Pith. sign in

REVIEW 5 major objections 6 minor 29 references

Distributed Model Predictive Control Design for Multi-agent Systems via Bayesian Optimization

T0 review · 5 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read This paper proves that a distributed model predictive controller can reproduce the optimal policy of a multi-agent system, even when every local model is wrong, by modifying the local costs and learning them with coordinated Bayesian…

desk verdict A useful algorithmic combination for DMPC tuning, but the claimed proofs of optimality and convergence are not delivered; worth major revision, not publication as is. read the letter →

arxiv 2501.12989 v2 pith:WO2EZ5JJ submitted 2025-01-22 eess.SY cs.SY

Hossein Nejatbakhsh Esfahani , Kai Liu , Javad Mohammadpour Velni This is my paper · ORCID
classification eess.SYcs.SY
keywords distributedmodelpredictivecontrolmulti-agentsystemsBayesianoptimizationADMMMarkovdecisionprocesscostfunctionlearningmismatchclosed-loopperformance
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

This paper claims that a distributed model predictive controller can be tuned, using only closed-loop data and not an accurate model, to reproduce the optimal policy of the true multi-agent Markov decision process. The central theorem shows that no matter how inaccurate the local prediction models are, one can modify each agent's terminal and stage costs so that the N-step value function of the local MPC exactly equals the true local value function. Because that modification is built from the unknown true value function, the paper then parameterizes the DMPC costs and learns the parameters with an ADMM-coordinated multi-agent Bayesian optimizer, proving monotone decrease of the associated augmented Lagrangian. A sympathetic reader should care because this gives a principled, decomposition-friendly route to optimal distributed control under model mismatch, exactly the regime where standard DMPC degrades.

What carries the argument

The load-bearing object is the telescoping cost modification: choose the modified terminal cost as the true value function $V^{\pi}_i(\hat{s}^{k+N})$ and the modified stage cost as the temporal-difference-like difference $V^{\pi}_i(\hat{s}^{\ell}) - \gamma \mathbb{E}[V^{\pi}_i(\hat{s}^{\ell+1})]$. Summing these along an $N$-step trajectory cancels every intermediate term and leaves the current true value $V^{\pi}_i(\tilde{s}^k)$, regardless of the model used to generate the trajectory, provided the expectation is finite on $\Xi$ as in Assumption 2. In the deterministic DMPC variant, the same cancellation is enforced by forcing the slack variable to start at the neighbor's influence $W_{ij}(s^k_j)$ and propagate through $\delta^{\ell}_i$. The coordinated learning part rests on the ADMM split of the global closed-loop objective into local Bayesian-optimization problems with a consensus constraint $\zeta_i = \bar{\zeta}$, using the expected-improvement acquisition whose smoothness and monotonicity are proved in Lemmas 1 and 2.

What would settle it

Take a two-agent linear system with a known quadratic optimal value function $V^\star_i$. Implement the modified costs from Theorem 1 ($\hat{V}^{f}_i = V^{\pi}_i$ at the terminal state, $\hat{L}_i = V^{\pi}_i - \gamma \mathbb{E}[V^{\pi}_i(\text{next})]$) in the DMPC scheme (28) under a deliberately wrong model, simulate the closed loop, and compare the resulting $N$-step value function with $V^{\pi}_i(\tilde{s}^k)$ at every visited augmented state in $\Xi$; any mismatch would falsify the claimed equality. Separately, on a non-convex closed-loop performance landscape with a known global optimum, run Algorithm 1 and record whether the iterates converge to that optimum; persistent convergence to a worse local optimum would show that the monotone-decrease guarantee is weaker than what the paper's convergence language suggests.

