Pith. sign in

REVIEW 2 major objections 5 minor 36 references

Adaptive Control of Positive Systems with Application to Learning SSP

T0 review · 2 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read A data-driven controller for positive systems keeps the value vector within explicit bounds.

desk verdict Novel data-driven Q-learning setup for positive systems, but Theorem 1's lower-bound constant can be non-positive under the stated assumption, so the main robustness guarantee is unproven as written. read the letter →

arxiv 2412.17012 v3 pith:JKI5JUR3 submitted 2024-12-22 math.OC cs.SYeess.SY

Fethi Bencherki , Anders Rantzer This is my paper · ORCID
classification math.OCcs.SYeess.SY MSC 93C4093D0993C05
keywords adaptivecontrolpositivesystemsdata-drivenQ-learningStochasticShortestPathrobustmodel-freeBellmanequationlinearcostnetworks
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 proposes an adaptive controller for positive linear systems, systems whose state stays nonnegative like buffers, queues, or probabilities, solving an infinite-horizon optimal control problem online without first identifying the dynamics. The controller updates a Q-factor from data correlation matrices with a forgetting factor and solves a data-driven algebraic equation that is the model-free form of the Bellman equation. The central result, Theorem 1, shows that if the estimated model's mismatch to the true plant stays below an explicit bound measured by $\rho$, and if $\rho$ times a stabilizability constant $\beta$ is small, then the data-driven value vector $p(t)$ stays within multiplicative factors of the true optimal value vector $p$; Corollary 1 turns this into a finite bound on cumulative cost. Because the problem class is equivalent to Stochastic Shortest Path, the same policy can be compared with model-free SSP algorithms and is shown numerically to achieve lower regret. The value of the result is an explicit, quantitative trade-off between excitation and robustness to unmodeled dynamics.

What carries the argument

The central objects are the data correlation matrices $\Sigma(t)$ and $\bar{\Sigma}(t)$ with forgetting factor $\lambda$: $\Sigma$ accumulates weighted outer products of past state-input pairs, and $\bar{\Sigma}$ accumulates weighted products of the next state with the current state-input pair. Their ratio $\bar{\Sigma}(t)\Sigma^{-1}(t)$ is an online estimate $(\hat{A}(t),\hat{B}(t))$ of the dynamics, and the difference from $(A,B)$ is $\tilde{\Sigma}(t)\Sigma^{-1}(t)$, the model mismatch controlled by condition (19). The data-driven algebraic equation in the $q$-parameter is the model-free Bellman equation in Q-factor form, and Lemma 1 shows that value iteration on $q$ is algebraically equivalent to value iteration on $p$, so solving (12) is a direct substitute for knowing $(A,B)$. The proof machinery is perturbation analysis of the algebraic equation (3)/(17) together with the comparison lemma, Lemma 2, which says that any supersolution of the Bellman equation gives an upper bound on the optimal value.

What would settle it

Evaluate the constants at a pair with $\rho\beta\in(1/2,1)$, for example $\beta=1.5$ and $\rho=0.4$: then $\hat{\alpha}=1-0.6/0.4=-0.5$, so the claimed lower bound $\hat{\alpha}p\leq p(t)$ is vacuous and the proof's division by $\hat{\alpha}$ is illegitimate. A concrete check would simulate a plant in $\mathcal{M}_\beta$ with a disturbance sequence satisfying (19) at that $\rho$ and test whether $p(t)$ still satisfies a non-vacuous two-sided bound; if it does, the theorem's range can be repaired, and if not, the stated range must be restricted.

Watch

Extended reading notes

Core claim

The paper claims that a policy extracted from the solution $q(t)$ of the data-driven algebraic equation $\left(q(t)-\begin{bmatrix}s\\r\end{bmatrix}\right)^\top \Sigma(t)=\min_{K(t)\in\mathcal{K}(E)} q(t)^\top \begin{bmatrix}I\\ K(t)\end{bmatrix} \bar{\Sigma}(t)$ stabilizes every plant in the model set $\mathcal{M}_\beta$, provided the perturbation bound (19) holds with $\rho\beta<1$. Under that condition, the associated value vector $p(t)=\begin{bmatrix}I & K^\top(t)\end{bmatrix}q(t)$ satisfies $\hat{\alpha}p\leq p(t)\leq \check{\alpha}^{-1}p$ with $\check{\alpha}=1-\rho\beta$ and $\hat{\alpha}=1-\check{\alpha}^{-1}\rho\beta$. This two-sided inequality is the paper's core: it says the adaptive estimate never drifts too far from the true optimal cost vector, and it yields the cumulative-cost bound of Corollary 1, in which a factor $\gamma\to 1$ as $\rho\to 0$ closes the suboptimality gap. The argument works through a perturbation analysis of the algebraic equation, comparing the true plant $(A,B)$ with the estimated pair $(\hat{A}(t),\hat{B}(t))=\bar{\Sigma}(t)\Sigma^{-1}(t)$.

