pith. sign in

arxiv: 2604.07693 · v1 · submitted 2026-04-09 · 🧮 math.OC · cs.SY· eess.SY

On Linear Critical-Region Boundaries in Continuous-Time Multiparametric Optimal Control

Lida Lamakani , Efstratios N. Pistikopoulos This is my paper

Pith reviewed 2026-05-10 18:16 UTC · model grok-4.3

classification 🧮 math.OC cs.SYeess.SY
keywords multiparametric optimal controlcritical regionscontinuous-time systemsexplicit model predictive controlhyperplane boundariescostate linearityparametric programming
0
0 comments X

The pith

A critical-region boundary in continuous-time multiparametric optimal control is a hyperplane if and only if the relevant extremum occurs at the initial or terminal time.

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

The paper establishes when the dividing surfaces between different explicit control laws remain simple hyperplanes rather than curved surfaces in the space of initial conditions. This matters because hyperplanes can be found analytically from matrix exponentials, letting the entire partition be built without numerical search for each region. The key is that the costate evolves linearly from the initial state at any fixed time, so an extremum fixed at an endpoint produces a linear condition on the initial state. When the time of the extremum varies with the initial state, such as a switching instant, the boundary curves. They illustrate this by deriving the full three-dimensional partition for a third-order example analytically.

Core claim

We show that a boundary is a hyperplane if and only if the relevant extremum is attained at either the initial time or the terminal time, regardless of the initial condition. The reason is that the costate is a linear function of the initial state at any fixed time, so when the extremum is tied to a fixed endpoint, the boundary condition is linear and the boundary normal follows directly from two matrix exponentials and a linear solve. When the extremum occurs at a time that shifts with the initial condition, the boundary is generally curved.

What carries the argument

The timing of the extremum of a Lagrange multiplier or constraint function along the optimal trajectory, which determines whether the boundary condition remains linear in the initial state.

If this is right

  • The complete critical-region partition can be obtained analytically for low-order systems by checking only endpoint extrema.
  • Comparison to discrete-time shows the continuous-time partition has far fewer regions and remains unchanged under time discretization.
  • The boundary normal is computed explicitly from two matrix exponentials and a linear solve when the condition holds.
  • Online evaluation of the control law stays simple because each region uses a closed-form expression valid over a polyhedral domain.

Where Pith is reading between the lines

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

  • Extending the result to systems with state-dependent switching times might require hybrid analysis to locate interior extrema.
  • Applications in real-time control could benefit from precomputing only the linear boundaries and handling curved ones via approximation.
  • Similar linearity arguments may apply to parametric nonlinear programs where the parameter enters linearly in the costate equation.

Load-bearing premise

The costate is a linear function of the initial state at any fixed time.

What would settle it

Construct a specific optimal control problem where an extremum at the initial time produces a visibly curved boundary between critical regions.

Figures

Figures reproduced from arXiv: 2604.07693 by Efstratios N. Pistikopoulos, Lida Lamakani.

Figure 1
Figure 1. Figure 1: CT critical-region partition of Θ = [−2.6, 2.6] × [−0.9, 0.9] × [−0.7, 0.7]. Cube faces are colored by critical region; black lines are cross￾sections of the four hyperplanes (15)–(18). Straight edges on every face confirm the endpoint condition of Proposition 1. Computing eMf tf and eMstf and applying (11): ℓ1 : 0.2084 x01 + 0.8613 x02 + 0.2650 x03 = −0.4, (15) ℓ2 : 0.2084 x01 + 0.8613 x02 + 0.2650 x03 = … view at source ↗
Figure 3
Figure 3. Figure 3: DT partition at N = 10: 77 critical regions. The region count grows as more nodes bring new active-set combinations. principle extends to path and output constraints through an analogous argument, as discussed in Remark 1. The third-order demonstration gives the first three￾dimensional critical-region partition for this problem class, with four boundary hyperplanes computed analytically and verified throug… view at source ↗
Figure 2
Figure 2. Figure 2: DT partition at N = 5: 23 critical regions on the same cube. Each region corresponds to a distinct active-constraint pattern across the 5 grid nodes. boundaries are hyperplanes. The hyperplanarity is a separate property: it means the CT boundaries in this example are available analytically from matrix exponentials, whereas the DT boundaries require solving the full multiparametric QP. Table IV also illustr… view at source ↗
read the original abstract

When an optimal control problem is solved for all possible initial conditions at once, the initial-state space splits into critical regions, each carrying a closed-form control law that can be evaluated online without solving any optimization. This is the multiparametric approach to explicit control. In the continuous-time setting, the boundaries between these regions are determined by extrema of Lagrange multipliers and constraint functions along the optimal trajectory. Whether a boundary is a hyperplane, computable analytically, or a curved manifold that requires numerical methods has a direct effect on how the partition is built. We show that a boundary is a hyperplane if and only if the relevant extremum is attained at either the initial time or the terminal time, regardless of the initial condition. The reason is that the costate is a linear function of the initial state at any fixed time, so when the extremum is tied to a fixed endpoint, the boundary condition is linear and the boundary normal follows directly from two matrix exponentials and a linear solve. When the extremum occurs at a time that shifts with the initial condition, such as a switching time or an interior stationary point, the boundary is generally curved. We demonstrate the result on a third-order system, obtaining the complete three-dimensional critical-region partition analytically for the first time in this problem class. A comparison with a discrete-time formulation shows how sharply the region count grows under discretization, while the continuous-time partition remains unchanged.

