REVIEW 4 major objections 4 minor 18 references
Optimal Dynamic Control of Bounded Jacobian Discrete-Time Systems via Interval Observers
T0 review · 4 major / 4 minor · reviewed 2026-08-16 · deepseek-v4-flash
Pith's one-line read Nonlinear dynamic feedback control stabilizes uncertain discrete-time systems by stabilizing an interval observer around the state, with a separation principle for gain design and LMI conditions giving ISS and an H-infinity bound.
desk verdict A sensible dynamic extension of the author's static interval-observer controller, but Lemma 1's comparison step needs a cooperativity condition that is never imposed, so the main stability and H∞ guarantees are not established. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The central object is the interval framer (13), a pair of coupled recursions for upper and lower state bounds built from tight mixed-monotone decomposition functions $\varphi_d$ and $\psi_d$; the same decompositions give matrices $F_\varphi$ and $F_\psi$ that bound the nonlinear increments via inequality (5). The framer error system (14) is a positive system once $L$ stabilizes $|A-LC|$. The argument then moves to the comparison augmented system (22), whose matrix $\tilde{A}$ is block lower triangular; this block structure is what lets observer and controller gains be designed separately. The SDP (23) is obtained by applying congruence transformations and a change of variables to known Lipschitz-nonlinear ISS conditions, eliminating bilinear products of the decision variables with $B$.
What would settle it
Fix a system satisfying Assumption 1, choose $L$ from the SDP in Proposition 3, solve SDP (23), and simulate the full closed loop (8), (13), (16) from the vertices of the initial interval with noise sequences $w_k,v_k$ held at their extreme values. If the true state ever leaves $[\underline{x}_k,\overline{x}_k]$, if the framer error $\varepsilon_k$ becomes negative, or if $\lVert z_k\rVert_2^2 > \mu^* \lVert \tilde{w}_k\rVert_2^2$ for any $k$, the claimed ISS property and bound (24) are false.
Extended reading notes
Core claim
On the paper's own terms, the discovery is that the interval-framer idea can be turned into a dynamic control design for bounded-Jacobian discrete-time systems without measuring the state. Writing $f(x)=Ax+\varphi(x)$ and $g(x)=Cx+\psi(x)$ with $\varphi,\psi$ Jacobian-sign-stable, the framer equations (13) produce bounds $\underline{x}_k \le x_k \le \overline{x}_k$, and the framer error evolves as $\varepsilon_{k+1}=|A-LC|\varepsilon_k + \delta^\varphi_k + |L|\delta^\psi_k + |LV|\delta_v + |W|\delta_w$. Lemma 1 claims a separation principle: the augmented system of framers, tracking errors, and controller states has a comparison system (22) whose state matrix is block lower triangular, so a stabilizing observer gain $L$ for the framer error system plus subsequently chosen dynamic gains stabilize the whole closed loop. Theorem 1 converts this into an SDP (23), with gains recovered from (25), and states that the closed-loop system is ISS and obeys $\lVert z_k\rVert_2^2 \le \mu^* \lVert \tilde{w}_k\rVert_2^2$. The paper notes in Remark 2 that the separation is not 'full,' since $L$ remains an input to the controller synthesis.
Load-bearing premise
The load-bearing premise is that the component-wise bounding comparison system (22) can stand in for the real closed-loop dynamics—stability of that bounding system is taken to imply stability of the true interval and error systems—but the transfer is assumed rather than proved, and the comparison matrix $\tilde{A}$ contains sign-indefinite entries.
Editorial extensions
If this is right
- The true state remains inside the computed interval at every time step, so the same controller can certify state bounds while regulating the system.
- Observer and controller gains are synthesized in two separate convex steps, which keeps the design tractable as the state dimension grows.
- The synthesis produces a numerical $H_\infty$ attenuation level $\mu^*$, allowing a designer to trade interval tightness against worst-case disturbance amplification.
- Nonlinear measurement functions and both state and measurement noise are handled, going beyond earlier interval-observer control formulations restricted to linear, noise-free outputs.
- The example indicates that the dynamic controller's extra gains yield substantially tighter closed-loop intervals than the static feedback design, not just stability.
Reading between the lines
- Editorial extension: the comparison step in Lemma 1 would be fully justified if $\tilde{A}$'s off-diagonal entries and $\lambda(z)$ were component-wise nonnegative; checking this property on a given example is a direct test, and when it holds, simpler linear comparison tools could replace the Lipschitz-based SDP.
