Efficient computation of optimal actions
Open accessAbstract
Optimal choice of actions is a fundamental problem relevant to fields as diverse as neuroscience, psychology, economics, computer science, and control engineering. Despite this broad relevance the abstract setting is similar: we have an agent choosing actions over time, an uncertain dynamical system whose state is affected by those actions, and a performance criterion that the agent seeks to optimize. Solving problems of this kind remains hard, in part, because of overly generic formulations. Here, we propose a more structured formulation that greatly simplifies the construction of optimal control laws in both discrete and continuous domains. An exhaustive search over actions is avoided and the problem becomes linear. This yields algorithms that outperform Dynamic Programming and Reinforcement Learning, and thereby solve traditional problems more efficiently. Our framework also enables computations that were not possible before: composing optimal control laws by mixing primitives, applying deterministic methods to stochastic systems, quantifying the benefits of error tolerance, and inferring goals from behavioral data via convex optimization. Development of a general class of easily solvable problems tends to accelerate progress--as linear systems theory has done, for example. Our framework may have similar impact in fields where optimal choice of actions is relevant.
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?
Human hand and arm movements are controlled near-optimally.
“Similarly, human hand and arm movements are controlled near-optimally.”
From A quantum microtubule substrate of consciousness is experimentally supported and solves the binding and epiphenomenalism problems (2025) · cites this paper as Todorov 2009
The introductory example (page unavailable) describes a nervous system generating muscle activations to maximize movement performance, but does not say that human hand or arm movements are controlled near-optimally. The Discussion (page unavailable) concerns computational methods for biomechanics rather than evidence about human movement. Searches for near-optimal human hand and arm movement found no passage establishing the claim. (LLM, conf=0.74)
Experimental — accuracy not yet measured
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
Reinforcement Learning: An Introduction
Sutton, R. S., & Barto, A. G. (1998) · IEEE Transactions on Neural Networks · doi:10.1109/tnn.1998.712192 · Cited by 27,225 · see citing works
Reinforcement Learning: An Introduction
Sutton, R. S., & Barto, A. (2005) · IEEE Transactions on Neural Networks · doi:10.1109/tnn.2004.842673 · Cited by 25,263 · see citing works
(1995) · Cited by 10,558 · see citing works
Watkins, C. J. C. H., & Dayan, P. (1992) · Machine Learning · doi:10.1007/bf00992698 · Cited by 9,083 · see citing works
Bryson, A. E., & Ho, Y. (2018) · Routledge eBooks · doi:10.1201/9781315137667 · Cited by 6,031 · see citing works
Optimality principles in sensorimotor control
Todorov, E. (2004) · Nature Neuroscience · doi:10.1038/nn1309 · Cited by 1,746 · see citing works
Self-citation(2000) · doi:10.2460/ajvr.67.2.323 · Cited by 1,621 · see citing works
Optimal Control and Estimation
Stengel, R. F. (1994) · Cited by 1,597 · see citing works
Numerical Methods for Stochastic Control Problems in Continuous Time
Kushner, H. J., & Dupuis, P. G. (2001) · Stochastic modelling and applied probability · doi:10.1007/978-1-4613-0007-6 · Cited by 677 · see citing works
Linearly-solvable Markov decision problems
Todorov, E. (2007) · The MIT Press eBooks · doi:10.7551/mitpress/7503.003.0176 · Cited by 352 · see citing works
Self-citationLearning physics-based motion style with nonlinear inverse optimization
Liu, C. K., Hertzmann, A., & Popović, Z. (2005) · ACM Transactions on Graphics · doi:10.1145/1073204.1073314 · Cited by 341 · see citing works
Schiemenz, B. (1980) · European Journal of Operational Research · doi:10.1016/0377-2217(80)90149-6 · Cited by 313 · see citing works
Linear Theory for Control of Nonlinear Stochastic Systems
Kappen, H. J. (2005) · Physical Review Letters · doi:10.1103/physrevlett.95.200201 · Cited by 307 · see citing works
General duality between optimal control and estimation
Todorov, E. (2008) · 2008 47th IEEE Conference on Decision and Control · doi:10.1109/cdc.2008.4739438 · Cited by 248 · see citing works
Self-citationOptimal Control and Nonlinear Filtering for Nondegenerate Diffusion Processes
Fleming, W. H., & Mitter, S. K. (1982) · Stochastics · doi:10.1080/17442508208833228 · Cited by 110 · see citing works
Optimal Portfolios: Stochastic Models For Optimal Investment And Risk Management In Continuous Time
Korn, R. (1997) · RePEc: Research Papers in Economics · doi:10.1142/3548 · Cited by 93 · see citing works
