Journal article
Decoupling non-sequential stochastic control problems
Systems & control letters, Vol.16(1), pp.65-69
1991
DOI: 10.1016/0167-6911(91)90031-9
Abstract
A stochastic control problem is said to be
sequential when its control actions can be ordered in advance such that no action depends on itself or subsequent actions. When no such order exists it is termed
non-sequential. At first glance, a problem's non-sequentiality would seem to preclude its recursive decomposition. This paper considers the possibility of reducing a class of unconstrained non-sequential problems, modeled within the framework of Witsenhausen's intrinsic model, to equivalent sequential problems that are (in theory) amenable to recursive decomposition. It is shown that a sequential reduction exists whenever a causality property and a mild measurability condition are satisfied. This result implies that a large class of
N-action non-sequential problems are no harder to solve than the hardest (
N + 1)-action sequential problems.
Details
- Title: Subtitle
- Decoupling non-sequential stochastic control problems
- Creators
- Mark S Andersland - Department of Electrical and Computer Engineering, The University of Iowa, Iowa City, IA 52242-1595, USA
- Resource Type
- Journal article
- Publication Details
- Systems & control letters, Vol.16(1), pp.65-69
- Publisher
- Elsevier B.V
- DOI
- 10.1016/0167-6911(91)90031-9
- ISSN
- 0167-6911
- eISSN
- 1872-7956
- Language
- English
- Date published
- 1991
- Academic Unit
- Electrical and Computer Engineering
- Record Identifier
- 9984083220902771
Metrics
16 Record Views