EANT+KALMAN: An Efficient Reinforcement Learning Method for Continuous State Partially Observable Domains

Yohannes Kassahun, José de Gea Fernández, Jan Hendrik Metzen, Mark Edgington, Frank Kirchner
In: Andreas Dengel, K. Berns, Thomas Breuel, Frank Bomarius, Thomas Roth-Berghofer (eds.) KI 2008: Advances in Artificial Intelligence volume 5243, Lecture Notes in Artificial Intelligence, Pages 241-248, Kaiserslautern, Germany, Springer, Berlin/ Heidelberg, 2008

Abstract:

In this contribution we present an extension of a neuroevolutionary method called Evolutionary Acquisition of Neural Topologies (EANT) [11] that allows the evolution of solutions taking the form of a POMDP agent (Partially Observable Markov Decision Process) [8]. The solution we propose involves cascading a Kalman filter [10] (state estimator) and a feed-forward neural network. The extension (EANT+KALMAN) has been tested on the double pole balancing without velocity benchmark, achieving significantly better results than the to date published results of other algorithms.

BibTex:

@inproceedings{ KASS2008,
	Title = {EANT+KALMAN: An Efficient Reinforcement Learning Method for Continuous State Partially Observable Domains},
	Author = {Yohannes Kassahun and José de Gea Fernández and Jan Hendrik Metzen and Mark Edgington and Frank Kirchner},
	Editor = {Andreas Dengel, K. Berns, Thomas Breuel, Frank Bomarius, Thomas Roth-Berghofer},
	BookTitle = {KI 2008: Advances in Artificial Intelligence},
	Year = {2008},
	Series = {Lecture Notes in Artificial Intelligence},
	Publisher = {Springer},
	Publisher = {5243},
	Pages = {241-248}
}

     
Last modified:: 30.08.2016