Pith. sign in

REVIEW 2 major objections 5 minor 16 references

Continuous-time primal-dual gradient dynamics for convex optimal control form a port-Hamiltonian PDE system whose equilibria are the Pontryagin conditions and that can be used for sub-optimal control.

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

2026-07-14 17:45 UTC pith:BQ2Y2CWP

load-bearing objection Clean, explicit PDE lift of port-Hamiltonian PDGCT to linear-convex optimal control; the structural claim is solid, the infinite-dimensional convergence claim is only sketched. the 2 major comments →

arxiv 2606.18818 v2 pith:BQ2Y2CWP submitted 2026-06-17 math.OC

Suboptimal control by primal-dual gradient dynamics

Arjan van der Schaft This is my paper
classification math.OC MSC 49J1593B5237J2549M29
keywords primal-dual gradient dynamicsport-Hamiltonian systemsoptimal controlPontryagin minimum principleinfinite-dimensional systemssuboptimal controlshifted Hamiltonian
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.

The paper shows that the continuous-time primal-dual gradient algorithm, already known to be an incremental port-Hamiltonian system for static constrained convex optimization, extends in a natural way to the infinite-dimensional setting of linear optimal control with strictly convex cost. Interpreting the linear dynamics as equality constraints produces a system of partial differential equations in two times: ordinary physical time and an algorithmic time. That PDE system is itself an infinite-dimensional port-Hamiltonian system; its equilibria are exactly the first-order necessary conditions of Pontryagin’s minimum principle, and a shifted Hamiltonian decreases along trajectories. Under the natural mixed boundary conditions and controllability of the plant, this decrease indicates convergence to the optimal control. The same PDE formulation is proposed as a starting point for deriving computationally lighter, sub-optimal controllers, for example by discretizing the two-time dynamics or by stopping the algorithmic evolution early.

Core claim

The continuous-time primal-dual gradient algorithm applied to a linear optimal-control problem with strictly convex running cost is an infinite-dimensional incremental port-Hamiltonian system of PDEs whose equilibria coincide with the Pontryagin first-order conditions and whose shifted Hamiltonian is non-increasing, thereby indicating asymptotic convergence to the optimal trajectory under the mixed boundary conditions of the minimum principle and controllability of (A,B).

What carries the argument

The formally skew-adjoint operator J that couples the state, control and co-state along physical time (Eq. 18), together with the shifted Hamiltonian functional built from the quadratic energy variables; their combination yields the infinite-dimensional port-Hamiltonian PDE (Eq. 23) whose dissipation inequality is used for the convergence argument.

Load-bearing premise

That a weak LaSalle invariance principle holds in the infinite-dimensional function spaces once the largest invariant set inside the zero-dissipation set has been shown to be only the optimal trajectory.

What would settle it

Construct an explicit linear controllable plant and strictly convex cost for which the PDE dynamics (17)–(23) with the stated boundary conditions remain bounded away from the unique Pontryagin solution for all algorithmic time; or prove that no such counter-example exists in the chosen Sobolev-type spaces.

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

If this is right

  • Numerical schemes for the two-time port-Hamiltonian PDE can be truncated early to obtain real-time sub-optimal controllers with built-in passivity guarantees.
  • The same PDE system can be interconnected with other port-Hamiltonian plants (physical or optimizer dynamics) while preserving a shifted total Hamiltonian as a Lyapunov function.
  • State and input constraints can be incorporated by replacing gradients with subdifferentials inside the same port-Hamiltonian structure.
  • Distributed optimal control problems become power-preserving interconnections of infinite-dimensional primal-dual gradient systems.

Where Pith is reading between the lines

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

  • Because the dissipation inequality is exact, any consistent spatial discretization of the PDE should inherit a discrete passivity property that can be exploited for certified early termination in model-predictive control.
  • The two-time structure suggests a natural multi-rate implementation: fast algorithmic iterations nested inside a slower physical-time receding-horizon loop.
  • If the same construction works for nonlinear plants whose dynamics remain port-Hamiltonian, the approach would cover a much larger class of optimal-control problems without losing the Lyapunov argument.

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

2 major / 5 minor

Summary. The note extends the known incremental port-Hamiltonian formulation of continuous-time primal-dual gradient dynamics from static constrained convex optimization to finite-horizon convex optimal control with linear dynamics. Interpreting the dynamics as infinite-dimensional equality constraints yields a two-time-scale system of PDEs (physical time t and algorithmic time τ) that is formally incremental port-Hamiltonian (Eqs. 17 and 23). Equilibria of the algorithmic dynamics coincide with the first-order conditions of Pontryagin’s minimum principle (13). A shifted Hamiltonian functional is constructed whose formal time derivative is non-positive under the mixed boundary conditions (25); Proposition 3.2 shows that the largest invariant set inside the zero-derivative set is the single optimal trajectory when (A,B) is controllable. The author argues that the PDE formulation can serve as a starting point for sub-optimal control schemes and for interconnection with other port-Hamiltonian systems.

