Pith. sign in

REVIEW 2 major objections 2 minor 92 references

In LQR control with unknown dynamics and online mode revelation, an algorithm achieves expected regret O(|M|^{1/4} n_s^{3/4} + n_m) while keeping state norms bounded via (α,β)-controllability.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.3

2026-06-26 11:19 UTC pith:HNTQLMLQ

load-bearing objection Paper gives a concrete regret bound for safe multi-mode LQR switching with unknown dynamics but the infinite-visit assumption for the benchmark needs explicit handling for finite n_s. the 2 major comments →

arxiv 2606.22223 v2 pith:HNTQLMLQ submitted 2026-06-20 eess.SY cs.SY

Regret-Guaranteed Safe Switching: LQR Setting with Unknown Dynamics

Jafar Abbaszadeh Chekan , S. Rasoul Etesami , Cedric Langbort This is my paper
classification eess.SY cs.SY
keywords switched systemsLQR controlregret minimizationunknown dynamicsdwell timesafe switchingmode switchingcontrollability
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper addresses learning-based switching control for linear quadratic regulator systems where each mode's parameters are unknown in advance and the next mode appears only at the instant of switching. It first solves the known-model case by applying per-mode Riccati gains together with dwell times that meet a controllability condition, under the assumption that every mode is visited infinitely often. For the unknown case it introduces an online algorithm that estimates dwell-time errors to keep the difference from the known-model benchmark cost low. A reader would care because the result supplies a concrete way to maintain stability and near-optimal cost without requiring a priori models in switched environments.

Core claim

Under the (α,β)-controllability requirement, the benchmark policy for known models uses discrete algebraic Riccati equation gains in each mode and selects dwell times that satisfy the controllability condition; the online algorithm for unknown models then estimates dwell-time errors to produce an expected regret of O(|M|^{1/4} n_s^{3/4} + n_m) relative to that benchmark.

What carries the argument

The (α,β)-controllability condition that enforces bounded state norms through sufficient dwell times, together with the dwell-time error estimation step inside the regret-minimizing switching rule.

Load-bearing premise

Each mode must be visited infinitely often.

What would settle it

A sequence of mode switches in which either the state norm violates the prescribed (α,β) bound or the observed cumulative cost difference grows faster than the stated order.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • The known-model benchmark applies Riccati gains and dwell times satisfying controllability whenever modes recur infinitely often.
  • The online policy keeps state norms bounded while incurring only sublinear regret in the number of switches.
  • Malignant switches contribute an additive linear term to total regret.
  • The regret expression separates the contribution of the number of modes from the number of switches.

Where Pith is reading between the lines

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

  • The bound implies that regret remains sublinear even when the number of modes grows, provided the number of switches grows slower than n_s to the fourth power.
  • The same dwell-time estimation idea could be tested on performance criteria other than quadratic cost.
  • When the infinite-visit assumption does not hold, the algorithm may need an explicit exploration schedule to restore the regret guarantee.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 2 minor

Summary. The paper studies regret minimization for safe switching control in multi-mode LQR systems with a priori unknown dynamics. Modes are revealed only at switching instants. It introduces (α,β)-controllability to enforce state-norm bounds, derives a benchmark policy for the known-model case (DARE gains plus dwell times satisfying the controllability condition, under the assumption that each mode is visited infinitely often), and proposes an online algorithm for the unknown-model case whose expected regret against this benchmark is O(|M|^{1/4} n_s^{3/4} + n_m) by estimating dwell-time errors.

