Disturbance-adaptive Model Predictive Control for Bounded Average Constraint Violations

Colin N. Jones; Jicheng Shi

Disturbance-adaptive Model Predictive Control for Bounded Average Constraint Violations

T0 review · 1 major / 1 minor · reviewed 2026-05-22 · grok-4.3

Challenge this review Download PDF Read on arXiv ↗

Pith's one-line read Disturbance-adaptive MPC updates its model from observed violations to bound long-term average constraint breaches even without knowing the true disturbance distribution.

desk verdict DAD-MPC adapts the disturbance model from violations to keep average constraint bounds in stochastic MPC, but the abstract gives no mechanism showing why robust invariance survives those updates. read the letter →

arxiv 2503.24169 v3 submitted 2025-03-31 eess.SY cs.SY

Jicheng Shi , Colin N. Jones This is my paper

classification eess.SYcs.SY

keywords modelpredictivecontrolstochasticsystemsaverageconstraintsdisturbanceadaptationrobustinvariancedata-drivenconstraintviolations

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper presents a disturbance-adaptive model predictive control method for linear systems under stochastic disturbances whose distribution is unknown. The controller revises its internal disturbance model whenever state constraints are violated in operation. A robust invariance technique keeps the optimization feasible from one step to the next and produces either asymptotic or robust guarantees that the fraction of violations over time stays below a chosen limit. These guarantees remain valid even when the adapted model is still inexact, which in turn permits the use of sample-based disturbance estimators. Numerical examples show lower total cost than prior methods at the same target violation rate.

What carries the argument

disturbance-adaptive model predictive control (DAD-MPC) that revises the disturbance model from measured violations and applies robust invariance to enforce recursive feasibility and average-violation bounds.

What would settle it

A closed-loop trajectory in which the long-run fraction of state-constraint violations exceeds the target bound despite repeated model updates from the measured violations.

Watch

Extended reading notes

Core claim

DAD-MPC adjusts the disturbance model based on measured constraint violations. Using a robust invariance method, DAD-MPC ensures recursive feasibility and guarantees asymptotic or robust bounds on average constraint violations. The bounds hold even with an inaccurate disturbance model, which allows for data-driven disturbance quantification methods to be used.

Load-bearing premise

The robust invariance property continues to hold after the disturbance model is revised from observed violations.

Editorial extensions

If this is right

The same guarantees apply when the disturbance model is obtained from any data-driven procedure such as conformal prediction.
Recursive feasibility is preserved at every time step even though the disturbance model changes.
Closed-loop cumulative cost is lower than existing methods at identical target violation rates.
The approach works for any linear system where only samples of the disturbance are available rather than its exact distribution.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The adaptation rule could be combined with online set-membership estimation to tighten bounds further as more data arrive.
Similar model-update logic might be applied to tube-based or scenario MPC formulations that already use robust invariant sets.
The method supplies a concrete way to trade off average violation allowance against performance in safety-critical control loops.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

The paper introduces DAD-MPC, which updates the disturbance model online from observed state-constraint violations and pairs that with a robust invariance approach to deliver recursive feasibility plus bounds on average violations. The bounds are claimed to hold even when the model stays inaccurate, which lets the method use data-driven tools such as conformal prediction instead of requiring the true distribution. Simulations reportedly show lower closed-loop cost than existing methods at several target violation rates while still meeting the average bounds. That combination of adaptation and invariance for average constraints is the concrete new piece. It targets a real limitation in constrained stochastic control where exact disturbance laws are rarely available. The practical angle is that you can start with a rough model and let violation data tighten it without losing the guarantees. The soft spot is exactly the one the stress test flags. The abstract states that recursive feasibility and the violation bounds remain after each model update, yet supplies no definition of the invariance set, no explicit update rule, and no argument that the bounds are preserved when the model is still wrong. Without those steps it is impossible to check whether the adaptation can loosen or break the invariance. Since only the abstract is available, the central mathematical support cannot be verified. This work is aimed at the MPC community working on stochastic or robust control with average constraints. Readers who already use invariance or conformal prediction methods might extract the high-level idea and the simulation comparison. A serious referee would be appropriate only if the full paper supplies the missing derivation for how the updated model keeps the invariance intact; on the abstract alone the claims are too thin to assess.

Referee Report

1 major / 1 minor

Summary. The paper proposes a disturbance-adaptive model predictive control (DAD-MPC) framework for stochastic linear time-invariant systems subject to constraints on the average number of state-constraint violations, without knowledge of the disturbance distribution. DAD-MPC adjusts the disturbance model online based on measured violations and employs a robust invariance method to guarantee recursive feasibility together with asymptotic or robust bounds on average violations; these bounds are asserted to remain valid even when the disturbance model is inaccurate, thereby permitting data-driven quantification techniques such as conformal prediction. Simulation results are reported to show reduced closed-loop cumulative cost relative to existing methods while satisfying the average-violation bounds across different target rates.

