On the Computational Complexity of Stochastic Controller Optimization in POMDPs

被引：40

作者：

Vlassis, Nikos ^{[1
]}

Littman, Michael L. ^{[2
]}

Barber, David ^{[3
]}

机构：

[1] Univ Luxembourg, Luxembourg Ctr Syst Biomed, 7 Ave Hauts Fourneaux, L-4362 Esch Belval, Luxembourg

[2] Brown Univ, Dept Comp Sci, Providence, RI 02912 USA

[3] UCL, Dept Comp Sci, London WC1E 6BT, England

来源：

ACM TRANSACTIONS ON COMPUTATION THEORY | 2012年 / 4卷 / 04期

基金：

美国国家科学基金会;

关键词：

Performance; Partially observable Markov decision process; stochastic controller; bilinear program; computational complexity; Motzkin-Straus theorem; sum-of-square-roots problem; matrix fractional program; computations on polynomials; nonlinear optimization;

D O I：

10.1145/2382559.2382563

中图分类号：

TP301 [理论、方法];

学科分类号：

081202 ;

摘要：

We show that the problem of finding an optimal stochastic blind controller in a Markov decision process is an NP-hard problem. The corresponding decision problem is NP-hard in PSPACE and SQRT-SUM-hard, hence placing it in NP would imply breakthroughs in long-standing open problems in computer science. Our result establishes that the more general problem of stochastic controller optimization in POMDPs is also NP-hard. Nonetheless, we outline a special case that is convex and admits efficient global solutions.

引用

页数：8

共 50 条

[1] Reinforcement learning for POMDPs based on action values and stochastic optimization
Perkins, TJ
[J]. EIGHTEENTH NATIONAL CONFERENCE ON ARTIFICIAL INTELLIGENCE (AAAI-02)/FOURTEENTH INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE (IAAI-02), PROCEEDINGS, 2002, : 199 - 204
[2] Computational complexity in robust controller synthesis
Hara, S
Yamada, Y
[J]. LEARNING, CONTROL AND HYBRID SYSTEMS: FESTSCHRIFT IN HONOR OF BRUCE ALLEN FRANCIS AND MATHUKUMALLI VIDYASAGAR ON THE OCCASION OF THEIR 50TH BIRTHDAYS, 1999, 241 : 59 - 80
[3] Computational complexity of stochastic programming problems
Dyer, M
Stougie, L
[J]. MATHEMATICAL PROGRAMMING, 2006, 106 (03) : 423 - 432
[4] Stochastic analog networks and computational complexity
Siegelmann, HT
[J]. JOURNAL OF COMPLEXITY, 1999, 15 (04) : 451 - 475
[5] Computational complexity of stochastic programming problems
Martin Dyer
Leen Stougie
[J]. Mathematical Programming, 2006, 106 : 423 - 432
[6] Optimizing fixed-size stochastic controllers for POMDPs and decentralized POMDPs
Amato, Christopher
Bernstein, Daniel S.
Zilberstein, Shlomo
[J]. AUTONOMOUS AGENTS AND MULTI-AGENT SYSTEMS, 2010, 21 (03) : 293 - 320
[7] TEMPERATURE CONTROLLER OPTIMIZATION BY COMPUTATIONAL INTELLIGENCE
Cojbasic, Zarko M.
Ristanovic, Milan R.
Markovic, Nemanja R.
Tesanovic, Stefan Z.
[J]. THERMAL SCIENCE, 2016, 20 : S1541 - S1552
[8] A comment on "computational complexity of stochastic programming problems"
Hanasusanto, Grani A.
Kuhn, Daniel
Wiesemann, Wolfram
[J]. MATHEMATICAL PROGRAMMING, 2016, 159 (1-2) : 557 - 569
[9] ON THE COMPUTATIONAL COST AND COMPLEXITY OF STOCHASTIC INVERSE SOLVERS
Faliszewski, Piotr
Smolka, Maciej
Schaefer, Robert
Paszynski, Maciej
[J]. COMPUTER SCIENCE-AGH, 2016, 17 (02): : 225 - 264
[10] A comment on “computational complexity of stochastic programming problems”
Grani A. Hanasusanto
Daniel Kuhn
Wolfram Wiesemann
[J]. Mathematical Programming, 2016, 159 : 557 - 569

← 1 2 3 4 5 →