DocumentCode :
2383479
Title :
Reinforcement field
Author :
Chiu, Po-Hsiang ; Huber, Manfred
Author_Institution :
Dept. of Comput. Sci. & Eng., Univ. of Texas at Arlington, Arlington, TX, USA
fYear :
2011
fDate :
9-12 Oct. 2011
Firstpage :
2567
Lastpage :
2574
Abstract :
Complex control tasks involving varying or evolving system dynamics often pose a great challenge to mainstream reinforcement learning algorithms. Specifically, in most standard methods, actions are often assumed to be a concrete and fixed set that applies to the state space in a predefined manner. Consequently, without resorting to a substantial re-learning procedure, the derived policy lacks the ability to adapt to variations in action outcomes or shifts in the action set. In addition, the standard action representation and its attendant state transition mechanism limit the applicability of the RL framework in complex domains primarily due to the intractability of the resulting large state space and lack of the facility to generalize the learned policy to the unknown parts of the state space. This paper proposes an alternative view of reinforcement learning by establishing the notion of the reinforcement field through a collection of policy-embedded particles gathered during the policy learning process. The reinforcement field serves as a policy generalization mechanism through the use of kernel functions as a state correlation hypothesis in combination with Gaussian process regression as a value function approximator.
Keywords :
Gaussian processes; adaptive control; function approximation; generalisation (artificial intelligence); learning (artificial intelligence); learning systems; regression analysis; state-space methods; Gaussian process regression; action outcome; action representation; action set shift; attendant state transition mechanism; control task; evolving system dynamics; kernel function; policy generalization; policy learning; policy-embedded particles; reinforcement field; reinforcement learning algorithm; relearning procedure; state correlation hypothesis; state space; value function approximator; variation adaptation; varying system dynamics; Aerospace electronics; Correlation; Gaussian processes; Ground penetrating radar; Kernel; Mathematical model; Vectors; Gaussian process regression; kernel methods; parametric actions; reinforcement learning;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Systems, Man, and Cybernetics (SMC), 2011 IEEE International Conference on
Conference_Location :
Anchorage, AK
ISSN :
1062-922X
Print_ISBN :
978-1-4577-0652-3
Type :
conf
DOI :
10.1109/ICSMC.2011.6084063
Filename :
6084063
Link To Document :
بازگشت