2014
DOI: 10.3182/20140824-6-za-1003.01962
|View full text |Cite
|
Sign up to set email alerts
|

Universal approximators for direct policy search in multi-purpose water reservoir management: A comparative analysis

Abstract: This study presents a novel approach which combines direct policy search and multi-objective evolutionary algorithms to solve high-dimensional state and control space water resources problems involving multiple, conflicting, and non-commensurable objectives. In such a multi-objective context, the use of universal function approximators is generally suggested to provide flexibility to the shape of the control policy. In this paper, we comparatively analyze Artificial Neural Networks (ANN) and Radial Basis Funct… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1
1

Citation Types

0
12
0

Year Published

2015
2015
2020
2020

Publication Types

Select...
6

Relationship

1
5

Authors

Journals

citations
Cited by 18 publications
(12 citation statements)
references
References 22 publications
0
12
0
Order By: Relevance
“…• Basic and Improved Operating Policies : we designed both the BOP and IOP sets by solving Problem 3 with the Evolutionary Multi-Objective Direct Policy Search (EMODPS) method , an approximate dynamic programming approach that combines direct policy search, nonlinear approximating networks, and multi-objective evolutionary algorithms. In particular, we parameterized the operating policy as Gaussian radial basis functions, because they have been demonstrated to be effective in solving this type of multi-objective policy design problems ( Giuliani et al, 2014a;2014b ), particularly when exogenous information is directly used for conditioning the operations ( Giuliani et al, 2015 ). To perform the optimization, we use the self-adaptive Borg Multi-Objective Evolutionary Algorithm (MOEA) ( Hadka and Reed, 2013 ), which has been shown to be highly robust in solving multi-objective optimal control problems, where it met or exceeded the performance of other state-of-the-art MOEAs ( Zatarain-Salazar et al, 2016 ).…”
Section: Experiments Strategy and Settingmentioning
confidence: 99%
“…• Basic and Improved Operating Policies : we designed both the BOP and IOP sets by solving Problem 3 with the Evolutionary Multi-Objective Direct Policy Search (EMODPS) method , an approximate dynamic programming approach that combines direct policy search, nonlinear approximating networks, and multi-objective evolutionary algorithms. In particular, we parameterized the operating policy as Gaussian radial basis functions, because they have been demonstrated to be effective in solving this type of multi-objective policy design problems ( Giuliani et al, 2014a;2014b ), particularly when exogenous information is directly used for conditioning the operations ( Giuliani et al, 2015 ). To perform the optimization, we use the self-adaptive Borg Multi-Objective Evolutionary Algorithm (MOEA) ( Hadka and Reed, 2013 ), which has been shown to be highly robust in solving multi-objective optimal control problems, where it met or exceeded the performance of other state-of-the-art MOEAs ( Zatarain-Salazar et al, 2016 ).…”
Section: Experiments Strategy and Settingmentioning
confidence: 99%
“…Their accuracy strongly depends on the choices of the class of functions used to parameterize the operating policy and on the efficiency of the algorithm used to optimize the policy parameters. In this work, we use Gaussian Radial Basis Functions (RBFs) to parameterize the operating policy as they are capable of representing functions for a large class of problems [e.g., Mhaskar and Micchelli , ; Busoniu et al ., ] and have been demonstrated to be more effective than other universal approximators [ Giuliani et al ., ]. The number of basis of the RBFs is set equal to N = M + 1, where M is the dimension of the policy input vector scriptIt=(t,boldxt,boldIt).…”
Section: The Hoa Binh Reservoir Case Studymentioning
confidence: 99%
“…As demonstrated in Figure , the inherent robustness of an optimal solution found by the RBF model in this case allows for a measure of uncertainty in the knowledge of the future climate exposure when determining which management alternative should be implemented. This supports the evidence that the policies identified using the RBF model represent reservoir decisions well [ Giuliani et al ., , ], but for other water resource systems, alternative models may be more appropriate. Where possible, the model chosen should not overcalibrate to a particular hydrometeorological state, and provide robustness across neighboring states in the same manner as the RBFs, to reduce the reliance on exact knowledge of the climate state should a decision be implemented.…”
Section: Discussionmentioning
confidence: 99%