robotic manipulators, arXiv preprint arXiv:2508.06313 (2025)

    work page arXiv 2025

  21. [21]

    L. Ding, H. Xing, H. Gao, A. Torabi, W. Li, M. Tavakoli, Vdc- basedadmittancecontrolofmulti-dofmanipulatorsconsideringjoint flexibility via hierarchical control framework, Control Engineering Practice 124 (2022) 105186

    work page 2022

  22. [22]

    Hejrati, P

    M. Hejrati, P. Mustalahti, J. Mattila, Robust immersive bilateral teleoperation of beyond-human-scale systems with enhanced trans- parency and sense of embodiment, arXiv preprint arXiv:2505.14486 (2025)

    work page arXiv 2025

  23. [23]

    O. Khatib, A unified approach for motion and force control of robot manipulators: The operational space formulation, IEEE Journal on RoboticsandAutomation3(1)(1987)43–53.doi:10.1109/JRA.1987. 1087068

    work page doi:10.1109/jra.1987 1987

  24. [24]

    Morton, M

    D. Morton, M. Pavone, Safe, task-consistent manipulation with op- erational space control barrier functions, in: 2025 IEEE/RSJ Interna- tional Conference on Intelligent Robots and Systems (IROS), IEEE, 2025, pp. 187–194

    work page 2025

  25. [25]

    1867– 1874

    S.Teng,W.Clark,A.Bloch,R.Vasudevan,M.Ghaffari,Liealgebraic cost function design for control on lie groups, in: 2022 IEEE 61st Conference on Decision and Control (CDC), IEEE, 2022, pp. 1867– 1874

    work page 2022

  26. [26]

    R. M. Murray, Z. Li, S. S. Sastry, A mathematical introduction to robotic manipulation, CRC press, 2017

    work page 2017

  27. [27]

    Featherstone, Rigid body dynamics algorithms, Springer, 2008

    R. Featherstone, Rigid body dynamics algorithms, Springer, 2008

    work page 2008

  28. [28]

    F.C.Park,J.E.Bobrow,S.R.Ploen,Aliegroupformulationofrobot dynamics,TheInternationaljournalofroboticsresearch14(6)(1995) 609–618

    work page 1995

  29. [29]

    LEE, Geometric methods for dynamic model-based identifica- tion and control of multibody systems, Ph.D

    T. LEE, Geometric methods for dynamic model-based identifica- tion and control of multibody systems, Ph.D. thesis, Department of mechanical and Aerospace Engineering, Seoul National University (2019). First Author et al.:Preprint submitted to ElsevierPage 13 of 13

    work page 2019