Significance. If the indicated convergence can be made rigorous, the paper supplies a clean structural bridge between continuous-time primal-dual methods and optimal control that inherits passivity and compositionality properties of port-Hamiltonian systems. This is potentially useful for “instant MPC”-style sub-optimal controllers and for distributed or interconnected optimization, and it clarifies the infinite-dimensional constructions already appearing in the recent literature (especially Gernandt–Schaller). The formal derivation of the skew-adjoint operator J, the energy balance, and the identification of equilibria with the Pontryagin conditions are clean and self-contained. The contribution is therefore of genuine interest as a short note, provided the convergence claim is either completed or carefully qualified.

major comments (2)
  1. After Proposition 3.2 the manuscript only asserts that a weak infinite-dimensional LaSalle principle is “expected” for suitable function spaces and technical conditions. The central claim of convergence to the optimal control therefore remains incomplete. Either a precise function-space setting and a reference (or sketch) of the applicable invariance principle should be supplied, or the abstract and conclusions should be rephrased to state only that the largest invariant set is the optimal trajectory and that asymptotic stability is indicated but not proved.
  2. The energy balance (24) contains residual boundary terms that vanish only under the mixed conditions (25). Remark 3.3 correctly notes that without these conditions one obtains a boundary-controlled port-Hamiltonian system, yet the paper never returns to the question of how such boundary ports would be used for interconnection or for free-endpoint problems. A short clarification of the intended scope (fixed versus free terminal state) would strengthen the claim that the PDE system is ready for interconnection.
minor comments (5)
  1. The title uses “Suboptimal” while the abstract and body use “sub-optimal”; consistent hyphenation would improve polish.
  2. References [14] and [15] appear to be identical; one should be removed.
  3. In the display of the operator J (18) the placement of the partial-derivative symbols is slightly ambiguous; a short remark that they act only on the co-state component would help the reader.
  4. The phrase “compromizing between” (Section 2) is a typographical error for “compromising between.”
  5. A brief pointer to the precise domains that make J skew-adjoint (already referenced to [5]) would make the note more self-contained.

Circularity Check

0 steps flagged

No significant circularity: the PDE port-Hamiltonian form is obtained by rewriting the standard infinite-dimensional primal-dual vector field in energy variables; self-citations supply background structure only.

full rationale

The derivation chain is self-contained and non-circular. Section 2 recalls the classical continuous-time primal-dual gradient algorithm for static convex optimization and rewrites it as an incremental port-Hamiltonian system by the change of variables z = Qq, µ = Λλ and the quadratic Hamiltonian (5); the shifted Hamiltonian (7) then yields the standard Lyapunov decrease (9). Section 3 applies exactly the same construction to the optimal-control Lagrangian L obtained from the integral of F = p⊤(Ax + Bu - ẋx) + K. The variational derivatives produce the PDE system (17) whose operator J is formally skew-adjoint by construction; the energy variables (20) and Hamiltonian functional (21) simply rename the same vector field as an infinite-dimensional incremental port-Hamiltonian system. Equilibria of (17) are identical to the Pontryagin first-order conditions (13) by direct substitution, and the energy balance (24) follows from formal skew-adjointness plus convexity of K. Self-citations ([2,3,5,10–14]) supply the general definition of (incremental) port-Hamiltonian systems and the infinite-dimensional Stokes-Dirac framework; they are not used to force uniqueness of the optimal trajectory or to define any quantity that later reappears as a prediction. The only incompleteness is the sketched infinite-dimensional LaSalle argument after Proposition 3.2, which is an open technical gap rather than a circular step. Consequently the circularity score is 1 (minor background self-citation, no load-bearing reduction).

Axiom & Free-Parameter Ledger

0 free parameters · 4 axioms · 1 invented entities

The central claim rests on standard convex-analysis and port-Hamiltonian facts plus a few domain-specific modeling choices (linear dynamics, strict convexity of K, controllability). No free parameters are fitted; the only invented entity is the two-time-scale PDE itself, which is derived rather than postulated. The weakest external reliance is the infinite-dimensional LaSalle principle left unproved.

axioms (4)
  • domain assumption K(x,u) is strictly convex and differentiable; the dynamics are linear and time-invariant.
    Invoked from the opening of Section 3 onward to guarantee uniqueness of the Pontryagin solution and monotonicity of the gradient map.
  • domain assumption The pair (A,B) is controllable (equivalently (B⊤,A⊤) is observable).
    Used in Proposition 3.2 to conclude that the only trajectory remaining in the zero-dissipation set is the optimal co-state.
  • standard math The operator J defined in (18) is formally skew-adjoint on a suitable dense domain (with boundary conditions (25)).
    Taken from the Stokes-Dirac structure literature [14] and the domain discussion in [5]; needed for the energy balance (24).
  • ad hoc to paper A weak infinite-dimensional LaSalle invariance principle holds in the chosen function spaces.
    Explicitly left as an expectation after Proposition 3.2; without it the convergence claim remains only formal.
invented entities (1)
  • Two-time-scale port-Hamiltonian PDE system (physical time t and algorithmic time τ) for optimal control no independent evidence
    purpose: To realize the continuous-time primal-dual gradient algorithm as an infinite-dimensional incremental port-Hamiltonian system whose equilibria recover Pontryagin optima.
    Derived from the variational derivatives of the integral Lagrangian; not an extra physical postulate, yet the concrete PDE object is new to the paper.

