Robust model-based control: an experimental case study

Three forms of robust model-based control were experimentally evaluated. Algorithm evaluation was prompted by a requirement for controllers with good high-speed tracking accuracy in uncertain payload environments. The test case was a PUMA-560 operating over the standard test suite. The tracking performance of the robust algorithms was compared, with and without payload, to that of a nonadaptive model-based controller with fixed proportional-derivative (PD) feedback. The model-based controllers were made robust by the addition of an auxiliary input term, replacing the PD feedback with a feedback loop based on quantitative feedback theory (QFT), or an adaptive feedforward compensator based on Lyapunov theory. Experimental evaluation provided valuable insight into the potential and limitations of each method. All three techniques improved the tracking performance of the manipulator. Superior overall performance, computational simplicity, and a mathematically rigorous design and tuning procedure make the model-based controller with QFT feedback the algorithm of choice