Disturbance Observer-based Neural Network Integral Sliding Mode Control for a Constrained Flexible Joint Robotic Manipulator

In this paper, the tracking control problem of flexible joint robotic manipulator (FJRM) system subjected to system uncertainties and time-varying external disturbances (TVED) is addressed. A new disturbance observer-based neural network integral sliding mode controller with output constraints (DNISMCOC) that comprises the merits of neural networks, disturbance observer and integral sliding mode is proposed. Considering that the radial basis function neural network (RBFNN) has a fast learning convergence speed and great approximation ability, two matrices of RBFNN are utilized to estimate the parameter matrices of the dynamic model of FJRM. In view of the estimation errors of RBFNNs and TVED in the system, a disturbance observer is introduced to estimate the lump uncertainties which consist of them. Integral sliding mode is introduced for eliminating steady errors further. For ensuring security in some high-accuracy using occasions, a barrier lyapunov functions (BLF) is adopted to achieve output constraints of FJRM. To validate the effectiveness of the proposed control scheme, numerical simulations on 2-link FJRM are conducted. According to the comparisons among DNISMCOC and other state-of-the-art controllers, the superiorities of DNISMCOC in several aspects are proved.