Watch

Extended reading notes

Core claim

Under Assumption 2, that the expected value of the local value function stays finite on the relevant set of augmented states, Theorem 1 constructs, for any joint policy $\pi$, modified local terminal and stage costs $\hat{V}^{f}_i(\hat{s}^{k+N}) = V^{\pi}_i(\hat{s}^{k+N})$ and $\hat{L}_i(\hat{s}^{\ell},\pi_i(\hat{s}^{\ell})) = V^{\pi}_i(\hat{s}^{\ell}) - \gamma \mathbb{E}[V^{\pi}_i(\hat{s}^{\ell+1})]$ such that the $N$-step value function of the modified DMPC telescopes to exactly $V^{\pi}_i(\tilde{s}^k)$, the true local value function of the multi-agent MDP, for any horizon $N$. Corollary 1 carries the same equality over to the deterministic DMPC scheme (28), in which the initial slack variable is forced to equal the neighbor-coupling term $W_{ij}(s^k_j)$ and the slack is propagated along the horizon as $\bar{w}^{\ell+1}_i = \bar{w}^{\ell}_i + \delta^{\ell}_i$; the telescoping cancellation then leaves $V^{\pi}_i(\tilde{s}^k)$ even though the trajectories are generated by the wrong model $\hat{f}_i$. Since these constructions require the unknown $V^{\pi}_i$, the paper parameterizes the local costs (and optionally dynamics and constraints) as $V^{\theta_i}_i$, and proposes to learn the parameters $\theta_i$ by ADMM-based multi-agent Bayesian optimization, minimizing the closed-loop performance $J^N_i(\zeta)$ with consensus variables; Lemma 1 and Lemma 2 establish differentiability, gradient Lipschitzness, and monotonicity of the expected-improvement acquisition function, which yields monotone decrease of the augmented Lagrangian under the ADMM updates and stationarity conditions at a fixed point. Numerical examples on a three-agent linear system and on a wheeled-mobile-robot formation task show coupling constraints satisfied and formation achieved after learning, whereas the unlearned DMPC violates them.

Load-bearing premise

The load-bearing premise is that the parameterized DMPC family (33) contains a parameter choice $\theta^*$ whose value functions closely approximate the true optimal value functions; the paper assumes this expressivity without providing an approximation bound, and it is separate from the cost-modification theorem, which uses the true value function directly.

Editorial extensions

If this is right

  • If Theorem 1 and Corollary 1 hold, then model mismatch in the local prediction models does not prevent the distributed MPC architecture from representing the exact optimal value function; only the cost (and possibly parameterized dynamics and constraints) must be adjusted.
  • The proposed ADMM-MABO algorithm gives a parallel, communication-based way to learn these adjustments from closed-loop data, and the monotone-decrease result for the augmented Lagrangian provides a stationarity guarantee at any fixed point of the learning loop.
  • The examples demonstrate the practical consequence: coupling constraints that are violated under an unlearned dual-decomposition DMPC are satisfied after coordinated learning, and the learned scheme outperforms running independent Bayesian optimizers per agent.
  • The non-myopic acquisition function of Proposition 1 can be substituted into the same MABO structure without changing the architecture, so the framework is not tied to the greedy expected-improvement choice.

Reading between the lines

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

  • Editorial inference: the paper's existence proof does not quantify how expressive the parametric class in (33) must be; if that class cannot approximate $V^\star_i$ well, the learned DMPC may converge to a consistent but strictly suboptimal policy, and measuring this gap would be a direct way to judge practical utility.
  • Editorial inference: the monotone-decrease argument for the augmented Lagrangian is not a convergence guarantee for the non-convex BO objective; a natural testable extension is to benchmark ADMM-MABO against known global optima on non-convex closed-loop costs to see how often it finds the global rather than a local optimum.
  • Editorial inference: because the telescoping construction works through expectations, the same cost modification should extend to stochastic transition models (for example, GP-based MPC) by replacing the deterministic DMPC trajectories with their expectations; the paper restricts itself to deterministic local models.
  • Editorial inference: anchoring the slack propagation to $W_{ij}(s^k_j)$ suggests a lightweight variant where agents learn the propagation function $\delta^{\ell}_i$ itself rather than only the costs, which would let the MABO correct for model mismatch in the coupling dynamics too.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 6 minor