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 manuscript claims that in continuous-time multiparametric optimal control, a boundary between critical regions in initial-state space is a hyperplane if and only if the relevant extremum of a Lagrange multiplier or constraint occurs at the initial or terminal time. This follows because the costate is linear in the initial state at any fixed time, so the boundary condition yields a hyperplane whose normal is obtained from two matrix exponentials and a linear solve. When the extremum time varies with the initial condition (e.g., switching or interior stationary point), the boundary is generally curved. The result is demonstrated by deriving the complete three-dimensional critical-region partition analytically for a third-order system, with a comparison showing that discretization sharply increases the number of regions while the continuous-time partition remains unchanged.

Significance. If the central claim holds, the result would be significant for explicit MPC, as it supplies an analytical criterion for when boundaries can be computed exactly as hyperplanes rather than approximated numerically. The complete analytical 3-D partition for a third-order example is a concrete strength, as is the explicit contrast with discrete-time formulations. The derivation from standard costate linearity at fixed times provides a clean theoretical handle on the geometry of the partition.

major comments (2)
  1. Abstract: The statement that 'the costate is a linear function of the initial state at any fixed time' regardless of the initial condition is load-bearing for the hyperplane claim. In a constrained linear system the active-set sequence (and therefore the closed-form costate) depends on the initial state, so the map x0 ↦ λ(t*) is piecewise linear across critical regions. The manuscript must show that the identical linear map continues to hold on both sides of the putative hyperplane; otherwise the zero-level set of the boundary condition is linear only locally inside each region and the global boundary need not be a single hyperplane. Please supply the detailed proof steps that establish global linearity at fixed endpoint times.
  2. Third-order demonstration (abstract and associated section): The claim that the complete three-dimensional partition is obtained analytically for the first time requires explicit verification that all boundaries separating regions are hyperplanes precisely when the determining extremum is at an endpoint. The handling of potential interior extrema and the confirmation that no curved boundaries appear in this example should be expanded with the explicit costate expressions and boundary equations used.
minor comments (2)
  1. The system assumptions (linearity of dynamics, quadratic cost, form of state/input constraints) should be stated explicitly at the beginning of the manuscript rather than left implicit.
  2. A brief comparison table or figure caption clarifying the region count in the continuous-time versus discretized formulations would improve readability of the final example.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed and constructive report, which highlights the potential significance of the result for explicit MPC while identifying points that require clarification and expansion. We address each major comment below and will revise the manuscript to incorporate the requested details.

read point-by-point responses

  1. Referee: Abstract: The statement that 'the costate is a linear function of the initial state at any fixed time' regardless of the initial condition is load-bearing for the hyperplane claim. In a constrained linear system the active-set sequence (and therefore the closed-form costate) depends on the initial state, so the map x0 ↦ λ(t*) is piecewise linear across critical regions. The manuscript must show that the identical linear map continues to hold on both sides of the putative hyperplane; otherwise the zero-level set of the boundary condition is linear only locally inside each region and the global boundary need not be a single hyperplane. Please supply the detailed proof steps that establish global linearity at fixed endpoint times.

    Authors: We agree that the global linearity claim requires an explicit proof that the linear map for the costate at a fixed endpoint time is identical on both sides of the boundary. In the revised manuscript we will insert a new subsection (immediately following the statement of the main theorem) that derives this step by step: (i) within each critical region the active-set sequence is constant, yielding an affine closed-form expression for the costate at any fixed t; (ii) at a boundary defined by an endpoint extremum the two adjacent regions differ only by the marginal constraint that is exactly at its limit on the boundary; (iii) because the underlying linear dynamics and the fixed-time boundary condition are unaffected by that single marginal constraint, the matrix exponential and the linear solve that produce λ(t*) remain the same on both sides. Consequently the zero-level set of the boundary condition is a single hyperplane whose normal is obtained from the two matrix exponentials and the linear solve, as claimed. We will also add a short remark noting that this identity fails when the extremum time varies with x0, which is why interior or switching-time boundaries are generally curved. revision: yes

  2. Referee: Third-order demonstration (abstract and associated section): The claim that the complete three-dimensional partition is obtained analytically for the first time requires explicit verification that all boundaries separating regions are hyperplanes precisely when the determining extremum is at an endpoint. The handling of potential interior extrema and the confirmation that no curved boundaries appear in this example should be expanded with the explicit costate expressions and boundary equations used.