Significance. If the regret bound is rigorously established, the work supplies a concrete performance guarantee for adaptive switching control that simultaneously enforces safety constraints and competes with an offline benchmark. The explicit dependence on the number of switches and modes, together with the handling of malignant switches, would be a useful addition to the literature on regret bounds for switched linear systems.

major comments (2)
  1. [known-model benchmark definition and regret analysis] The benchmark policy is shown to be optimal only under the explicit assumption that each mode is visited infinitely often (known-model analysis). The regret bound is stated for a finite number of switches n_s; the unknown-model analysis must show how the infinite-visit property is either relaxed or carried forward (e.g., via limiting arguments) so that the finite-horizon comparison remains valid. Without this step the claimed O(|M|^{1/4} n_s^{3/4} + n_m) bound rests on an unverified transfer of the benchmark optimality condition.
  2. [unknown-model algorithm and regret theorem] The abstract states the regret bound and the key assumption but supplies no derivation steps, error-bar analysis, or proof sketch for how dwell-time error estimation produces the specific exponents 1/4 and 3/4. The central claim therefore rests on unverified steps whose correctness cannot be assessed from the given text; the manuscript must supply the missing intermediate inequalities or concentration arguments that justify the rate.
minor comments (2)
  1. [abstract and notation section] Notation for the set of modes is introduced as |M| in the abstract but should be consistently defined with a calligraphic M or similar in the main text.
  2. [abstract] The term 'malignant switches' is used in the regret expression without an explicit definition or reference to its first appearance.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the thoughtful and detailed report. The comments highlight important points on the transfer of the infinite-visit optimality condition to the finite-switch regret bound and on the need for explicit derivation steps. We address each major comment below and indicate where revisions will be made.

read point-by-point responses

  1. Referee: [known-model benchmark definition and regret analysis] The benchmark policy is shown to be optimal only under the explicit assumption that each mode is visited infinitely often (known-model analysis). The regret bound is stated for a finite number of switches n_s; the unknown-model analysis must show how the infinite-visit property is either relaxed or carried forward (e.g., via limiting arguments) so that the finite-horizon comparison remains valid. Without this step the claimed O(|M|^{1/4} n_s^{3/4} + n_m) bound rests on an unverified transfer of the benchmark optimality condition.

    Authors: The referee correctly identifies that benchmark optimality is proven under the infinite-visit assumption. In the manuscript the regret is defined directly against this benchmark policy, and the finite-n_s bound is obtained by controlling the per-mode cost deviation via the (α,β)-controllability condition and the dwell-time estimator; the infinite-visit property is used only to establish the benchmark itself, not to require infinite visits in the realized trajectory. To make the finite-horizon comparison fully rigorous we will add an explicit lemma that bounds the difference between the finite-visit cost and the infinite-visit benchmark cost, using a limiting argument on the tail of the dwell-time sequence. This will be inserted in the known-model section and referenced in the regret theorem statement. revision: yes

  2. Referee: [unknown-model algorithm and regret theorem] The abstract states the regret bound and the key assumption but supplies no derivation steps, error-bar analysis, or proof sketch for how dwell-time error estimation produces the specific exponents 1/4 and 3/4. The central claim therefore rests on unverified steps whose correctness cannot be assessed from the given text; the manuscript must supply the missing intermediate inequalities or concentration arguments that justify the rate.

    Authors: The full derivation, including the concentration inequality on dwell-time estimation error and the subsequent application of Hölder’s inequality that yields the |M|^{1/4} n_s^{3/4} term, appears in the appendix. We agree that a concise proof sketch is missing from the main body. We will insert a one-paragraph outline immediately after the algorithm description that highlights the key steps: (i) uniform bound on dwell-time estimation error via Hoeffding, (ii) propagation of this error into the per-switch cost deviation, and (iii) summation over n_s switches with the resulting 3/4 exponent. The abstract itself cannot accommodate the sketch, but the introduction will be expanded accordingly. revision: yes

