Formal Synthesis of Robust Koopman-Model Predictive Control: A Case Study in AC-DC Power Conversion

arXiv:2609.15197 · eess.SY, cs.SY · Submitted 2026-09-14 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Robotics Radio. Generated commentary on the latest robotics and control papers.

Rosa: I'm Rosa.

Dev: And I'm Dev, here with Taro.

Rosa: We're looking at the paper: "Formal Synthesis of Robust Koopman-Model Predictive Control".

Dev: Here's the gist of the paper. Formal synthesis of Robust Koopman-Model Predictive Control (RK-MPC), a novel datadriven approach to formal synthesis of systems with nonlinear dynamics This letter proposes a formal synthesis of the RK-MPC and apply it to an AC-DC converter Core Methodology The proposed work introduces a novel formulation in the RK-MPC that realizes data-driven formal synthesis of controllers for nonlinear systems.

Rosa: This involves encoding Signal Temporal Logic (STL) specifications as inequality constraints with max/min functions and formulating an optimization problem with both these inequality constraints and the Koopman model.

Dev: The formulation is designed to guarantee robust satisfaction of the STL specifications despite modeling error.

Rosa: Robustness Guarantee The robustness of the controller is ensured by tightening the inequality constraints associated with the STL specifications by incorporating a modeling error assumed in the SafEDMD framework.

Paper summary: Rosa: To recap where we are, we've talked about how this paper proposes a novel data-driven approach called Robust Koopman Model Predictive Control, or RK-MPC. The main thesis is that they formalize the control problem by incorporating specifications from Signal Temporal Logic directly into the optimization problem constraints.

Dev: They claim this formulation provides a way to prove closed-loop performance for nonlinear systems even when you have modeling errors in your Koopman model, which is crucial because those models are rarely perfect in reality.

Taro: So they're essentially building a formal bridge between the abstract idea of temporal logic specifications and the actual optimization problem that drives the physical system.

Rosa: And this bridge is made robust by tightening up those constraints using an assumed modeling error, which they manage with the SafEDMD framework to define these tightened predicate functions.

Dev: They apply this whole thing to a specific case study: designing a reliable AC-DC power converter. The control objectives they are targeting involve maintaining specific voltage levels and limiting current excursions, like keeping the mean DC voltage above two hundred fifty volts and currents below eighty two amps.

Taro: It’s interesting that they use these concrete numbers for their application, because it grounds the theoretical work in something tangible—a real power conversion system.

Rosa: And what they show is that numerically, this RK-MPC controller actually satisfies those requirements even when modeling error is present, when compared to a standard K-MPC approach.

Dev: The key result here is that the robustness and feasibility of the system depend on a tuning parameter called c, and they demonstrated that keeping c at zero point zero zero five ensures the specification holds in this study.

Taro: So for someone listening, it means if you're designing a nonlinear controller, you can use this method to get a formal guarantee about meeting your safety requirements under imperfect modeling conditions.

Rosa: It shows how data-driven methods can be used not just to find an optimal solution, but to formally verify its robustness against known types of errors.

Dev: This paper establishes the RK-MPC framework as a way to systematically incorporate temporal logic requirements into the optimization structure for nonlinear control design.

Conclusion: Taro: So wrapping up the idea from "Formal Synthesis of Robust Koopman-Model Predictive Control: A Case Study in AC-DC Power Conversion," it seems this work is about creating a method to formally synthesize controllers for nonlinear systems that are guaranteed to be robust against modeling inaccuracies.

Rosa: And the authors, Shun Hirose, Shiu Mochiyama, and Yoshihiko Susuki, put this into practice by applying it directly to a power converter problem where they have clear voltage and current limits.

Dev: The main implication is that we gain a systematic way to move from abstract specifications, like Signal Temporal Logic constraints, straight into the actual mathematical optimization problem for control design.

Taro: It gives us a concrete tool—the RK-MPC—that can handle nonlinearity and uncertainty simultaneously by incorporating those requirements directly into the constraint set.

