Neural Network-Based Terminal Sliding Mode Controller Design for Manipulator Systems

In the context of nonlinear and uncertain manipulator systems, achieving fast and accurate desired trajectories for each joint of the manipulator is crucial. Our study aims to improve current control methods by combining sliding mode control and neural network techniques. Specifically, we leverage the neural networks’ approximation capabilities to estimate both the modeling error and external disturbance in our research. These estimates are then integrated into the synovial control as additional gains. This approach allows for the sliding mode controller’s switch gain to adapt to changing modeling error and external disturbance, thereby avoiding the vibration production typically associated with traditional sliding mode controllers. It can be observed from the simulation results that the combination of the proposed nonsingular terminal sliding mode controller and neural network achieves fast and precise tracking of desired trajectories.