Load-bearing premise

The argument needs $\hat{\alpha}=1-(1-\rho\beta)^{-1}\rho\beta$ to be positive, which is equivalent to $\rho\beta<1/2$, because the proof divides by $\hat{\alpha}$; Theorem 1 only states $\rho\beta<1$, so for mismatch levels between half and one the lower bound on $p(t)$ is not actually established.

Editorial extensions

If this is right

  • If condition (19) holds after some time $t_0$, the adaptive policy stabilizes the closed loop in the sense that the true value vector $p(t)$ remains between $\hat{\alpha}p$ and $\check{\alpha}^{-1}p$, so the controller can be used online without an identification step.
  • The bound quantifies the excitation-robustness trade-off: plants that are harder to stabilize, corresponding to larger $\beta$, require smaller mismatch $\rho$, so more or better excitation must be injected.
  • As the mismatch $\rho$ tends to zero, the constants $\check{\alpha}$ and $\hat{\alpha}$ tend to $1$, and the suboptimality gap in Corollary 1 closes, recovering the optimal policy.
  • Corollary 1 gives a finite-time cumulative cost bound in which only the exploration noise and disturbances appear additively, so regret is controlled by the excitation injected and the disturbance magnitude.
  • For Stochastic Shortest Path instances, the same guarantees transfer via the equivalence of problem classes, and the numerical comparison indicates lower regret than $\epsilon$-decreasing Q-learning.

Reading between the lines

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

  • Condition (19) is computable online from $\Sigma(t)$ and $\bar{\Sigma}(t)$, so the controller could in principle monitor its own validity and increase exploration when the mismatch bound is approached; the paper does not develop this adaptation.
  • The proof's division by $\hat{\alpha}$ indicates that the meaningful robustness margin is $\rho\beta<1/2$, not $\rho\beta<1$ as stated in Theorem 1; sharpening or fixing this gap is a natural next step.
  • Because the algebraic equation is solved by value iteration or a linear program, the same framework could extend to other structured MDPs whose Bellman equation has an affine solution, such as average-cost positive systems or processing-network dispatching.
  • The regret comparison used $\epsilon$-decreasing exploration; equipping the policy with more efficient exploration strategies from the SSP literature is an untested but plausible improvement.
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

2 major / 5 minor

Summary. The paper proposes an adaptive, data-driven controller for a class of positive linear systems with linear costs, based on a 'data-driven algebraic equation' derived from the Q-function Bellman equation. The main results (Theorem 1, Theorem 2, Corollary 1) aim to establish that, under a small-model-mismatch condition (19) and rho*beta < 1, the adaptive policy's value function p(t) remains within constant factors of the optimal p, leading to a suboptimality bound on cumulative cost. A numerical experiment compares the policy with Q-learning on a stochastic shortest path example.

Significance. If established, the result would give a simple online scheme with explicit robustness margins and a quantitative excitation-robustness trade-off for an important class of positive systems, with a reproducible numerical study. The connection to SSP and the data-driven equation are interesting. However, the main theorem contains a load-bearing technical error that invalidates the stated guarantees; the contribution is therefore not yet supported.