Summary. The manuscript proposes a method for learning parameterized distributed model predictive control (DMPC) schemes using multi-agent Bayesian optimization (MABO). It first establishes a formal connection between DMPC and multi-agent Markov decision processes by modifying local terminal and stage cost functions, then introduces a parametric DMPC and an ADMM-based MABO algorithm for coordinated learning. The paper claims rigorous proofs of optimality and convergence of the learning method, and reports two numerical examples (a linear multi-agent system and a formation-control problem with wheeled mobile robots) showing improved closed-loop performance relative to non-learned or non-coordinated DMPC.

Significance. If the theoretical claims were sound, the paper would provide a useful bridge between DMPC and multi-agent Bayesian optimization, enabling data-driven tuning of distributed controllers under model mismatch. The algorithmic structure, particularly the ADMM-based coordinated BO formulation, is reasonable and the numerical examples are suggestive. However, the central theoretical contributions—the optimality of the modified DMPC and the convergence of the MABO algorithm—are not established by the proofs provided. The main value of the manuscript is therefore conditional on repairing the theory; in its current form, the claims substantially exceed what is demonstrated.

major comments (5)
  1. [Section III-B, Theorem 1 (Eqs. 15-18)] Theorem 1 is a definitional identity rather than a structural result. The proof chooses the modified terminal cost in (17a) and stage cost in (17b) as telescoping differences of the true value function V_i^pi, which forces Eq. (16) to hold by construction. The theorem therefore does not show that a DMPC scheme—whose costs are not chosen with oracle access to V_i^pi—can capture the true value function. To make the claim meaningful, the authors would need to characterize a class of cost functions for which the equality holds without constructing the costs from V_i^pi itself, or provide an approximation bound for using a surrogate value function.
  2. [Section III-B, Theorem 2 (Eqs. 20-25)] The proof of Theorem 2 contains an invalid exchange of minimization and expectation. In Eq. (20), the min over neighboring policies pi_j is moved outside the expectation; then Eq. (21) folds that minimum into the transition model as pi*_j, and Eq. (22) moves min over pi_i inside the expectation. In a stochastic multi-agent system, the neighboring policy affects the transition kernel P_i itself, so the minimax-type interchange requires additional assumptions such as a saddle-point property or a contracting Bellman operator on the relevant function space. Without such justification, the Bellman equation (24) and the conclusion pi*_i in arg min in (25) are not established.
  3. [Section IV-C, Lemmas 1-2 and Remark 5] The convergence analysis does not deliver the claimed result. Lemma 1 asserts Lipschitz continuity of the gradient of alpha_EI but the proof does not bound the derivatives of the GP posterior mean and standard deviation; the statement that these are 'bounded due to the smoothness of GP posterior predictions' is not a proof, and no value of L is given. Lemma 2 shows monotonicity of alpha_EI with respect to two scalar arguments, but this does not imply monotonic decrease of the augmented Lagrangian (66), which also involves lambda_i and the consensus term. Most importantly, Remark 5 explicitly states that convergence analysis of the non-convex ADMM is left to future work, which directly contradicts the abstract's claim of 'rigorous proofs establishing the optimality and convergence of the underlying learning method.'
  4. [Section III-C, parameterized DMPC (33)] The paper assumes without proof that the parameterized DMPC scheme (33) is expressive enough to represent the optimal policy, i.e., that there exists theta* such that V_i^{theta*}_i approximates V*_i. This assumption is structurally distinct from Theorem 1, which constructs modified costs using the true value function, and it is not derived from any approximation theorem. Since the learning algorithm can only search within the parametric class {V_i^{theta_i}}, the optimality claims for the learned controller require either a representability result or an explicit approximation bound. Without such a result, the numerical successes demonstrate tuning benefits but not convergence to an optimal policy.
  5. [Corollary 1 and Eq. (29)] Corollary 1 passes from the expectation-based formulation of Theorem 1 to the deterministic DMPC value function by assuming the trajectories are 'approximately deterministic,' but no quantitative condition or error bound is given for this approximation. Since the stochasticity of the true system is central to the multi-agent MDP formulation, this step is load-bearing for the claim that the DMPC scheme (28) captures V_i^pi. The proof of the corollary should state the approximation error explicitly and identify the regime in which it vanishes.
minor comments (6)
  1. [Section VI] The concluding remarks refer to 'dual composition'; this should presumably read 'dual decomposition.'
  2. [Section III-A, Eq. (9)] The notation 'pi_j in M_i' is not formally defined; it should specify that pi_j ranges over the policies of the neighboring agents in M_i.
  3. [Section II-B, Eq. (6)] The multiplier update appears with superscript I, but the stopping criteria in (7) are stated in terms of mu_i and the text switches between iteration indices; the notation could be made consistent.
  4. [Section IV-C, Lemma 1] The proof of Lemma 1 depends on properties of Phi and phi that are standard, but the statement 'both nabla mu and nabla sigma are bounded due to smoothness' should either be proven or replaced by an explicit assumption on the kernel and the fitted GP.
  5. [Section V, Example 1] The caption of Fig. 2 lists colors in the order gray, green, cyan, red, but the order of appearance in the sentence does not match the order in the figure; please align the description.
  6. [Section IV-A, Eq. (39)] The acquisition function is written as alpha_EI(theta) = -((J* - mu(theta)) Phi(Z) + sigma(theta) phi(Z)); this is the negative expected improvement, so the text should explicitly note the sign convention used for the minimization.