A Variational Approach to Nonlinear Estimation
Mitter, S. K., & Newton, N. J. (2003) · SIAM Journal on Control and Optimization · doi:10.1137/s0363012901393894 · Cited by 87 · see citing works
Compositionality of optimal control laws
Todorov, E. V. (2009) · Cited by 82 · see citing works
Self-citationEigenfunction approximation methods for linearly-solvable optimal control problems
Todorov, E. (2009) · IEEE Symposium on Adaptive Dynamic Programming and Reinforcement Learning · doi:10.1109/adprl.2009.4927540 · Cited by 44 · see citing works
Self-citation
Flagged first, then by citations
Research works printed but not linked: 3
- [12]Todorov E. Classic Maximum Principles and Estimation-Control Dualities for nonlinear Stochastic Systems. 2009 preprint. [[Google Scholar](https://scholar.google.com/scholar_lookup?journal=Classic%20Maximum%20Principles%20and%20Estimation-Control%20Dualities%20for%20nonlinear%20Stochastic%20Systems&author=E%20Todorov&publication_year=2009&)]
- [14]Todorov E. Efficient Algorithms for Inverse Optimal Control. 2009 preprint. [[Google Scholar](https://scholar.google.com/scholar_lookup?journal=Efficient%20Algorithms%20for%20Inverse%20Optimal%20Control&author=E%20Todorov&publication_year=2009&)]
- [15]Watkins C, Dayan P. Q-learning. Mach Learn. 1992;8:279–292. [[Google Scholar](https://scholar.google.com/scholar_lookup?journal=Mach%20Learn&title=Q-learning&author=C%20Watkins&author=P%20Dayan&volume=8&publication_year=1992&pages=279-292&)]
Cited by
Sorted by citations
The dopaminergic midbrain encodes the expected certainty about desired outcomes
Schwartenbeck, P., FitzGerald, T. H. B., Mathys, C., Dolan, R., & Friston, K. (2015) · Cerebral Cortex · doi:10.1093/cercor/bhu159 · Cited by 219 · see citing works
Petzschner, F. H., Weber, L. A. E., Gard, T., & Stephan, K. E. (2017) · Biological Psychiatry · doi:10.1016/j.biopsych.2017.05.012 · Cited by 181 · see citing works
Toward an integration of deep learning and neuroscience
Marblestone, A. H., Wayne, G., & Kording, K. P. (2016) · Frontiers in Computational Neuroscience · doi:10.3389/fncom.2016.00094 · Cited by 119 · see citing works
Optimal control of gene regulatory networks for morphogen-driven tissue patterning
Pezzotta, A., & Briscoe, J. (2023) · Cell Systems · doi:10.1016/j.cels.2023.10.004 · Cited by 34 · see citing works
Reward maximization through discrete active inference
Da Costa, L., Sajid, N., Parr, T., Friston, K., & Smith, R. (2023) · Neural Computation · doi:10.1162/neco_a_01574 · Cited by 35 · see citing works
Wiest, M. C. (2025) · Neuroscience of Consciousness · doi:10.1093/nc/niaf011 · Cited by 34 · see citing works
“In particular, trial-to-trial fluctuations in real movements tend to vary flexibly along task-irrelevant dimensions, which can change quickly as one adopts new goals (Todorov and Jordan 2002, Todorov 2009).”
Chemotaxis emerges as the optimal solution to cooperative search games
Pezzotta, A., Adorisio, M., & Celani, A. (2018) · Physical Review E · doi:10.1103/physreve.98.042401 · Cited by 17 · see citing works
A dynamic Bayesian model of homeostatic control
Penny, W., & Stephan, K. (2014) · Lecture Notes in Computer Science · doi:10.1007/978-3-319-11298-5_7 · Cited by 10 · see citing works
- Authors
- Todorov, E.
- Affiliations
- University of Washington (US)
- Corresponding author
- Todorov, Emanuel
- Venue
- Proceedings of the National Academy of Sciences
- Publisher
- National Academy of Sciences
- Published
- 2009-07-14
- Volume
- 106
- Issue
- 28
- Pages
- 11478-11483
- ISSN
- 0027-8424 (ISSN-L)
- 1091-6490
- PMC
- PMC2705278
- OpenAlex
- W2093524643
- Cited by
- 417
- References
- 19
- Citations by year
- 2026: 17 · 2025: 30 · 2024: 19 · 2023: 21 · 2022: 17 · 2021: 31 · 2020: 30 · 2019: 21 · 2018: 20 · 2017: 24 · 2016: 30 · 2015: 22 · 2014: 36 · 2013: 26 · 2012: 25
- FWCI
- 25.79
- Type
- journal article
- Language
- en
- Open access
- green
- www.ncbi.nlm.nih.gov
- Locations
- Proceedings of the National Academy of Sciences (publishedVersion)
- PubMed Central (publishedVersion)
- PubMed (publishedVersion)
- Funders
- National Science Foundation
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Reinforcement Learning in Robotics
- Physical Sciences › Computer Science › Computational Theory and Mathematics › Adaptive Dynamic Programming Control
- Physical Sciences › Engineering › Control and Systems Engineering › Advanced Control Systems Optimization
- Keywords
- optimal control, optimal action selection, stochastic systems, error tolerance, convex optimization, dynamic programming comparison, reinforcement learning comparison
- MeSH terms
- Algorithms, Information Science, Models, Theoretical, Problem Solving, Stochastic Processes, Uncertainty
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