major comments (2)
  1. [Section 3.1, Theorem 1 and Appendix A.3] The constant alpha_hat = 1 - rho*beta/alpha_check, with alpha_check = 1 - rho*beta, is positive only if rho*beta < 1/2. The theorem states the result for all rho*beta < 1, but the proof divides by alpha_hat to obtain p <= alpha_hat^{-1} p(t). For rho*beta in [1/2, 1), alpha_hat <= 0, so the division reverses the inequality and the two-sided bound (20) is not derived. For example, rho*beta = 3/4 gives alpha_hat = -2. Since Theorem 2 and Corollary 1 rely on (20), the main stability and cost guarantees are not established under the stated assumptions. The theorem should be restated with the condition rho*beta < 1/2, or the proof must be revised, and the downstream results adjusted accordingly.
  2. [Section 3.1, condition (19)] Condition (19) is assumed to hold for all t >= t0, which is a persistence-of-excitation and small-disturbance condition rather than a consequence of the proposed algorithm. Remark 5 acknowledges this, but the main theorems do not provide any online verifiable certificate for (19). Since the stability guarantee is conditional on this assumption, the paper should state more explicitly that (19) is an additional hypothesis on the exploration signal and the disturbance bound relative to the state, and it should discuss how a practitioner could check it online or what happens when it fails.
minor comments (5)
  1. [Section 4] The displayed transition matrices T(1)-T(4) and cost vectors c(1)-c(4) have inconsistent bracket placement and row lengths, making the SSP instance difficult to verify from the text.
  2. [Definition 1] The chain s <= p <= (beta min_i s_i) 1 <= beta s is unclear: the middle term is a scalar times the all-ones vector, and the final inequality uses elementwise order. Please clarify the intended ordering.
  3. [Corollary 1] The displayed bound would be more readable with explicit parentheses as gamma^{-1}( p^T x_{t0} + sum ... ); as written, the scope of gamma^{-1} is ambiguous.
  4. [Notation, Section 1.2] The definition of min{A,0} as 'the minimum element of A' conflicts with its later use as an elementwise operator in equations such as (3) and (5).
  5. [Section 2.2, equations (10)-(12)] The passage from (9) to (10) should justify why the min over K can be moved inside the row of stacked states; the validity relies on the minimizing K being the same for each column, which holds for the true q-parameter but should be stated explicitly.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the data-driven bounds follow from Bellman-equation algebra plus an explicit small-mismatch perturbation assumption.

full rationale

The paper's derivation chain is self-contained in the relevant sense. The data-driven algebraic equation (12) is constructed directly from the model-free Bellman equation (8)-(10) by multiplying through by data correlation matrices, and it does not encode the stability or robustness conclusion. Lemma 1 is an algebraic equivalence between q-value iteration and p-value iteration, established from the definitions of q, p, and the gain structure. Remark 4 and equation (17) restate the content of (12) in terms of the least-squares estimate [Ahat Bhat] = Sigma_bar Sigma^{-1}; this is an algebraic identity, not a fitted prediction. Theorem 1 is a perturbation result: under the explicit small-model-mismatch assumption (19), the proof compares the true Bellman equation (3) with the estimated one (17) and derives the two-sided bound (20) via the comparison Lemma 2, rather than assuming the bound. The cited results from Ohlin et al. and Li and Rantzer are used as background, as a problem-class characterization, or as a standard comparison fact with stated assumptions that do not include the target theorem; they are not invoked to force the adaptive bounds, so they are not load-bearing circularity. The reviewer-identified issue that alpha_hat = 1 - (rho beta)/(1 - rho beta) is positive only for rho beta < 1/2, while Theorem 1 states rho beta < 1, is a correctness gap in the proof's division step, not a circularity: the claimed conclusion is not equivalent to the inputs by construction. Accordingly, the circularity score is 0.

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

The central claim rests on several domain assumptions inherited from the positive-systems literature, the small-model-mismatch condition (19), and an implicit positivity condition on alpha_hat that the theorem omits. No parameters are fitted to data in the theoretical part; the numerical experiment uses standard hyperparameters.

assumptions (6)
  • domain assumption Assumption 1: (A+BK)x >= 0 for all K in K(E) and reachable x.
    Ensures the closed-loop dynamics remain positive, a standard condition for positive systems.
  • domain assumption Assumption 2: s > \bar{E}^T r.
    Ensures detectability and that the optimal policy is stabilizing.
  • domain assumption The plant belongs to M_beta, i.e., s <= p <= beta*s.
    Defines the class of stabilizable systems; used in the bounds on ||p||_infty.
  • domain assumption Condition (19) holds for all t >= t0.
    This is the small-model-mismatch condition; the theorems are proved only under it.
  • domain assumption The data correlation matrix Sigma(t) is invertible for all t >= t0.
    Needed to define the estimated dynamics [A_hat, B_hat] = \bar{Sigma} Sigma^{-1} and to solve the data-driven equation.
  • ad hoc to paper The positivity of alpha_hat = 1 - alpha_check^{-1}*rho*beta, i.e., rho*beta < 1/2.
    Required in the proof of Theorem 1 when dividing by alpha_hat; not stated in the theorem.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Adaptive Control of Positive Systems with Application to Learning SSP." pith.science (2026). https://pith.science/paper/JKI5JUR3

