Q-Learning Based on Dynamical Structure Neural Network for Robot Navigation in Unknown Environment

Authors:
Junfei Qiao;Ruiyuan Fan;Honggui Han;Xiaogang Ruan
Affiliations:
Institute of Intelligence System, College of Electronic Information and control Engineering, Beijing University of Technology, Beijing, China 100000;Institute of Intelligence System, College of Electronic Information and control Engineering, Beijing University of Technology, Beijing, China 100000;Institute of Intelligence System, College of Electronic Information and control Engineering, Beijing University of Technology, Beijing, China 100000;Institute of Intelligence System, College of Electronic Information and control Engineering, Beijing University of Technology, Beijing, China 100000
Venue:
ISNN 2009 Proceedings of the 6th International Symposium on Neural Networks: Advances in Neural Networks - Part III
Year:
2009

Citing 5
Cited 0

Application of SONQL for real-time learning of robot behaviors

Robotics and Autonomous Systems
A Comprehensive Survey of Multiagent Reinforcement Learning

IEEE Transactions on Systems, Man, and Cybernetics, Part C: Applications and Reviews
Cognitive navigation based on nonuniform Gabor space sampling, unsupervised growing networks, and reinforcement learning

IEEE Transactions on Neural Networks
Global Reinforcement Learning in Neural Networks

IEEE Transactions on Neural Networks
Integrating Temporal Difference Methods and Self-Organizing Neural Networks for Reinforcement Learning With Delayed Evaluative Feedback

IEEE Transactions on Neural Networks

Quantified Score

Hi-index	0.00

Visualization

Abstract

An automation learning and navigation strategy based on dynamical structure neural network and reinforcement learning was proposed in this paper. The neural network can adjust its structure according to the complexity of the working environment. New nodes or even new hidden-layers can be inserted or deleted during the training process. In such a way, the mapping relations between environment states and responding action were established, and the dimension explosion problem was solved at the same time. Simulation and Pioneer3-DX mobile robot navigation experiments were done to test the proposed algorithm. Results show that the robot can learn the correct action and finish the navigation task without people's guidance, and the performance was better than artificial potential field method.