Improved Sliding Mode Control for a Robotic Manipulator With Input Deadzone and Deferred Constraint
针对存在系统不确定性、输入死区和外部扰动的n连杆机器人操作臂，提出了基于神经网络的滑模控制方案，通过误差转移函数和障碍函数实现有限时间收敛，无需初始状态满足预设约束。
In this article, neural network (NN)-based sliding mode control schemes are proposed for an n-link robotic manipulator with system uncertainties, input deadzone, and external perturbations. A novel error-shifting function is proposed to release initial conditions. NNs are employed to approximate the unknown parameters of both system uncertainties and input deadzone. To update the sliding mode scheme, two advanced sliding mode surfaces with error-shifting function and barrier function are proposed to reduce the dependency of prior information and to realize a finite time convergence result, collectively. It should be pointed out that the proposed methods do not require initial states to satisfy the prescribed constraint caused by the barrier function and can be applied under unknown initial conditions. Furthermore, finite-time convergence for both tracking errors and NN weights is guaranteed. The effectiveness of the proposed schemes is demonstrated by simulation and experiments on the KINOVA robot.