Circularity Check

0 steps flagged

Benchmark policy defined independently via DARE; regret against external benchmark with no self-reduction

full rationale

The known-model benchmark is constructed from the discrete algebraic Riccati equation for gains and explicit (α,β)-controllability dwell-time conditions, stated under the infinite-visit assumption. Regret is defined as the cost difference to this offline benchmark, which does not depend on the online algorithm's fitted quantities or equations. The O(|M|^{1/4} n_s^{3/4} + n_m) bound is obtained by dwell-time error estimation in the unknown-model case. No quoted step shows a prediction reducing to a fitted input by construction, a self-citation chain, or an ansatz smuggled via prior work. The infinite-visit assumption is carried explicitly from the benchmark definition and does not create definitional circularity, though its finite-horizon applicability is a separate validity concern outside circularity analysis. The derivation remains self-contained.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the domain assumption that every mode is visited infinitely often and on the newly introduced (α,β)-controllability definition; no free parameters or invented entities are visible in the abstract.

axioms (1)
  • domain assumption each mode is visited infinitely often
    Explicitly stated in the abstract as a requirement for the benchmark policy and implicitly used for the regret analysis.

pith-pipeline@v0.9.1-grok · 5823 in / 1239 out tokens · 23232 ms · 2026-06-26T11:19:15.306867+00:00 · methodology

0 comments
read the original abstract

We consider learning-based control in LQR setting, where the parameters associated with each mode are a priori unknown. The next mode to be activated is revealed online only at the time of switching. The objective is to determine both the switching times and the control gains for each mode such that (1) the norm of the system state remains bounded according to a prescribed criterion, and (2) the accumulated cost is minimized. To formalize the state-norm requirement, we introduce the notion of $(\alpha,\beta)$-controllability for given parameters $\alpha$ and $\beta$. We first study the problem in a known model setting and show that, under the switching mechanism described above and under the assumption that each mode is visited infinitely often, the strategy that minimizes the average expected cost consists of applying, in each mode, the feedback gain obtained from the solution of the discrete algebraic Riccati equation, while selecting dwell times that sufficiently satisfy the controllability condition. We refer to this strategy as the benchmark policy. Next, we propose an algorithm for the unknown-model setting that minimizes the regret, defined as the difference between the cumulative cost incurred by the online algorithm and that of the offline benchmark. By accurately estimating dwell-time errors, our method achieves an expected regret of $\mathcal{O}(|\mathcal{M}|^{1/4} n_s^{3/4} + n_m)$, where $n_s$ denotes the number of switches, $|\mathcal{M}|$ is the number of modes, and $n_m$ is the number of malignant switches.

discussion (0)

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

Reference graph

Works this paper leans on

92 extracted references · 13 canonical work pages · 8 internal anchors

  1. [1]

    Online Least Squares Estimation with Self-Normalized Processes: An Application to Bandit Problems

    Online least squares estimation with self-normalized processes: An application to bandit problems , author=. arXiv preprint arXiv:1102.2670 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  2. [2]

    Proceedings of the 24th Annual Conference on Learning Theory , pages=

    Regret bounds for the adaptive control of linear quadratic systems , author=. Proceedings of the 24th Annual Conference on Learning Theory , pages=

  3. [3]

    Design Automation Conference , pages=

    Cyber-physical systems: the next computing revolution , author=. Design Automation Conference , pages=. 2010 , organization=

    2010

  4. [4]

    2008 The 28th International Conference on Distributed Computing Systems Workshops , pages=

    Secure control: Towards survivable cyber-physical systems , author=. 2008 The 28th International Conference on Distributed Computing Systems Workshops , pages=. 2008 , organization=

    2008

  5. [5]

    SIAM Journal on Control and Optimization , volume=

    Adaptive linear quadratic Gaussian control: the cost-biased approach revisited , author=. SIAM Journal on Control and Optimization , volume=. 1998 , publisher=

    1998

  6. [6]

    2016 International Joint Conference on Neural Networks (IJCNN) , pages=

    Game-theoretic tracking control for actuator attack attenuation in cyber-physical systems , author=. 2016 International Joint Conference on Neural Networks (IJCNN) , pages=. 2016 , organization=

    2016

  7. [7]