Rosa: The paper demonstrates that this approach works in practice on an AC-DC converter, showing that it meets its objectives even with modeling error, provided you tune the parameter c correctly.

Dev: So for anyone interested in control theory applied to nonlinear systems, this paper shows a way to formally verify robustness without needing a perfect model of the physical plant.

Taro: It's about building safety into the synthesis process itself rather than trying to fix it after the fact when things go wrong during operation.

Kyoto University

eess.SY, cs.SY

Submitted: 2026-09-14

Updated: 2026-11-10

Comments: 6 pages, 3 figures, 1 table

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 90/100

The gist: Formal synthesis of Robust Koopman-Model Predictive Control (RK-MPC), a novel datadriven approach to formal synthesis of systems with nonlinear dynamics This letter proposes a formal synthesis of the…

Key concepts

Koopman Model
A mathematical framework that allows complex nonlinear systems to be approximated by linear systems in a higher-dimensional space. This approximation helps in designing controllers by simplifying the control problem, even when the actual system dynamics are nonlinear.
Signal Temporal Logic (STL)
A formal language used to specify desired behaviors for a system over time, such as 'the voltage must stay above X' or 'the current must not exceed Y' during certain periods. These specifications are translated into mathematical constraints for the controller.
Robust Satisfaction
The guarantee that the designed controller will successfully meet all specified performance requirements even when there is some inaccuracy in how the system was modeled. This robustness is achieved by adding extra safety margins to the constraints.

Terminology

Summary

Formal synthesis of Robust Koopman-Model Predictive Control (RK-MPC), a novel datadriven approach to formal synthesis of systems with nonlinear dynamics This letter proposes a formal synthesis of the RK-MPC and apply it to an AC-DC converter

Core Methodology

The proposed work introduces a novel formulation in the RK-MPC that realizes data-driven formal synthesis of controllers for nonlinear systems. This involves encoding Signal Temporal Logic (STL) specifications as inequality constraints with max/min functions and formulating an optimization problem with both these inequality constraints and the Koopman model. p. 0 The formulation is designed to guarantee robust satisfaction of the STL specifications despite modeling error. p. 0

Robustness Guarantee

The robustness of the controller is ensured by tightening the inequality constraints associated with the STL specifications by incorporating a modeling error assumed in the SafEDMD framework. Specifically, Theorem 2 proves that this formulation guarantees robust satisfaction of the STL specifications despite modeling error. p. 0 This is achieved by introducing tightened predicate functions corresponding to the predicates and formulating the tightened robustness function. p. 2

Application to Power Conversion

The study applies this formal synthesis to a reliable design of an AC-DC power converter. The plant is modeled as a nonlinear control-affine system describing an AC-DC converter. p. 3 The control objectives are specified using STL, such as maintaining the mean value of the DC voltage at or above 250 V and limiting the excursion of the AC current to less than 82 A. p. 3

Numerical Validation

Numerical results demonstrate that the proposed RK-MPC controller meets the specification despite modeling error. The comparison with the standard K-MPC controller shows that for varying degrees of modeling error, the proposed formulation guarantees robust satisfaction of the specification. p. 5 The robustness and feasibility are shown to be affected by the tuning parameter c, being ensured as long as c = 0. p. 5005 in this case study. p. 0

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Formal synthesis of Robust Koopman-Model Predictive Control enables data-driven formal synthesis of controllers for nonlinear systems by encoding Signal Temporal Logic specifications as inequality constraints with max/min functions. This allows the system to guarantee robust satisfaction of the STL specifications despite the modeling error.

  2. The proposed formulation solves an optimization problem (5a) that incorporates Koopman model dynamics (5b), input constraints (5d), and temporal logic specifications through tightened robustness functions (5e and 5f). This means the AI system can explicitly account for modeling error in its control strategy to guarantee performance even when the underlying nonlinear model is imperfect.

  3. The improved system can reliably design AC-DC power converters by setting STL specifications for AC current and DC voltage, leading to a closed-loop response that satisfies the STL formula phi (Theorem 2). This ensures mission-critical reliability against load fluctuations and transient events like voltage sags, as demonstrated in the application section.

Related papers