Adaptive learning approach of integrating evolution fuzzy-neural networks and Q-learning for mobile robots

In the paper, an adaptive learning approach of integrating evolution fuzzy-neural networks and Q-learning is developed so that a mobile robot can adapt itself to a real and complex environment. Specifically, based on Q-value and an evolution method that adjusts their parameter values of the fuzzy-neural networks, the mobile robot evolves better strategies to adapt to the environment. However, in most studies of evolution learning, the learning of mobile robots often requires a simulator and an enormous amount of evolution time so as to perform a task. Therefore, we are to integrate Q-learning into the evolution fuzzy-neural networks to avoid the requirement of the simulator. Experiment results of a mobile robot illustrate the performance of the proposed approach.