- Editorial extension: since Lemma 1's separation is one-way, jointly optimizing $L$ and the controller gains is a natural next step that the paper itself flags in Remark 2.
- Editorial extension: the same set-valued state estimate could drive event-triggered or model-predictive controllers; the conclusion names switched and hybrid systems and MPC as future work.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The manuscript considers discrete-time nonlinear systems with bounded Jacobians, nonlinear measurements, and bounded state/measurement noise. Rather than stabilizing the uncertain plant directly, it constructs an interval framer/observer (13) whose upper and lower trajectories bound the true state, and then designs a nonlinear dynamic output-feedback controller (16) for the framer system. The main claims are a separation principle (Lemma 1), an LMI-based synthesis procedure (Theorem 1) that guarantees input-to-state stability of the closed-loop comparison system and the H-infinity bound (24), and a numerical example showing tighter intervals than a static controller from the author's prior work.
Significance. If the main results were fully established, the paper would be a useful extension of the author's earlier static interval-observer control method to dynamic nonlinear control, with the advertised separation principle enabling separate design of observer and controller gains. The manuscript has clear strengths: it uses mixed-monotone decomposition tools in a nontrivial way, the problem formulation is explicit about noise and nonlinear observations, and the numerical example reports concrete matrices and achieved attenuation levels. However, the central proof of the separation principle is incomplete in a load-bearing way, and the current draft does not establish the ISS claims that the LMI synthesis is meant to guarantee. The contribution is therefore conditional on a repairable but nontrivial revision of Lemma 1 and the associated comparison argument.
major comments (4)
- [Section IV-B, Lemma 1, Eq. (22)] The proof of Lemma 1 asserts that the matrix \tilde A in (22) is block lower triangular and uses this to separate the design of L from the design of the controller gains. This assertion is incorrect: the displayed blocks include \tilde A_{34} = K_b + K_x^\nu F_\varphi and \tilde A_{35} = -(K_b + K_x^\nu F_\varphi) in row 3, while rows 4 and 5 contain \tilde A_{43} = B C_c, \tilde A_{53} = B C_c, as well as nonzero couplings to e and \bar{e} such as \tilde A_{41}, \tilde A_{42}, \tilde A_{51}, \tilde A_{52}. These entries create a dependency cycle x_c -> \bar{x},\underline{x} -> x_c, so no reordering of the state vector makes \tilde A block triangular. The spectral argument in the proof therefore does not support separate design of the observer gain and the dynamic controller gains as stated.
- [Section IV-B, Lemma 1, Eq. (21)-(22)] The proof transfers ISS of the comparison system (22) to ISS of the original closed-loop framer and error systems. Such a transfer requires the comparison dynamics to be cooperative, i.e. the right-hand side must be monotone nondecreasing in the comparison state z_k. No monotonicity property is proved, and it is in fact not satisfied by the displayed entries: \tilde A_{21} = (A-LC)^\ominus + F_\varphi - |L|F_\psi subtracts the nonnegative matrix |L|F_\psi, and \tilde A_{45} = -(A-LC)^\ominus - B K_d is sign-indefinite in general. With negative comparison coefficients, boundedness (or ISS) of the comparison system does not imply boundedness (or ISS) of the actual framer/error system. The comparison principle used in the proof of Lemma 1 is therefore not valid without additional explicit cooperativity or monotonicity conditions.
- [Section III, paragraph after Eq. (9)] The claim that F_\varphi can 'without loss of generality' be assumed invertible by increasing its diagonal elements is not harmless. The Lipschitz constant used in Theorem 1 is \gamma = \|F_\varphi\|_\infty, and the bound (5) is used with this F_\varphi in the comparison inequalities (21). Increasing diagonal entries changes \gamma and changes the comparison matrix \tilde A through the terms involving F_\varphi. The manuscript does not show that the modified F_\varphi preserves the required decomposition bound with the same \gamma, nor that the resulting comparison system remains a valid upper bound for the original framer/error dynamics. This point needs a proof or an explicit restatement of the assumptions under which the modification is valid.