    IEEE Control Systems Magazine , volume=

    Game theory-based control system algorithms with real-time reinforcement learning: How to solve multiplayer games online , author=. IEEE Control Systems Magazine , volume=. 2017 , publisher=

    2017

  8. [8]

    IEEE Transactions on Automatic Control , volume=

    Cooperative Q-learning for rejection of persistent adversarial inputs in networked linear quadratic systems , author=. IEEE Transactions on Automatic Control , volume=. 2017 , publisher=

    2017

  9. [9]

    IEEE Access , volume=

    Data-Driven Integral Reinforcement Learning for Continuous-Time Non-Zero-Sum Games , author=. IEEE Access , volume=. 2019 , publisher=

    2019

  10. [10]

    IEEE transactions on cybernetics , year=

    Resilient and Robust Synchronization of Multiagent Systems Under Attacks on Sensors and Actuators , author=. IEEE transactions on cybernetics , year=

  11. [11]

    Online learning for linearly parametrized control problems , author=

  12. [12]

    Poisoning Attacks against Support Vector Machines

    Poisoning attacks against support vector machines , author=. arXiv preprint arXiv:1206.6389 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  13. [13]

    Advances in neural information processing systems , pages=

    Certified defenses for data poisoning attacks , author=. Advances in neural information processing systems , pages=

  14. [14]

    Preprints of the 1st workshop on Secure Control Systems , pages=

    False data injection attacks in control systems , author=. Preprints of the 1st workshop on Secure Control Systems , pages=

  15. [15]

    Learning Linear-Quadratic Regulators Efficiently with only $\sqrt{T}$ Regret

    Learning Linear-Quadratic Regulators Efficiently with only sqrt(T) Regret , author=. arXiv preprint arXiv:1902.06223 , year=

    work page internal anchor Pith review Pith/arXiv arXiv 1902

  16. [16]

    2014 , institution=

    Architecture-based self-adaptation for moving target defense , author=. 2014 , institution=

    2014

  17. [17]

    2017 IEEE Power & Energy Society Innovative Smart Grid Technologies Conference (ISGT) , pages=

    Moving target defense for securing smart grid communications: Architecture, implementation & evaluation , author=. 2017 IEEE Power & Energy Society Innovative Smart Grid Technologies Conference (ISGT) , pages=. 2017 , organization=

    2017

  18. [18]

    Proceedings of the First ACM Workshop on Moving Target Defense , pages=

    Comparing different moving target defense techniques , author=. Proceedings of the First ACM Workshop on Moving Target Defense , pages=

  19. [19]

    2015 54th IEEE Conference on Decision and Control (CDC) , pages=

    Detecting integrity attacks on control systems using a moving target approach , author=. 2015 54th IEEE Conference on Decision and Control (CDC) , pages=. 2015 , organization=

    2015

  20. [20]

    Model-Free Linear Quadratic Control via Reduction to Expert Prediction

    Model-free linear quadratic control via reduction to expert prediction , author=. arXiv preprint arXiv:1804.06021 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  21. [21]

    2016 54th Annual Allerton Conference on Communication, Control, and Computing (Allerton) , pages=

    A moving target approach for identifying malicious sensors in control systems , author=. 2016 54th Annual Allerton Conference on Communication, Control, and Computing (Allerton) , pages=. 2016 , organization=

    2016

  22. [22]

    International conference on decision and game theory for security , pages=

    Game-theoretic approach to feedback-driven multi-stage moving target defense , author=. International conference on decision and game theory for security , pages=. 2013 , organization=

    2013

  23. [23]

    IEEE Communications Surveys & Tutorials , year=

    Toward Proactive, Adaptive Defense: A Survey on Moving Target Defense , author=. IEEE Communications Surveys & Tutorials , year=

  24. [24]

    IEEE Transactions on Automatic Control , year=

    A Moving Target Defense Control Framework for Cyber-Physical Systems , author=. IEEE Transactions on Automatic Control , year=