    Authors: We will expand the third-order example section with the explicit closed-form costate trajectories (derived from the fixed active-set sequences in each region), the precise algebraic boundary equations obtained by setting the relevant multiplier or constraint to its extremal value at t=0 or t=T, and a table listing every separating surface together with the endpoint time at which its defining extremum occurs. We will also add a short paragraph confirming that, for the chosen third-order linear system, all candidate interior stationary points lie outside the feasible initial-state domain or violate the optimality conditions, so no curved boundaries arise. These additions will supply the explicit verification requested while preserving the analytic character of the partition. revision: yes

Circularity Check

0 steps flagged

No circularity: central claim follows from standard linear-system costate properties

full rationale

The paper derives that boundaries are hyperplanes precisely when extrema occur at fixed initial or terminal times because the costate at any fixed time is linear in the initial state. This linearity is a standard consequence of the linear dynamics and costate equation in linear-quadratic optimal control, not a quantity defined inside the paper or fitted to its own outputs. No equations reduce the boundary claim to a self-referential definition, a renamed empirical pattern, or a load-bearing self-citation whose validity depends on the present result. The derivation chain is therefore independent of the target statement and remains self-contained against external benchmarks from classical optimal control theory.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the standard property that the costate is linear in the initial state at fixed times for linear dynamics; this is treated as background rather than derived or fitted inside the paper.

axioms (1)
  • domain assumption The costate is a linear function of the initial state at any fixed time.
    Invoked directly to obtain linearity of the boundary condition when the extremum occurs at a fixed endpoint.

pith-pipeline@v0.9.0 · 5567 in / 1228 out tokens · 66066 ms · 2026-05-10T18:16:08.120693+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

12 extracted references · 12 canonical work pages

  1. [1]

    The explicit linear quadratic regulator for constrained systems,

    A. Bemporad, M. Morari, V . Dua, and E. N. Pistikopoulos, “The explicit linear quadratic regulator for constrained systems,”Automatica, vol. 38, no. 1, pp. 3–20, 2002

    work page 2002

  2. [2]

    An algorithm for multi-parametric quadratic programming and explicit mpc solutions,

    P. Tøndel, T. A. Johansen, and A. Bemporad, “An algorithm for multi-parametric quadratic programming and explicit mpc solutions,” Automatica, vol. 39, no. 3, pp. 489–497, 2003

    work page 2003

  3. [3]

    E. N. Pistikopoulos, N. A. Diangelakis, and R. Oberdieck,Multi- parametric optimization and control. John Wiley & Sons, 2020

    work page 2020

  4. [4]

    Multi-parametric model-based control: theory and applications,

    E. N. Pistikopoulos, M. C. Georgiadis, and V . Dua, “Multi-parametric model-based control: theory and applications,”(No Title), 2007

    work page 2007

  5. [5]

    Model predictive control: theory, computation, and design,

    J. B. Rawlings, D. Q. Mayne, and M. M. Diehl, “Model predictive control: theory, computation, and design,”(No Title), 2020

    work page 2020

  6. [6]

    Design of robust model-based controllers via parametric programming,

    V . Sakizlis, N. M. Kakalis, V . Dua, J. D. Perkins, and E. N. Pis- tikopoulos, “Design of robust model-based controllers via parametric programming,”Automatica, vol. 40, no. 2, pp. 189–201, 2004

    work page 2004

  7. [7]

    Explicit hybrid model-predictive control: The exact solution,

    R. Oberdieck and E. N. Pistikopoulos, “Explicit hybrid model-predictive control: The exact solution,”Automatica, vol. 58, pp. 152–159, 2015

    work page 2015

  8. [8]

    Explicit model predictive control: A connected-graph approach,

    R. Oberdieck, N. A. Diangelakis, and E. N. Pistikopoulos, “Explicit model predictive control: A connected-graph approach,”Automatica, vol. 76, pp. 103–112, 2017

    work page 2017

  9. [9]

    Multiparametric continuous- time optimal control via pontryagin’s maximum principle: explicit solu- tions and comparisons with discrete-time formulations,

    L. Lamakani and E. N. Pistikopoulos, “Multiparametric continuous- time optimal control via pontryagin’s maximum principle: explicit solu- tions and comparisons with discrete-time formulations,”arXiv preprint arXiv:2603.16887, 2026

    work page arXiv 2026

  10. [10]

    Explicit solutions to op- timal control problems for constrained continuous-time linear systems,

    V . Sakizlis, J. Perkins, and E. Pistikopoulos, “Explicit solutions to op- timal control problems for constrained continuous-time linear systems,” IEE Proceedings-Control Theory and Applications, vol. 152, no. 4, pp. 443–452, 2005

    work page 2005

  11. [11]

    L. S. Pontryagin,Mathematical theory of optimal processes. Routledge, 2018

    work page 2018

  12. [12]

    Ppopt-multiparametric solver for explicit mpc,

    D. Kenefake and E. N. Pistikopoulos, “Ppopt-multiparametric solver for explicit mpc,” inComputer Aided Chemical Engineering. Elsevier, 2022, vol. 51, pp. 1273–1278

    work page 2022