Significance. If the stated recursive-feasibility and average-violation guarantees can be established, the framework would constitute a useful advance in adaptive MPC for systems with probabilistic average constraints. The explicit tolerance of inaccurate disturbance models and the compatibility with conformal-prediction-style quantification would be practically relevant strengths, allowing online adaptation without requiring an exact disturbance distribution. At present these strengths remain hypothetical because the manuscript supplies no technical development of the invariance construction or the update rule.

major comments (1)

[Abstract] Abstract: the central claim that robust invariance is preserved (and recursive feasibility retained) after online updates of the disturbance model from observed violations is asserted without any derivation, assumption list, invariance definition, or proof sketch. This is load-bearing for every guarantee stated in the abstract.

minor comments (1)

[Abstract] Abstract: the description of the DAD-MPC update rule and the precise form of the robust invariant set could be expanded by one or two sentences to give readers an immediate indication of the adaptation mechanism.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the detailed review and the identification of the central claim in the abstract. We address the major comment below.

read point-by-point responses

Referee: [Abstract] Abstract: the central claim that robust invariance is preserved (and recursive feasibility retained) after online updates of the disturbance model from observed violations is asserted without any derivation, assumption list, invariance definition, or proof sketch. This is load-bearing for every guarantee stated in the abstract.

Authors: The abstract is a concise summary of the paper's main results and is not intended to contain derivations or proofs. The full manuscript provides the complete technical development, including the assumption list, the definition of the robust invariant set, the online disturbance model update rule based on measured violations, and the proof that recursive feasibility and the asymptotic/robust average-violation bounds are retained under the stated conditions (even for inaccurate models). These elements appear in the sections developing the DAD-MPC scheme and the associated invariance arguments. We are willing to add a brief pointer to the relevant theorem in the abstract if the editor deems it helpful for readability. revision: no

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: abstract states guarantees via robust invariance without any visible self-referential definitions, fitted predictions, or load-bearing self-citations.

full rationale

The provided abstract claims that DAD-MPC uses a robust invariance method to ensure recursive feasibility and bounds on average violations, even with inaccurate models. No equations, parameter fits, or derivation steps are shown. No self-citations are referenced as load-bearing, and the guarantees are presented as following from the method without reducing to input data or prior author results by construction. This is the common case of a self-contained high-level claim with no detectable circularity in the given text.

Assumptions & free parameters 0 free parameters · 2 assumptions · 1 invented entities

Review limited to abstract; no numerical free parameters are stated. The two domain assumptions below are the minimal background invoked to support the claimed guarantees.

assumptions (2)

domain assumption The plant is a stochastic linear time-invariant system
Stated in the first sentence of the abstract as the system class under consideration.
domain assumption Robust invariance can be maintained after disturbance-model updates driven by observed violations
Invoked to obtain recursive feasibility and the average-violation bounds described in the abstract.

invented entities (1)

DAD-MPC framework
purpose: Online adjustment of the disturbance model from measured constraint violations to enforce average-violation bounds
New controller architecture introduced in the abstract; no external falsifiable evidence supplied.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Disturbance-adaptive Model Predictive Control for Bounded Average Constraint Violations." pith.science (2026). https://pith.science/paper/2503.24169

@misc{pith2026250324169,
  author       = {Pith},
  title        = {Pith review of: Disturbance-adaptive Model Predictive Control for Bounded Average Constraint Violations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2503.24169}},
  note         = {Machine review of arXiv:2503.24169}
}

read the original abstract

This paper considers stochastic linear time-invariant systems subject to constraints on the average number of state-constraint violations over time without knowing the disturbance distribution. We present a novel disturbance-adaptive model predictive control (DAD-MPC) framework, which adjusts the disturbance model based on measured constraint violations. Using a robust invariance method, DAD-MPC ensures recursive feasibility and guarantees asymptotic or robust bounds on average constraint violations. Additionally, the bounds hold even with an inaccurate disturbance model, which allows for data-driven disturbance quantification methods to be used, such as conformal prediction. Simulation results demonstrate that the proposed approach reduces closed-loop cumulative cost compared to state-of-the-art methods across different target violation rates, while satisfying average violation bounds.

Discussion (0). Sign in to comment.

Pith tools

Reviewed May 22, 2026 · model on record in the stance chip above.