Experimental issues of finite time compensation of dynamic friction for robots

Compensation of dynamic friction draws the attention in the high precision research community since the seminal paper of Panteley, which employs the model of Canudas. Later, the Canudas model was further studied to uncover some undesirable dynamical properties of the nature of dynamic friction, which this model captures very well, and those properties have known some time ago by the tribology research community. However, it is precisely the ability to reproduce hard nonlinearities and presliding regime at very small displacements that makes this very simple model a viable choice for engineers dealing with friction. Altogether, this control scheme allows to compensate asymptotically hard nonlinearities due to friction at several velocity regimes. However, the asymptotic behavior of this scheme may be unsuitable for fast precision tasks, wherein bidirectional motion is involved. Thus, ideally, finite time convergence instead of asymptotic convergence is a better control choice. In this paper, the experimental verification of an adaptive second order sliding mode controller for robust compensation of dynamic friction in finite time is presented. All system parameters are considered unknown. The closed-loop system exhibits a well-posed terminal attractors with chattering-free controller, and experiments on a planar robot actuated by direct-drive motors are presented. Further implication of this technique are analyzed and discussed.