  25. [25]

    Proceedings of 1994 American Control Conference-ACC'94 , volume=

    Adaptive linear quadratic control using policy iteration , author=. Proceedings of 1994 American Control Conference-ACC'94 , volume=. 1994 , organization=

    1994

  26. [26]

    Least-Squares Temporal Difference Learning for the Linear Quadratic Regulator

    Least-squares temporal difference learning for the linear quadratic regulator , author=. arXiv preprint arXiv:1712.08642 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  27. [27]

    International conference on machine learning , pages=

    Global convergence of policy gradient methods for the linear quadratic regulator , author=. International conference on machine learning , pages=. 2018 , organization=

    2018

  28. [28]

    Towards provable control for unknown linear dynamical systems , author=

  29. [29]

    Advances in Neural Information Processing Systems , pages=

    Efficient reinforcement learning for high dimensional linear quadratic systems , author=. Advances in Neural Information Processing Systems , pages=

  30. [30]

    Learning-based Control of Unknown Linear Systems with Thompson Sampling

    Learning-based control of unknown linear systems with thompson sampling , author=. arXiv preprint arXiv:1709.04047 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  31. [31]

    International Conference on Machine Learning , pages=

    Improved regret bounds for thompson sampling in linear quadratic control problems , author=. International Conference on Machine Learning , pages=

  32. [32]

    Bayesian Optimal Control of Smoothly Parameterized Systems: The Lazy Posterior Sampling Algorithm

    Bayesian optimal control of smoothly parameterized systems: The lazy posterior sampling algorithm , author=. arXiv preprint arXiv:1406.3926 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  33. [33]

    Foundations of Computational Mathematics , pages=

    On the sample complexity of the linear quadratic regulator , author=. Foundations of Computational Mathematics , pages=. 2017 , publisher=

    2017

  34. [34]

    Proceedings of the tenth annual conference on Computational learning theory , pages=

    PAC adaptive control of linear systems , author=. Proceedings of the tenth annual conference on Computational learning theory , pages=

  35. [35]

    Optimism-Based Adaptive Regulation of Linear-Quadratic Systems

    Optimism-based adaptive regulation of linear-quadratic systems , author=. arXiv preprint arXiv:1711.07230 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  36. [36]

    arXiv preprint arXiv:2004.00241 , year=

    Regret Bounds for LQ Adaptive Control Under Database Attacks (Extended Version) , author=. arXiv preprint arXiv:2004.00241 , year=

    work page arXiv 2004

  37. [37]

    IEEE control systems magazine , volume=

    Basic problems in stability and design of switched systems , author=. IEEE control systems magazine , volume=. 1999 , publisher=

    1999

  38. [38]

    Automatica , volume=

    Adaptive fault-tolerant control for actuator failures: a switching strategy , author=. Automatica , volume=. 2017 , publisher=

    2017

  39. [39]

    Asia-Pacific Journal of Chemical Engineering , volume=

    Adaptive fault-tolerant control based on hybrid redundancy , author=. Asia-Pacific Journal of Chemical Engineering , volume=. 2012 , publisher=

    2012

  40. [40]

    Automatica , volume=

    Adaptive actuator fault tolerant control for uncertain nonlinear systems with multiple actuators , author=. Automatica , volume=. 2015 , publisher=

    2015

  41. [41]

    arXiv preprint arXiv:2007.12291 , year=

    Explore More and Improve Regret in Linear Quadratic Regulators , author=. arXiv preprint arXiv:2007.12291 , year=

    work page arXiv 2007

  42. [42]

    2007 46th IEEE Conference on Decision and Control , pages=

    Optimal controller initialization for switching between stabilizing controllers , author=. 2007 46th IEEE Conference on Decision and Control , pages=. 2007 , organization=

    2007

  43. [43]