Circularity Check

2 steps flagged · score 8.0 of 10

Theorems 1 and Corollary 1 prove their conclusions by defining the modified costs to telescope the true value function; the claim that the modified DMPC captures the true multi-agent MDP is a definitional equality, not a derived result.

  1. self definitional [Section III-B, Theorem 1, Eqs. (15)-(18)]
    "Let us choose the modified local terminal and stage costs as ˆV f_i(ˆs^{k+N}) = V^π_i(ˆs^{k+N}), (17a) ˆL_i(ˆs^ℓ, π_i(ˆs^ℓ)) = { V^π_i(ˆs^ℓ) − γV^+_i(ˆs^ℓ, a^ℓ_i) if |V^+_i(ˆs^ℓ, a^ℓ_i)|<∞, ∞ Otherwise }, where V^+_i(ˆs^ℓ, a^ℓ_i) = E[V^π_i(ˆs^{ℓ+1}) | a^ℓ = π(s^ℓ)]. ... Using a telescoping sum, the centralized N-step value function (15) is rewritten as ... = V^π_i(˜s^k)."

    Theorem 1's conclusion (16) is obtained by defining the modified costs in terms of the very value function V^π_i it claims to reproduce. The stage cost (17b) is exactly the one-step Bellman residual V^π_i − γE[V^π_i(s')]; substituting it into (15) makes every interior term cancel, leaving V^π_i(˜s^k) for any N and for any imperfect transition model. Thus the existence of modified costs is an identity, not a derived prediction: the equality is forced by the definition of the costs.

  2. self definitional [Section III-B, Corollary 1, Eqs. (29)-(32)]
    "We then select the modified terminal and stage costs to be the same as those in Theorem 1 as ˆT^{µ⋆}_i = V^π_i(ˆx^{k+N}_i, ¯w^{k+N}_i), (30a), ˆL^{µ⋆}_i = V^π_i(ˆx^k_i, ¯w^k_i) − γV^π_i(ˆx^{k+1}_i, ¯w^{k+1}_i). (30b) Using (30), we then rewrite the modified value function (29) as ... Using a telescoping sum, we then obtain ... ˆV^{DMPC} = V^π_i(˜s^k)."

