Cooperation Learning for Behaviour-based Neural-fuzzy Controller in Robot Navigation
| Authors: | Li Jianing, Institute of Automation, Chinese Academy of Sciences, China
Yi Jianqiang, Institute of Automation, Chinese Academy of Sciences, China
Zhao Dongbin, Institute of Automation, Chinese Academy of Sciences, China
Xi Guangcheng, Institute of Automation, Chinese Academy of Sciences, China |
|---|
| Topic: | 4.3 Robotics |
|---|
| Session: | Mobile Robots II |
|---|
| Keywords: | Mobile robots; Behaviour; Neural network; Fuzzy control; Learning algorithms; Sensors |
|---|
Abstract
Based on the previously proposed extended neural-fuzzy network, this paper presents a cooperation scheme of training data based hybrid learning and reinforcement learning for constructing sensor-based behaviour modules in robot navigation. In order to solve reinforcement learning problem, a reinforcement-based neural-fuzzy control system (RNFCS) is provided, which consists of a neural-fuzzy controller (NFC) and a neural-fuzzy predictor (NFP). By estimating the “desired output”, reinforcement learning is treated and realized from the point of view of training data based learning. Computer simulations are conducted to illustrate the effectiveness of our method.
