Kinematic model identification of autonomous mobile robot using dynamical recurrent neural networks

Abstract

The kinematic model identification algorithm of the autonomous mobile robot using dynamical recurrent neural network is presented. The structure and the learning algorithm of dynamical recurrent neural network are analyzed. Data in the training set are from an autonomous mobile robot and are filtered with /spl alpha/-/spl beta/ filter. Experiments on the autonomous mobile robot show that it is practically feasible to represent the nonlinearities and dynamic characters of the autonomous mobile robot. The forward model is also an essential segment to control the autonomous mobile robot motion with generalized predictive control algorithm.