Experimental Comparison of Nonlinear Controllers for Flexible Joint Manipulators

In this paper, the authors present an experimental comparison of several nonlinear controllers for flexible joint manipulators. The experimental setup is a 2-DOF manipulator with an elasticity at the second joint. This work follows and confirms a previous com parative study based on theoretical and numerical investigations. The main conclusions of the study are as follows. (1) Some non linear controllers may behave quite differently when applied to a particular device, despite their apparent theoretical resemblance. (2) The compensation of nonlinearities (Coriolis and centrifugal torques), as well as joint flexibilities, may improve the closed-loop performance for certain desired trajectories, despite a rather high joint stiffness. (3) However, when the joint stiffness is sufficiently large, the controllers that do not take into account the joint flexi bility in the design can, in certain cases, behave as well as those that incorporate it. Thus, future experimental work with a smaller joint stiffness is needed. (4) Finally, the complex structure of cer tain nonlinear controllers for flexible joint robots (backstepping and energy-shaping schemes) may be an obstacle to their closed-loop behavior enhancement. The experimental results clearly show that the inputs of such schemes chatter and have a magnitude larger than those of simpler controllers (like singular perturbation-based or Slotine and Li controllers).