Model-based sliding functions design for sliding mode robot control

Abstract

This paper introduces a novel manifold design for sliding mode control, applicable to second-order mechanical systems in which nonlinear dynamics can be formalised into that of robotic manipulators. The new approach shows that model-based sliding manifold design substantially simplifies the torque control law, which ultimately becomes linear in terms of joint angles and rates. Additionally, this approach allows the decoupling of the chattering effect on the torque inputs on each axis. A new property related to the gravity term is introduced and is used for stability analysis and model validation. Simulation results compare the introduced approach to the conventional linear manifold design and demonstrate that the new approach reduces transient constraints on torque input and is more robust to matched uncertainties for low inertia robots.