Finite-Data Performance Guarantees for the Output-Feedback Control of an Unknown System
Open accessAbstract
As the systems we control become more complex, first-principle modeling becomes either impossible or intractable, motivating the use of machine learning techniques for the control of systems with continuous action spaces. As impressive as the empirical successes of these methods have been, strong theoretical guarantees of performance, safety, or robustness are few and far between. This paper takes a step towards providing such guarantees by establishing finite-data performance guarantees for the robust output-feedback control of an unknown FIR SISO system. In particular, we introduce the “Coarse-ID control” pipeline, which is composed of a system identification step followed by a robust controller synthesis procedure, and analyze its end-to-end performance, providing quantitative bounds on the performance degradation suffered due to model uncertainty as a function of the number of experiments used to identify the system. We conclude with numerical examples demonstrating the effectiveness of our method.
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
There has been recent progress in RL for control, which could be useful for fine-tuning geoengineering interventions such as deciding where and when to release aerosols.
“Additionally, there has been recent progress in RL for control [21,84,545], which could be useful for fine-tuning geoengineering interventions such as deciding where and when to release aerosols.”
From Tackling climate change with machine learning (2023) · cites this paper as [84]
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Cites
Boyd, S., & Vandenberghe, L. (2004) · doi:10.1017/cbo9780511804441 · Cited by 56,601 · see citing works
Lennart Ljung, A. (1998) · Wiley Encyclopedia of Electrical and Electronics Engineering · doi:10.1002/047134608x.w1046 · Cited by 4,118 · see citing works
Global Convergence of Policy Gradient Methods for the Linear Quadratic RegulatorPreprint
Fazel, M., Ge, R., Kakade, S. M., & Mesbahi, M. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1801.05039 · Cited by 387 · see citing works
Positive Trigonometric Polynomials and Signal Processing Applications
Dumitrescu, B. (2017) · Signals and communication technology · doi:10.1007/978-3-319-53688-0 · Cited by 216 · see citing works
Linear system identification via atomic norm regularization
Shah, P., Bhaskar, B. N., Tang, G., & Recht, B. (2012) · IEEE Conference on Decision and Control (CDC) · doi:10.1109/cdc.2012.6426006 · Cited by 101 · see citing works
Self-citationFinite sample properties of system identification methods
Campi, M. C., & Weyer, E. (2002) · IEEE Transactions on Automatic Control · doi:10.1109/tac.2002.800750 · Cited by 97 · see citing works
On the Sample Complexity of the Linear Quadratic RegulatorPreprint
Dean, S., Mania, H., Matni, N., et al. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1710.01688 · Cited by 94 · see citing works
Self-citationGoal-driven dynamics learning via Bayesian optimization
Bansal, S., Calandra, R., Xiao, T. G., et al. (2017) · Proceedings of the IEEE Conference on Decision & Control, including the Symposium on Adaptive Processes · doi:10.1109/cdc.2017.8264425 · Cited by 74 · see citing works
Scalable system level synthesis for virtually localizable systems
Matni, N., Wang, Y.-S., & Anderson, J. (2017) · Proceedings of the IEEE Conference on Decision & Control, including the Symposium on Adaptive Processes · doi:10.1109/cdc.2017.8264168 · Cited by 52 · see citing works
Self-citationNon-Asymptotic Analysis of Robust Control from Coarse-Grained IdentificationPreprint
Tu, S., Boczar, R., Packard, A., et al. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1707.04791 · Cited by 51 · see citing works
Self-citationNonparametric estimation of transfer functions: rates of convergence and adaptation
Goldenshluger, A. (1998) · IEEE Transactions on Information Theory · doi:10.1109/18.661510 · Cited by 32 · see citing works
Goal-Driven Dynamics Learning via Bayesian OptimizationPreprint
Bansal, S., Calandra, R., Xiao, T. G., et al. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1703.09260 · Cited by 31 · see citing works
On the design of LQR kernels for efficient controller learning
Marco, A., Hennig, P., Stefan Schaal, A., & Trimpe, S. (2017) · IEEE Conference on Decision and Control (CDC) · doi:10.1109/cdc.2017.8264429 · Cited by 26 · see citing works
A System Level Approach to Controller SynthesisPreprint
Wang, Y.-S., Matni, N., & Doyle, J. C. (2016) · arXiv (Cornell University) · doi:10.48550/arxiv.1610.04815 · Cited by 19 · see citing works
Self-citation
Flagged first, then by citations
Research works printed but not linked: 3
- Y. Duan, X. Chen, R. Houthooft, J. Schulman, and P. Abbeel, "Benchmarking deep reinforcement learning for continuous control," in International Conference on Machine Learning, 2016, pp. 1329-1338.
- S. Levine, C. Finn, T. Darrell, and P. Abbeel, "End-to-end training of deep visuomotor policies," JMLR, vol. 17, no. 1, pp. 1334-1373, 2016.
- S. Boucheron, G. Lugosi, and P. Massart, Concentration Inequalities: A Nonasymptotic Theory of Inde- pendence. Oxford University Press, 2013.
Cited by
Sorted by citations
Tackling climate change with machine learning
Rolnick, D., Donti, P. L., Kaack, L. H., Kochanski, K., Lacoste, A., Sankaran, K., Ross, A. S., Milojevic-Dupont, N., Jaques, N., Waldman-Brown, A., et al. (2023) · ACM Computing Surveys · doi:10.1145/3485128 · Cited by 336 · see citing works
- Authors
- Boczar, R., Matni, N., & Recht, B.
- Affiliations
- University of California, Berkeley (US)
- Venue
- 2018 IEEE Conference on Decision and Control (CDC)
- Publisher
- IEEE
- Published
- 2018-12
- Pages
- 2994-2999
- ISSN
- 0191-2216 (ISSN-L)
- OpenAlex
- W2794593261
- Cited by
- 33
- References
- 14
- Citations by year
- 2025: 2 · 2023: 3 · 2022: 3 · 2021: 9 · 2020: 6 · 2019: 8 · 2018: 2
- FWCI
- 9.38
- Type
- conference paper
- Language
- en
- Open access
- green
- arxiv.org
- Locations
- IEEE Conference on Decision and Control (CDC) (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Topics
- Physical Sciences › Engineering › Control and Systems Engineering › Control Systems and Identification
- Physical Sciences › Engineering › Control and Systems Engineering › Advanced Control Systems Optimization
- Physical Sciences › Engineering › Control and Systems Engineering › Iterative Learning Control Systems
- Keywords
- output feedback control, unknown systems, robust control, system identification, model uncertainty, data-driven control, continuous action spaces
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
