Abstract: The system is decomposed into the input-output subsystem and the zero dynamics subsystem for flexible manipulators with uncertain parameters through the output of manipulators system is redefined. The neural sliding mode control strategy is proposed for the input-output subsystem. The weights of the radial basis function (RBF) neural network is adjusted online to meet sliding mode arrived conditions. It can guarantee the output of the network to approach non-singular terminal sliding mode control input of the system and avoid chattering phenomenon. The zero dynamics subsystem is approximately linearized nearby the equilibrium point to select the design control parameters. Simulation experiment results demonstrate that the method fulfills the accurate point to point control of the flexible manipultor, and the elastic vibration at the end of flexible manipulator is eliminated.

Key words: Flexible manipulator, Neural network, Non-singular terminal sliding mode