    Corollary 1 repeats the same definitional construction for the DMPC scheme (28): the terminal and stage costs are set directly from V^π_i so that the value function telescopes to V^π_i(˜s^k). No property of the proposed DMPC, of the dual decomposition, of the ADMM multipliers, or of the imperfect model is used beyond boundary conditions; the 'capture' result is true by construction, not by any feature of the distributed MPC mechanism.

full rationale

The main circularity is in Section III: Theorem 1 and Corollary 1 construct modified cost functions as the true value function and its one-step Bellman residual, so the conclusion that the modified DMPC value equals the true multi-agent MDP value is an algebraic identity (Eqs. 17-18 and 30-32). The paper is transparent that this construction is not tractable and then moves to a parameterized scheme (33), for which expressiveness is assumed rather than derived; that is a gap, but not itself circular. The correctness concerns about the convergence analysis—Remark 5 explicitly defers non-convex ADMM convergence to future work, and Theorem 2 interchanges min and expectation without justification—are rigorousness problems, not circularity, so they are not counted in the score. No load-bearing self-citation chain appears: prior works [8]-[12] are background for the MDP-MPC fusion idea. The central Section III result is nonetheless forced by definition, so the paper receives a circularity score of 8 rather than a lower score, despite the numerical studies having independent empirical content.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The central claim rests on several unstated numeric choices (BO/ADMM/MPC hyperparameters) and on unproven structural assumptions, chiefly that the parametric DMPC can represent the optimal policy and that the ADMM-based MABO converges despite non-convexity. The existence theorems are tautological constructions using the true value function, so the practical method is predominantly an empirical tuning procedure.

free parameters (6)
  • θ_i (local MPC parameters) = not reported
    Learned via BO to minimize closed-loop cost; the main tunable parameters in (33).
  • GP hyperparameters (l, σ_J^2, σ_n) = not reported
    Kernel length scale, signal variance, and noise variance in (36) must be chosen or estimated; not specified.
  • ADMM penalty ρ = not reported
    Penalty parameter in (52); affects convergence and consensus accuracy.
  • Slack penalty weights p, p_f = not reported
    Weights in (33a) said to be sufficiently large but no values given.
  • Dual decomposition step size β_i = not reported
    Step size in (6) for multiplier updates.
  • Prediction horizon N and discount factor γ = not reported
    MPC horizon and MDP discount factor used throughout the formulation and examples.
assumptions (5)
  • domain assumption Assumption 1: subsystems have the same state-space dimension.
    Used throughout to define neighbor coupling dimensions in (1d) and Remark 1; limits generality but not the core mechanism.
  • domain assumption Assumption 2: the set Ξ of states with finite expected true value functions is non-empty.
    Technical condition for Theorem 1 to hold; standard boundedness assumption.
  • ad hoc to paper The parameterized DMPC (33) is expressive enough to represent the optimal policy.
    No approximation theorem is given; the paper assumes there exists θ* such that V^{θ*}_i approximates V⋆_i without proof.
  • ad hoc to paper The ADMM-based MABO converges to the global optimum of (50).
    Convergence is claimed but not proven; Remark 5 defers non-convex ADMM convergence to future work.
  • domain assumption Each agent has access to all multipliers through the communication network.
    Remark 4 states each local value function depends on global parameters through multipliers, requiring full communication.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Distributed Model Predictive Control Design for Multi-agent Systems via Bayesian Optimization." pith.science (2026). https://pith.science/paper/WO2EZ5JJ

@misc{pith2026250112989,
  author       = {Pith},
  title        = {Pith review of: Distributed Model Predictive Control Design for Multi-agent Systems via Bayesian Optimization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WO2EZ5JJ}},
  note         = {Machine review of arXiv:2501.12989}
}
read the original abstract