- [Section IV-C, Theorem 1, proof of (26)-(27)] The proof of Theorem 1 is abbreviated: it invokes [18, Lemma 3] and then applies similarity transformations and changes of variables without showing that the transformed LMIs are equivalent to the original ones, especially with respect to the block-diagonal congruence transformations and the definition of \Gamma. In particular, the relation diag_5(\tilde K^*) = (Q^{-1} \Theta^*)^\top is stated without derivation, and the dimensions of the matrices in (23) and (26)-(27) are not checked against the stated 20n/\hat{n} block sizes. Since the ISS guarantee depends entirely on this LMI step, the proof needs to be written out in full or the missing equivalence argument must be provided.
minor comments (4)
- [Section V, Example data] In the matrix A displayed in Section V, the entry shown as '01384' appears to be a typo and should likely read '0.1384'.
- [Equation (24)] The signal \tilde{w}_k in the H-infinity bound (24) is never defined; the reader must infer that it denotes the augmented noise vector \eta_k from (22). Please define it explicitly.
- [Lemma 1 proof] The proof states that 'by applying the last inequality in (21), it is straightforward to see that (20) admits a linear comparison system with state matrix \tilde A_u'. The derivation is not shown, and given the coupling terms introduced in (18), this step is not immediate; it should be expanded or the relevant inequalities should be written out.
- [References] Reference [2] contains a typo ('Teansaction'), and several key results are cited as 'follows the lines of [8]' or 'accepted' ([15]); please update the reference list with complete publication data and identify precisely which parts of the proofs rely on these prior results.
Circularity Check
No significant circularity: the dynamic control design is self-contained apart from legitimate prior-work citations.
full rationale
The paper's claimed contribution is dynamic output-feedback stabilization of a bounded-Jacobian discrete-time system via an interval observer. The derivation chain uses prior decomposition results [16], an observer-gain SDP from [8], a comparison-system separation argument in Lemma 1, and an LMI stability lemma from [18]. None of these steps defines its target claim into its inputs: no parameter is fitted to data and then renamed as a prediction, and the H-infinity bound (24) is the SDP objective rather than a fitted value. The self-citations to [8], [15], and [16] are load-bearing in the sense of being used as tools, but they are prior published results whose stated assumptions do not include the present dynamic-control result, so they are independent evidence under the review rules. The weak point noted by the reader—that Lemma 1's comparison system may require a cooperativity property and that the displayed matrix A~ is not block lower triangular—is a correctness or rigor concern, not a circularity: the comparison system is not claimed to be identical to the closed-loop system by construction, and no equation in the paper reduces the ISS claim to the LMI feasibility problem. Therefore no circular step can be exhibited, and the honest finding is no significant circularity.
Assumptions & free parameters
free parameters (2)
- alpha =
0.1 (example, user-chosen)
- epsilon_0 =
0.001 (example)
assumptions (6)
- domain assumption Assumption 1: known initial state bounds x0,xbar0; known Jacobian bounds for f and g; output y_k known at all times.
- domain assumption Mixed-monotone decomposition results of [16] (Propositions 1 and 2) hold, giving additive decompositions f=Ax+phi, g=Cx+psi with JSS phi,psi and F_phi,F_psi satisfying (5).
- ad hoc to paper F_phi can be assumed invertible without loss of generality by increasing its diagonal elements, preserving (5).
- domain assumption The observer gain L computed from the SDP in [8] (as recalled in Proposition 3) stabilizes the framer error system (14).
- domain assumption ISS of the comparison system (22) implies ISS and closed-loop stability of the original augmented error and framer systems, requiring a valid monotone comparison principle.
- domain assumption [18, Lemma 3] provides sufficient LMI conditions for ISS of Lipschitz nonlinear discrete-time systems of the form (22).
Cite this review
Pith. "Pith review of Optimal Dynamic Control of Bounded Jacobian Discrete-Time Systems via Interval Observers." pith.science (2026). https://pith.science/paper/ALRC67QP
@misc{pith2026250421832,
author = {Pith},
title = {Pith review of: Optimal Dynamic Control of Bounded Jacobian Discrete-Time Systems via Interval Observers},
year = {2026},
howpublished = {\url{https://pith.science/paper/ALRC67QP}},
note = {Machine review of arXiv:2504.21832}
}
read the original abstract
This paper presents an optimal dynamic control framework for bounded Jacobian nonlinear discrete-time (DT) systems with nonlinear observations affected by both state and process noise. Rather than directly stabilizing the uncertain system, we focus on stabilizing an interval observer in a higher dimensional space, whose states bound the true system states. Our nonlinear dynamic control method introduces added flexibility over traditional static and linear approaches, effectively compensating for system nonlinearities and enabling potentially tighter closed-loop intervals. Additionally, we establish a separation principle that allows for the design of observer and control gains. We further derive tractable matrix inequalities to ensure system stability in the closed-loop configuration. The simulation results show that the proposed dynamic control approach significantly outperforms a static counterpart method.