@misc{pith2026241217012,
  author       = {Pith},
  title        = {Pith review of: Adaptive Control of Positive Systems with Application to Learning SSP},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JKI5JUR3}},
  note         = {Machine review of arXiv:2412.17012}
}
read the original abstract

An adaptive controller is proposed and analyzed for the class of infinite-horizon optimal control problems in positive linear systems presented in (Ohlin et al., 2024b). This controller is derived from the solution of a "data-driven algebraic equation" constructed using the model-free Bellman equation from Q-learning. The equation is driven by data correlation matrices that do not scale with the number of data points, enabling efficient online implementation. Consequently, a sufficient condition guaranteeing stability and robustness to unmodeled dynamics is established. The derived results also provide a quantitative characterization of the interplay between excitation level and robustness to unmodeled dynamics. The class of optimal control problems considered here is equivalent to Stochastic Shortest Path (SSP) problems, allowing for a performance comparison between the proposed adaptive policy and model-free algorithms for learning the stochastic shortest path, as demonstrated in the numerical experiment.

Figures

Figures reproduced from arXiv: 2412.17012 by the authors.

Figure 1
Figure 1. Each plot represents the average over 100 repeated runs, with the shaded area indicating the 95% confidence interval. represents the optimal cumulative cost obtained by applying the optimal policy to the system subject to disturbances w. In the SSP do￾main, an episode terminates when the goal state (i = 4) is reached, whereas in Problem 1, an episode ends when the state vector becomes sufficiently small, indicating … view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

36 extracted references · 24 canonical work pages

  1. [1]

    Model-based and data-based output feedback for external positivity

    Abed AlRahman Al Makdah and Fabio Pasqualetti. Model-based and data-based output feedback for external positivity. In 2024 IEEE 63rd Conference on Decision and Control (CDC), pages 4119--4124, 2024. doi:10.1109/CDC56724.2024.10885918

    arXiv 2024

  2. [2]

    Dynamic programming

    Richard Bellman. Dynamic programming. science, 153 0 (3731): 0 34--37, 1966

    work page 1966

  3. [4]

    Data-driven adaptive dispatching policies for processing networks

    Fethi Bencherki and Anders Rantzer. Data-driven adaptive dispatching policies for processing networks. IEEE Control Systems Letters, 8: 0 2841--2846, 2024. doi:10.1109/LCSYS.2024.3516637

    arXiv 2024

  4. [5]

    Piecewise-linear lyapunov functions for structural stability of biochemical networks

    Franco Blanchini and Giulia Giordano. Piecewise-linear lyapunov functions for structural stability of biochemical networks. Automatica, 50 0 (10): 0 2482--2493, 2014. URL https://www.sciencedirect.com/science/article/pii/S0005109814003288

    work page 2014

  5. [6]

    Adaptive linear quadratic control using policy iteration

    Steven J Bradtke, B Erik Ydstie, and Andrew G Barto. Adaptive linear quadratic control using policy iteration. In Proceedings of 1994 American Control Conference-ACC'94, volume 3, pages 3475--3479. IEEE, 1994. doi:10.1109/ACC.1994.735224

    work page arXiv 1994

  6. [7]

    Modelling methodology for physiology and medicine

    Ewart Carson and Claudio Cobelli. Modelling methodology for physiology and medicine. Newnes, 2013. URL https://doi.org/10.1016/C2012-0-06031-0

    work page doi:10.1016/c2012-0-06031-0 2013

  7. [8]

    Stabilization of positive linear systems

    Patrick De Leenheer and Dirk Aeyels. Stabilization of positive linear systems. Systems & control letters, 44 0 (4): 0 259--271, 2001. URL https://www.sciencedirect.com/science/article/pii/S0167691101001463

    work page 2001

  8. [9]

    Formulas for data-driven control: Stabilization, optimality, and robustness

    Claudio De Persis and Pietro Tesi. Formulas for data-driven control: Stabilization, optimality, and robustness. IEEE Transactions on Automatic Control, 65 0 (3): 0 909--924, 2019. doi:10.1109/TAC.2019.2959924

    arXiv 2019

Show all 36 references
  1. [10]