This paper introduces a new approach that leverages Multi-agent Bayesian Optimization (MABO) to design Distributed Model Predictive Control (DMPC) schemes for multi-agent systems. The primary objective is to learn optimal DMPC schemes even when local model predictive controllers rely on imperfect local models. The proposed method invokes a dual decomposition-based distributed optimization framework, incorporating an Alternating Direction Method of Multipliers (ADMM)-based MABO algorithm to enable coordinated learning of parameterized DMPC schemes. This enhances the closed-loop performance of local controllers, despite discrepancies between their models and the actual multi-agent system dynamics. In addition to the newly proposed algorithms, this work also provides rigorous proofs establishing the optimality and convergence of the underlying learning method. Finally, numerical examples are given to demonstrate the efficacy of the proposed MABO-based learning approach.

Figures

Figures reproduced from arXiv: 2501.12989 by the authors.

Figure 1
Figure 1. An overview of the proposed MABO-based DMPC. [PITH_FULL_IMAGE:figures/full_fig_p008_1.png] view at source ↗
Figure 3
Figure 3. The gray lines show the evolution of the first states du [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figure 4
Figure 4. The gray lines show the evolution of control signals d [PITH_FULL_IMAGE:figures/full_fig_p010_4.png] view at source ↗
Figures from the paper (6 more)
Figure 6
Figure 6. Figure 6: Moreover, the BO-DMPC cannot drive the first state [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]
Figure 5
Figure 5. Figure 5: The evolution of the local closed-loop performance o [PITH_FULL_IMAGE:figures/full_fig_p011_5.png]
Figure 6
Figure 6. Figure 6: The evolution of the coupling constraints is shown he [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]
Figure 7
Figure 7. Figure 7: The evolution of the first states is shown here compari [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]
Figure 9
Figure 9. Figure 9: Formation control using the proposed learning-base [PITH_FULL_IMAGE:figures/full_fig_p012_9.png]
Figure 10
Figure 10. Figure 10: The evolution of the local closed-loop performance [PITH_FULL_IMAGE:figures/full_fig_p012_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

29 extracted references · 27 canonical work pages

  1. [1]

    A survey of distributed optimiz ation,

    T. Y ang, X. Yi, J. Wu, Y . Y uan, D. Wu, Z. Meng, Y . Hong, H. Wan g, Z. Lin, and K. H. Johansson, “A survey of distributed optimiz ation,” Annual Reviews in Control , vol. 47, pp. 278–305, 2019

    work page 2019

  2. [2]

    Bemporad and D

    A. Bemporad and D. Barcelli, Decentralized Model Predictive Control . London: Springer London, 2010, pp. 149–178

    work page 2010

  3. [3]

    Distributed model predict ive control: An overview and roadmap of future research opportunities,

    R. Negenborn and J. Maestre, “Distributed model predict ive control: An overview and roadmap of future research opportunities,” IEEE Control Systems Magazine, vol. 34, no. 4, pp. 87–97, 2014

    work page 2014

  4. [4]

    and Shames, I

    Farokhi, F. and Shames, I. and Johansson, K. H., Distributed MPC Via Dual Decomposition and Alternative Direction Method of Mul tipliers, Maestre, Jos´ e M. and Negenborn, Rudy R., Ed. Springer Nethe rlands, 2014

    work page 2014

  5. [5]

    Nonlinear dist ributed model predictive control with dual decomposition and event -based communication approach,

    D. Hammami, S. Maraoui, and K. Bouzrara, “Nonlinear dist ributed model predictive control with dual decomposition and event -based communication approach,” Transactions of the Institute of Measurement and Control, vol. 42, no. 15, pp. 2929–2940, 2020

    work page 2020

  6. [6]