    IEEE transactions on automatic control , volume=

    Adaptive control using multiple models , author=. IEEE transactions on automatic control , volume=. 1997 , publisher=

    1997

  44. [44]

    IEEE Transactions on Automatic Control , volume=

    Model and controller selection policies based on output prediction errors , author=. IEEE Transactions on Automatic Control , volume=. 1996 , publisher=

    1996

  45. [45]

    IEEE Transactions on Control Systems Technology , volume=

    Dynamic transfer among alternative controllers and its relation to antiwindup controller design , author=. IEEE Transactions on Control Systems Technology , volume=. 1996 , publisher=

    1996

  46. [46]

    Automatica , volume=

    Linear quadratic bumpless transfer , author=. Automatica , volume=. 2000 , publisher=

    2000

  47. [47]

    2016 American control conference (ACC) , pages=

    Fault tolerant control for over-actuated systems: An adaptive correction approach , author=. 2016 American control conference (ACC) , pages=. 2016 , organization=

    2016

  48. [48]

    Microprocessors and Microsystems , volume=

    A gateway for time-triggered control networks , author=. Microprocessors and Microsystems , volume=. 2007 , publisher=

    2007

  49. [49]

    Advances in Neural Information Processing Systems , volume=

    Certainty equivalence is efficient for linear quadratic control , author=. Advances in Neural Information Processing Systems , volume=

  50. [50]

    Mathematics of Control, Signals, and Systems , volume=

    On time-triggered and event-based control of integrator systems over a shared communication system , author=. Mathematics of Control, Signals, and Systems , volume=. 2013 , publisher=

    2013

  51. [51]

    2006 , publisher=

    Diagnosis and fault-tolerant control , author=. 2006 , publisher=

    2006

  52. [52]

    IFAC Proceedings Volumes , volume=

    Fault-tolerant control: the 1997 situation , author=. IFAC Proceedings Volumes , volume=. 1997 , publisher=

    1997

  53. [53]

    International Conference on Machine Learning , pages=

    Logarithmic regret for learning linear quadratic regulators efficiently , author=. International Conference on Machine Learning , pages=. 2020 , organization=

    2020

  54. [54]

    arXiv preprint arXiv:2105.14709 , year=

    Joint Stabilization and Regret Minimization through Switching in Systems with Actuator Redundancy , author=. arXiv preprint arXiv:2105.14709 , year=

    work page arXiv

  55. [55]

    2018 IEEE Conference on Decision and Control (CDC) , pages=

    A moments based approach to designing MIMO data driven controllers for switched systems , author=. 2018 IEEE Conference on Decision and Control (CDC) , pages=. 2018 , organization=

    2018

  56. [56]

    Automatica , volume=

    Data driven stability analysis of black-box switched linear systems , author=. Automatica , volume=. 2019 , publisher=

    2019

  57. [57]

    Automatica , volume=

    Online learning of data-driven controllers for unknown switched linear systems , author=. Automatica , volume=. 2022 , publisher=

    2022

  58. [58]

    Automatica , volume=

    A convex optimization approach to synthesizing state feedback data-driven controllers for switched linear systems , author=. Automatica , volume=. 2022 , publisher=

    2022

  59. [59]

    International Journal of Robust and Nonlinear Control , volume=

    Supervisory switching-based prescribed performance control of unknown nonlinear systems against actuator failures , author=. International Journal of Robust and Nonlinear Control , volume=. 2020 , publisher=

    2020

  60. [60]

    International Journal of Systems Science , volume=

    A general fault tolerant linear quadratic control strategy under actuator outages , author=. International Journal of Systems Science , volume=. 2010 , publisher=

    2010

  61. [61]

    Proceedings of the 2001 American Control Conference.(Cat

    Stabilizing a linear system by switching control with dwell time , author=. Proceedings of the 2001 American Control Conference.(Cat. No. 01CH37148) , volume=. 2001 , organization=

    2001

  62. [62]