    Analysis and synthesis of interconnected positive systems

    Yoshio Ebihara, Dimitri Peaucelle, and Denis Arzelier. Analysis and synthesis of interconnected positive systems. IEEE Transactions on Automatic Control, 62 0 (2): 0 652--667, 2016. doi:10.1109/TAC.2016.2558287

    2016

  2. [11]

    Minimax linear optimal control of positive systems

    Alba Gurpegui, Emma Tegling, and Anders Rantzer. Minimax linear optimal control of positive systems. IEEE Control Systems Letters, 2023. doi:10.1109/LCSYS.2023.3341344

    2023

  3. [12]

    Nonnegative and compartmental dynamical systems

    Wassim M Haddad, VijaySekhar Chellaboina, and Qing Hui. Nonnegative and compartmental dynamical systems. Princeton University Press, 2010

    2010

  4. [13]

    Modeling the three stages in hiv infection

    Esteban A Hernandez-Vargas and Richard H Middleton. Modeling the three stages in hiv infection. Journal of theoretical biology, 320: 0 33--40, 2013. URL https://www.sciencedirect.com/science/article/pii/S0022519312006170

    2013

  5. [14]

    Data informativity for distributed positive stabilization

    Takumi Iwata, Shun-ichi Azuma, Ryo Ariizumi, and Toru Asai. Data informativity for distributed positive stabilization. IEEE Control Systems Letters, 2024. doi:10.1109/LCSYS.2024.3404219

    2024

  6. [15]

    Learning-enabled robust control with noisy measurements

    Olle Kjellqvist and Anders Rantzer. Learning-enabled robust control with noisy measurements. In Learning for Dynamics and Control Conference, pages 86--96. PMLR, 2022 a . URL https://proceedings.mlr.press/v168/kjellqvist22a.html

    2022

  7. [16]

    Minimax adaptive estimation for finite sets of linear systems

    Olle Kjellqvist and Anders Rantzer. Minimax adaptive estimation for finite sets of linear systems. In 2022 American Control Conference (ACC), pages 260--265. IEEE, 2022 b . doi:10.23919/ACC53348.2022.9867474

    2022

  8. [17]

    Perturbation analysis of the discrete riccati equation

    Michail M Konstantinov, P Hr Petkov, and Nikolai D Christov. Perturbation analysis of the discrete riccati equation. Kybernetika, 29 0 (1): 0 18--29, 1993

    1993

  9. [18]

    Perturbation theory for matrix equations

    Mihail Konstantinov, D Wei Gu, Volker Mehrmann, and Petko Petkov. Perturbation theory for matrix equations. Gulf Professional Publishing, 2003

    2003

  10. [19]

    Exact dynamic programming for positive systems with linear optimal cost

    Yuchao Li and Anders Rantzer. Exact dynamic programming for positive systems with linear optimal cost. IEEE Transactions on Automatic Control, 2024. doi:10.1109/TAC.2024.3420716

    2024

  11. [20]

    Behavioral systems theory in data-driven analysis, signal processing, and control

    Ivan Markovsky and Florian D \"o rfler. Behavioral systems theory in data-driven analysis, signal processing, and control. Annual Reviews in Control, 52: 0 42--64, 2021. URL https://www.sciencedirect.com/science/article/pii/S1367578821000754

    2021

  12. [22]

    Heuristic search for linear positive systems

    David Ohlin, Anders Rantzer, and Emma Tegling. Heuristic search for linear positive systems. arXiv preprint arXiv:2410.17220, 2024 a

    2024

  13. [23]

    Optimal control of linear cost networks

    David Ohlin, Emma Tegling, and Anders Rantzer. Optimal control of linear cost networks. European Journal of Control, page 101068, 2024 b . URL https://www.sciencedirect.com/science/article/pii/S0947358024001286

    2024

  14. [24]

    Data-driven representations of conical, convex, and affine behaviors

    Alberto Padoan, Florian D \"o rfler, and John Lygeros. Data-driven representations of conical, convex, and affine behaviors. In 2023 62nd IEEE Conference on Decision and Control (CDC), pages 596--601. IEEE, 2023. doi:10.1109/CDC49753.2023.10383687

    2023

  15. [25]

    Controller synthesis for positive linear systems with bounded controls

    M Ait Rami and Fernando Tadeo. Controller synthesis for positive linear systems with bounded controls. IEEE Transactions on Circuits and Systems II: Express Briefs, 54 0 (2): 0 151--155, 2007. doi:10.1109/TCSII.2006.886888