    Cautious mode l predic- tive control using gaussian process regression,

    L. Hewing, J. Kabzan, and M. N. Zeilinger, “Cautious mode l predic- tive control using gaussian process regression,” IEEE Transactions on Control Systems Technology, vol. 28, no. 6, pp. 2736–2743, 2020

    work page 2020

  7. [7]

    Exact mul tiple-step predictions in gaussian process-based model predictive co ntrol: Ob- servations, possibilities, and challenges,

    M. Pfefferkorn, M. Maiworm, and R. Findeisen, “Exact mul tiple-step predictions in gaussian process-based model predictive co ntrol: Ob- servations, possibilities, and challenges,” in 2022 American Control Conference (ACC), 2022, pp. 2829–2836

    work page 2022

  8. [8]

    Reinforcement learning-based control and state esti- mation using model predictive control and moving horizon es timation,

    H. N. Esfahani, “Reinforcement learning-based control and state esti- mation using model predictive control and moving horizon es timation,” PhD thesis, Norwegian University of Science and Technology (NTNU), 2024

    work page 2024

Show all 29 references
  1. [9]

    Data-driven economic nmpc using re inforce- ment learning,

    S. Gros and M. Zanon, “Data-driven economic nmpc using re inforce- ment learning,” IEEE Transactions on Automatic Control , vol. 65, no. 2, pp. 636–648, 2020

    2020

  2. [10]

    Performance-oriented data-driven control: Fusing koopm an operator and mpc-based reinforcement learning,

    H. Nejatbakhsh Esfahani, U. V aidya, and J. Mohammadpou r V elni, “Performance-oriented data-driven control: Fusing koopm an operator and mpc-based reinforcement learning,” IEEE Control Systems Letters , vol. 8, pp. 3021–3026, 2024

    2024

  3. [11]

    Cooperative multi-agen t q-learning using distributed mpc,

    H. N. Esfahani and J. M. V elni, “Cooperative multi-agen t q-learning using distributed mpc,” IEEE Control Systems Letters , vol. 8, pp. 2193– 2198, 2024

    2024

  4. [12]

    Learning-ba sed safety crit- ical model predictive control using stochastic control bar rier functions,

    H. N. Esfahani, S. Ahmadi, and J. M. V elni, “Learning-ba sed safety crit- ical model predictive control using stochastic control bar rier functions,” in 2024 American Control Conference (ACC) , 2024, pp. 1658–1663

    2024

  5. [13]

    Bayesian optimization for policy search in control systems,

    B. Letham et al. , “Bayesian optimization for policy search in control systems,” in American Control Conference (ACC) . IEEE, 2019, pp. 3004–3009

    2019

  6. [14]

    Energy-efficient hvac control via bayesian optimization and mpc,

    J. Gao et al. , “Energy-efficient hvac control via bayesian optimization and mpc,” Energy and Buildings , vol. 225, p. 110307, 2020

    2020

  7. [15]

    Bayesian optimization for nonlinear mpc,

    A. Maheshwari et al. , “Bayesian optimization for nonlinear mpc,” Chemical Engineering Research and Design , vol. 175, pp. 181–192, 2022

    2022

  8. [16]

    Taking the human out of the loop: A review of bayesian optimi zation,

    B. Shahriari, K. Swersky, Z. Wang, R. P . Adams, and N. de F reitas, “Taking the human out of the loop: A review of bayesian optimi zation,” Proceedings of the IEEE , vol. 104, no. 1, pp. 148–175, 2016

    2016

  9. [17]

    Parameter tuning in mpc using bayesia n optimiza- tion,

    T. Chu and F. Y ou, “Parameter tuning in mpc using bayesia n optimiza- tion,” Computers & Chemical Engineering , vol. 137, p. 106869, 2020

    2020

  10. [18]

    Learning-based tuning of mpc using bayesian optimization,

    A. Balakrishna et al. , “Learning-based tuning of mpc using bayesian optimization,” IEEE Transactions on Automatic Control , vol. 66, no. 12, pp. 6041–6047, 2021

