Title :
Markov decision Processes with fractional costs
Author :
Ren, Zhiyuan ; Krogh, Bruce H.
Author_Institution :
Signal Electron. & Embedded Syst. Lab., Gen. Electr. Global Res. Center, Niskayuna, NY, USA
fDate :
5/1/2005 12:00:00 AM
Abstract :
Certain methods for constructing embedded Markov decision processes (MDPs) lead to performance measures that are the ratio of two long-run averages. For such MDPs with finite state and action spaces and under an ergodicity assumption, this note presents algorithms for computing optimal policies based on policy iterations, linear programming, value iterations and Q-learning.
Keywords :
Markov processes; iterative methods; linear programming; Q-learning; embedded Markov decision processes; ergodicity assumption; fractional costs; linear programming; optimal policy computation; policy iterations; value iterations; Cost function; Embedded system; Linear programming; Probability; Q factor; State-space methods; Fractional costs; Markov decision processes;
Journal_Title :
Automatic Control, IEEE Transactions on
DOI :
10.1109/TAC.2005.846520
