Vague Neural Network Based Reinforcement Learning Control System for Inverted Pendulum
Reinforcement learning is a class of model-free learning control method that can solve Markov decision problems. But it has some problems in applications, especially in MDPs of continuous state spaces. In this paper, based on the vague neural networks, we propose a Q-learning algorithm which is comprehensively considering the reward and punishment of the environment. Simulation results in cart-pole balancing problem illustrate the effectiveness of the proposed method.
KeywordsReinforcement Learning Inverted Pendulum Fuzzy Neural Network Continuous State Space Optimal Control Policy
Unable to display preview. Download preview PDF.
- 1.Horikawa, S., Furuhashi, T., Uchikawa, Y.: On fuzzy modeling using fuzzy neural network with the back-propagation algorithm. IEEE Trans. NN-3(5), 801–806 (1992)Google Scholar
- 4.Chen, S.M.: Fuzzy system reliability analysis based on vague set theory. In: 1997 IEEE International Conference on Computational Cybernetics and Simulation, vol. 2, pp. 1650–1655 (1997)Google Scholar
- 5.Zheng, Y., Luo, S., Lv, Z.: The negative effect on the control of inverted pendulum caused by the limit cycle in reinforcement learning. In: ICNN&B 2005, pp. 772–775 (2005)Google Scholar
- 6.Onat, A.: Q-learning with recurrent Neural Networks as a Controller for the Inverted Pendulum Problem. In: The Fifth International Conference on Neural Information Processing, October 21-23, pp. 837–840 (1998)Google Scholar