    2021

  11. [19]

    Gaussian processes for system identification in mpc,

    C. S. Ong et al. , “Gaussian processes for system identification in mpc,” Journal of Process Control , vol. 49, pp. 75–89, 2017

    2017

  12. [20]

    S tability- informed bayesian optimization for mpc cost function learn ing,

    S. Hirt, M. Pfefferkorn, A. Mesbah, and R. Findeisen, “S tability- informed bayesian optimization for mpc cost function learn ing,” IF AC- PapersOnLine, vol. 58, no. 18, pp. 208–213, 2024, 8th IFAC Conference on Nonlinear Model Predictive Control NMPC 2024

    2024

  13. [21]

    Effic ient performance-based mpc tuning in high dimensions using baye sian opti- mization over sparse subspaces,

    A. Kudva, M. T. Huynh, A. Mesbah, and J. A. Paulson, “Effic ient performance-based mpc tuning in high dimensions using baye sian opti- mization over sparse subspaces,” IF AC-PapersOnLine, vol. 58, no. 14, pp. 458–463, 2024, 12th IFAC Symposium on Advanced Control o f Chemical P...

    2024

  14. [22]

    V alue-decomposition networks for cooperativ e multi-agent learning,

    P . Sunehag, G. Lever, A. Gruslys, W. M. Czarnecki, V . F. Z ambaldi, M. Jaderberg, M. Lanctot, N. Sonnerat, J. Z. Leibo, K. Tuyls, and T. Graepel, “V alue-decomposition networks for cooperativ e multi-agent learning,” CoRR, vol. abs/1706.05296, 2017. [Online]. Available: http...

    2017 arXiv

  15. [23]

    Multi-agent blac k-box opti- mization using a bayesian approach to alternating directio n method of multipliers,

    D. Krishnamoorthy and J. A. Paulson, “Multi-agent blac k-box opti- mization using a bayesian approach to alternating directio n method of multipliers,” IF AC-PapersOnLine, vol. 56, no. 2, pp. 2232–2237, 2023, 22nd IFAC World Congress

    2023

  16. [24]

    Why non-myopic bayesian optimi zation is promising and how far should we look-ahead? a study via rollo ut,

    X. Y ue and R. A. Kontar, “Why non-myopic bayesian optimi zation is promising and how far should we look-ahead? a study via rollo ut,”

  17. [25]

    Efficient rollout strategies for bayesian optimization,

    E. Lee, D. Eriksson, D. Bindel, B. Cheng, and M. Mccourt, “Efficient rollout strategies for bayesian optimization,” in Proceedings of the 36th Conference on Uncertainty in Artificial Intelligence ( UAI), ser. Proceedings of Machine Learning Research, J. Peters and D. S ontag, Ed...

    2020

  18. [26]

    Bayesian optimiz ation with a finite budget: An approximate dynamic programming approach ,

    R. Lam, K. Willcox, and D. H. Wolpert, “Bayesian optimiz ation with a finite budget: An approximate dynamic programming approach ,” in Ad- vances in Neural Information Processing Systems , D. Lee, M. Sugiyama, U. Luxburg, I. Guyon, and R. Garnett, Eds., vol. 29. Curran As socia...

    2016

  19. [27]

    Bayesian optimization as a fle xible and efficient design framework for sustainable process systems ,

    J. A. Paulson and C. Tsay, “Bayesian optimization as a fle xible and efficient design framework for sustainable process systems ,” 2024. [Online]. Available: https://arxiv.org/abs/2401.16373

    2024 arXiv

  20. [28]

    Admm for nonconvex optimization under minima l continuity assumption,

    G. Y uan, “Admm for nonconvex optimization under minima l continuity assumption,” 2025

    2025

  21. [2022]

    Available: https://arxiv.org/abs/1911

    [Online]. Available: https://arxiv.org/abs/1911. 01004

    1911

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.