pith-pipeline@v1.1.0-grok45 · 13754 in / 2793 out tokens · 103406 ms · 2026-07-14T17:45:27.608243+00:00 · methodology

0 comments
read the original abstract

This note generalizes the port-Hamiltonian formulation of the continuous time primal-dual gradient algorithm for static constrained convex optimization to the convex optimal control problem.The resulting dynamics is shown to be a port-Hamiltonian system of partial differential equations, involving ordinary physical time as well 'algorithmic' time. Convergence to the optimal control solution is indicated, and it is argued that sub-optimal control strategies could be derived starting from the partial differential equation formulation.

Review history (2 revisions) →

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

16 extracted references

  1. [1]

    Arrow, L

    K. Arrow, L. Hurwitz, H. Uzawa, Studies in Linear and Non-Linear Programming, Stanford University Press, Stanford, California, 1958

    1958

  2. [2]

    Camlibel, A.J

    M.K. Camlibel, A.J. van der Schaft, Incrementally port-Hamiltonian systems, 52nd IEEE Conf. Decision and Control, pp. 2538–2543, Flo- rence, Italy, 2013

    2013

  3. [3]

    Camlibel, A.J

    M.K. Camlibel, A.J. van der Schaft, Port-Hamiltonian systems theory and monotonicity, SIAM J. Contr. Optimization 61(4):2193–2221, 2023

    2023

  4. [4]

    Cherukuri, E

    A. Cherukuri, E. Mallada, J. Cortes, Asymptotic convergence of con- strained primal-dual dynamics, Systems Control Lett., 87:10–15, 2016

    2016

  5. [5]

    Gernandt, M

    H. Gernandt, M. Schaller, Port-Hamiltonian structures in infinite- dimensional optimal control: Primal-dual gradient method and control- by-interconnection, Systems Control Lett., 197, 106030, 2025

    2025

  6. [6]

    Gernandt, T

    H. Gernandt, T. Preuster, M. Schaller, Optimization-based con- trol by interconnection of nonlinear port-Hamiltonian systems, arXiv 2602.06670, 2026

    arXiv 2026

  7. [7]

    Meinsma, A

    G. Meinsma, A. van der Schaft, A Course on Optimal Control, Springer Nature, Switzerland, 2023

    2023

  8. [8]

    T. Pham, N. Vu, I. Prodan, L. Lefevre, A combined Control by Interconnection-Model Predictive Control design for constrained Port- Hamiltonian systems, Systems Control Lett., 167, 105336, 2022

    2022

  9. [9]

    Rawlings, D.Q

    J.B. Rawlings, D.Q. Mayne, M.M. Diehl, Model Predictive Control: Theory, Computation, and Design, 2nd edition, Nob Hill Publishing Madison, WI, 2017

    2017

  10. [10]

    Stegink, C

    T. Stegink, C. De Persis, A. van der Schaft, Port-Hamiltonian for- mulation of the gradient method applied to smart grids, IFAC- PapersOnLine, 48(13):13–18, 2015

    2015

  11. [11]

    T. W. Stegink, C. De Persis, A.J. van der Schaft, A unifying energy- based approach to stability of power grids with market dynamics, IEEE Transactions on Automatic Control 62(6):2612-2622, 2017

    2017

  12. [12]

    A.J. van der Schaft, L 2-Gain and Passivity Techniques in Nonlinear Control, 3rd edition, Springer International Publishing, Communica- tions and Control Engineering series, 2017, Cham, Switzerland

    2017

  13. [13]

    van der Schaft and D

    A. van der Schaft and D. Jeltsema. Port-Hamiltonian systems theory: An introductory overview. Foundations and Trends in Systems and Control, 1(2-3):173–378, 2014

    2014

  14. [14]

    van der Schaft, B.M

    A.J. van der Schaft, B.M. Maschke, Hamiltonian formulation of dis- tributed parameter systems with boundary energy flow, J. of Geometry and Physics, 42:166–174, 2002

    2002

  15. [15]

    van der Schaft, B.M

    A.J. van der Schaft, B.M. Maschke, Hamiltonian formulation of distributed-parameter systems with boundary energy flow, Journal of Geometry and Physics, 42:166–194, 2002

    2002

  16. [16]

    Yoshida, M

    K. Yoshida, M. Inoue, T. Hatanaka, Instant MPC for linear systems and dissipativity-based stability analysis, IEEE Control Syst. Lett., 3(4):811–816, 2019

    2019