Figures
Reference graph
Works this paper leans on
- [1]
-
[2]
Distributed Resilient Interval Observer Synthesis for Nonlinear Discrete-Time Systems
M. Khajenejad, S. Brown, and S. Martinez. Distributed re - silient interval observer synthesis for nonlinear discret e-time sys- tems. IEEE Teansaction on Automatic Control, to appear , 2025. https://arxiv.org/pdf/2401.15511
work page Pith review arXiv 2025
-
[3]
A.M. Tahir and B. Ac ¸ıkmes ¸e. Synthesis of interval obse rvers for bounded Jacobian nonlinear discrete-time systems. IEEE Control Systems Letters , 2021
work page 2021
-
[4]
M. Khajenejad, Z. Jin, and S.Z. Y ong. Interval observers for simultane- ous state and model estimation of partially known nonlinear systems. In American Control Conference (ACC) , pages 2848–2854, 2021
work page 2021
-
[5]
L. Farina and S. Rinaldi. Positive linear systems: theory and applica- tions, volume 50. John Wiley & Sons, 2000
work page 2000
-
[6]
M. Kieffer and E. Walter. Guaranteed nonlinear state est imation for continuous-time dynamical models from discrete-time meas urements. IF AC Proceedings V olumes, 39(9):685–690, 2006
work page 2006
- [7]
-
[8]
M. Khajenejad and S.Z. Y ong. H∞-optimal interval observer syn- thesis for uncertain nonlinear dynamical systems via mixed -monotone decompositions. IEEE Control Systems Letters , 6:3008–3013, 2022
work page 2022
Show all 18 references
-
[9]
T. Pati, M. Khajenejad, S. Praveen Daddala, and S.Z. Y ong . l1-robust interval observer design for uncertain nonlinear dynamica l systems. IEEE Control Systems Letters , 6:3475–3480, 2022
2022
-
[10]
Y .E. Wang, Y . Gao, D. Wu, and B. Niu. Interval observer-b ased event- triggered control for switched linear systems. Journal of the Franklin Institute, 357(10):5753–5772, 2020
2020
-
[11]
Zhang, J
Z. Zhang, J. Shen, J. Zhang, and H. Qiu. Semi-global inte rval observer- based robust control of linear time-invariant systems subj ect to input saturation. Circuits, Systems, and Signal Processing, pages 1–24, 2024
2024
-
[12]
Efimov, T
D. Efimov, T. Raissi, and A. Zolghadri. Control of nonlin ear and LPV systems: interval observer-based framework. IEEE Transactions on Automatic Control , 58(3):773–778, 2013
2013
-
[13]
He and W
Z. He and W. Xie. Control of nonlinear systems based on in terval observer design. IET Control Theory & Applications , 12(4):543–548, 2018
2018
-
[14]
Efimov, A
D. Efimov, A. Polyakov, and X. Ping. Interval observatio n and control for continuous-time Persidskii systems. IEEE Transactions on Automatic Control , 2024
2024
-
[15]
Khajenejad
M. Khajenejad. Optimal feedback stabilizing control o f bounded jacobian discrete-time systems via interval observers. American Control Conference, accepted, https://arxiv.org/pdf/2410.09222, 2025
2025 arXiv
-
[16]
Khajenejad, F
M. Khajenejad, F. Shoaib, and S.Z. Y ong. Interval obser ver synthesis for locally Lipschitz nonlinear dynamical systems via mixe d-monotone decompositions. In American Control Conference (ACC), pages 2970– 2975, 2022
2022
-
[17]
Sootla, Y
A. Sootla, Y . Zheng, and A. Papachristodoulou. Block-d iagonal solutions to Lyapunov inequalities and generalisations of diagonal dominance. In IEEE 56th conference on decision and control (CDC) , pages 6561–6566. IEEE, 2017
2017
-
[18]
Abbaszadeh and Horacio J
M. Abbaszadeh and Horacio J. Marquez. LMI optimization approach to robust H∞ observer design and static output feedback stabilization for discrete-time nonlinear uncertain systems. International Journal of Robust and Nonlinear Control: IF AC-Affiliated Journal , 19(3):313– 340, 2009
2009
Reviewed August 16, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.