    IEEE Transactions on Automatic Control , volume=

    A moving target defense for securing cyber-physical systems , author=. IEEE Transactions on Automatic Control , volume=. 2020 , publisher=

    2020

  63. [63]

    IEEE Transactions on Automatic Control , volume=

    Learn and Control While Switching: Guaranteed Stability and Sublinear Regret , author=. IEEE Transactions on Automatic Control , volume=. 2024 , publisher=

    2024

  64. [64]

    IEEE Transactions on automatic control , volume=

    Nonlinear regulation: The piecewise linear approach , author=. IEEE Transactions on automatic control , volume=. 1981 , publisher=

    1981

  65. [65]

    Discrete Event Dynamic Systems , volume=

    Optimal control of hybrid switched systems: A brief survey , author=. Discrete Event Dynamic Systems , volume=. 2015 , publisher=

    2015

  66. [66]

    On stability, L2-gain and H

    Zhao, Jun and Hill, David J , journal=. On stability, L2-gain and H. 2008 , publisher=

    2008

  67. [67]

    IEEE Transactions on Automatic control , volume=

    Stability and stabilizability of switched linear systems: a survey of recent results , author=. IEEE Transactions on Automatic control , volume=. 2009 , publisher=

    2009

  68. [68]

    Systems & Control Letters , volume=

    Stability of switched systems: a Lie-algebraic condition , author=. Systems & Control Letters , volume=. 1999 , publisher=

    1999

  69. [69]

    Proceedings of the 38th IEEE conference on decision and control (Cat

    Stability of switched systems with average dwell-time , author=. Proceedings of the 38th IEEE conference on decision and control (Cat. No. 99CH36304) , volume=. 1999 , organization=

    1999

  70. [70]

    International Conference on Artificial Intelligence and Statistics , pages=

    Reinforcement learning with fast stabilization in linear dynamical systems , author=. International Conference on Artificial Intelligence and Statistics , pages=. 2022 , organization=

    2022

  71. [71]

    Advances in neural information processing systems , volume=

    Improved algorithms for linear stochastic bandits , author=. Advances in neural information processing systems , volume=

  72. [72]

    International conference on machine learning , pages=

    Thompson sampling for contextual bandits with linear payoffs , author=. International conference on machine learning , pages=. 2013 , organization=

    2013

  73. [73]

    Advances in Neural Information Processing Systems , volume=

    Optimal order simple regret for Gaussian process bandits , author=. Advances in Neural Information Processing Systems , volume=

  74. [74]

    IFAC-PapersOnLine , volume=

    Joint Stabilization and Regret Minimization through Switching in Over-Actuated Systems , author=. IFAC-PapersOnLine , volume=. 2022 , publisher=

    2022

  75. [75]

    International Conference on Machine Learning , pages=

    Online linear quadratic control , author=. International Conference on Machine Learning , pages=. 2018 , organization=

    2018

  76. [76]

    Proceedings of Machine Learning Research vol XX , volume=

    Online switching control with stability and regret guarantees , author=. Proceedings of Machine Learning Research vol XX , volume=

  77. [77]

    2022 American Control Conference (ACC) , pages=

    Data-driven control of markov jump systems: Sample complexity and regret bounds , author=. 2022 American Control Conference (ACC) , pages=. 2022 , organization=

    2022

  78. [78]

    Automatica , volume=

    Regret bounds for online-learning-based linear quadratic control under database attacks , author=. Automatica , volume=. 2023 , publisher=

    2023

  79. [79]

    arXiv preprint arXiv:2303.11489 , year=

    Data-driven mode detection and stabilization of unknown switched linear systems , author=. arXiv preprint arXiv:2303.11489 , year=

    work page arXiv

  80. [80]

    Learning for Dynamics and Control , pages=

    Regret bounds for adaptive nonlinear control , author=. Learning for Dynamics and Control , pages=. 2021 , organization=

    2021

Showing first 80 references.