DocumentCode
2256367
Title
Choosing the cost vector of the linear programming approach to approximate dynamic programming
Author
De Farias, Daniela Pucci ; Weber, Théophane
Author_Institution
Dept. of Mech. Eng., MIT, Cambridge, MA, USA
fYear
2008
fDate
9-11 Dec. 2008
Firstpage
67
Lastpage
72
Abstract
We consider the linear programming approach to approximate dynamic programming. In the general case of linear combination of features, we prove the existence of a solution which can be used to generate a policy with performance proportional to the strength of the architecture. In the special case of features defined on a partition of the state space, we give a simpler method to find the solution, as well as a stronger performance bound.
Keywords
approximation theory; dynamic programming; linear programming; cost vector; dynamic programming approximation; linear programming approach; state space; Cost function; Dynamic programming; Function approximation; Linear approximation; Linear programming; Mechanical engineering; Operations research; Space stations; State-space methods; Vectors;
fLanguage
English
Publisher
ieee
Conference_Titel
Decision and Control, 2008. CDC 2008. 47th IEEE Conference on
Conference_Location
Cancun
ISSN
0191-2216
Print_ISBN
978-1-4244-3123-6
Electronic_ISBN
0191-2216
Type
conf
DOI
10.1109/CDC.2008.4739452
Filename
4739452
Link To Document