    2007

  16. [26]

    On the kalman-yakubovich-popov lemma for positive systems

    Anders Rantzer. On the kalman-yakubovich-popov lemma for positive systems. IEEE Transactions on Automatic Control, 61 0 (5): 0 1346--1349, 2015. doi:10.1109/TAC.2015.2465571

    2015

  17. [27]

    Minimax adaptive control for a finite set of linear systems

    Anders Rantzer. Minimax adaptive control for a finite set of linear systems. In Learning for Dynamics and Control, pages 893--904. PMLR, 2021. URL https://proceedings.mlr.press/v144/rantzer21a.html

    2021

  18. [28]

    A data-driven R iccati equation

    Anders Rantzer. A data-driven R iccati equation. In Proceedings of the 6th Annual Learning for Dynamics & Control Conference, volume 242 of Proceedings of Machine Learning Research, pages 504--513. PMLR, 15--17 Jul 2024. URL https://proceedings.mlr.press/v242/rantzer24a.html

    2024

  19. [29]

    A tutorial on positive systems and large scale control

    Anders Rantzer and Maria Elena Valcher. A tutorial on positive systems and large scale control. In 2018 IEEE Conference on Decision and Control (CDC), pages 3686--3697. IEEE, 2018. doi:10.1109/CDC.2018.8618689

    2018

  20. [30]

    An online learning analysis of minimax adaptive control

    Venkatraman Renganathan, Andrea Iannelli, and Anders Rantzer. An online learning analysis of minimax adaptive control. In 2023 62nd IEEE Conference on Decision and Control (CDC), pages 1034--1039. IEEE, 2023

    2023

  21. [31]

    Data-driven positive stabilization of linear systems

    Bahram Shafai, Anahita Moradmand, and Milad Siami. Data-driven positive stabilization of linear systems. In 2022 8th International Conference on Control, Decision and Information Technologies, volume 1, pages 1031--1036. IEEE, 2022. doi:10.1109/CoDIT55151.2022.9804005

    2022

  22. [32]

    A positive systems model of tcp-like congestion control: asymptotic results

    Robert Shorten, Fabian Wirth, and Douglas Leith. A positive systems model of tcp-like congestion control: asymptotic results. IEEE/ACM transactions on networking, 14 0 (3): 0 616--629, 2006. doi:10.1109/TNET.2006.876178. URL https://doi.org/10.1109/TNET.2006.876178

    2006

  23. [33]

    Perturbation theory for algebraic riccati equations

    Ji-Guang Sun. Perturbation theory for algebraic riccati equations. SIAM Journal on Matrix Analysis and Applications, 19 0 (1): 0 39--65, 1998. doi:10.1137/S0895479895291303. URL https://doi.org/10.1137/S0895479895291303

    1998 doi

  24. [34]

    Reinforcement learning: An introduction

    Richard S Sutton and Andrew G Barto. Reinforcement learning: An introduction. MIT press, 2018

    2018

  25. [35]

    Statistical learning theory for control: A finite-sample perspective

    Anastasios Tsiamis, Ingvar Ziemann, Nikolai Matni, and George J Pappas. Statistical learning theory for control: A finite-sample perspective. IEEE Control Systems Magazine, 43 0 (6): 0 67--97, 2023. doi:10.1109/MCS.2023.3310345

    2023

  26. [36]

    Data-driven identification and control of positive systems

    Yueyang Wang and Bahram Shafai. Data-driven identification and control of positive systems. In Integrated Systems: Data Driven Engineering, pages 289--307. Springer, 2024

    2024

  27. [37]

    On boundedness of q-learning iterates for stochastic shortest path problems

    Huizhen Yu and Dimitri P Bertsekas. On boundedness of q-learning iterates for stochastic shortest path problems. Mathematics of Operations Research, 38 0 (2): 0 209--227, 2013. URL http://www.jstor.org/stable/24540850

    2013

  28. [38]

    Data-enabled policy optimization for direct adaptive learning of the lqr

    Feiran Zhao, Florian D \"o rfler, Alessandro Chiuso, and Keyou You. Data-enabled policy optimization for direct adaptive learning of the lqr. arXiv preprint arXiv:2401.14871, 2024. URL https://arxiv.org/abs/2401.14871

    2024 arXiv

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.