Trajectory tracking for tracked mobile robot via zero-bias fuzzy neural network

This paper proposes a zero-bias fuzzy neural network controller for trajectory tracking of tracked mobile robots. The fuzzy rules are constructed in the framework of the neural networks. It allows online learning by using backpropagation. The conventional fuzzy neural network controllers can have nonlinear gain but might have biases, which means that the controllers output a non-zero value even if the error between the reference and the output of the plant is zero. The proposed method is based on a membership function composed of trigonometric functions instead of ordinary sigmoid functions, which allows asymmetric controllers without any bias. The effectiveness of the proposed method is validated through simulations. The simulation results show there is no bias in the proposed controller.