Generalized Dynamic Predictive Control for Nonparametric Uncertain Systems With Application to Series Elastic Actuators

One weakness of the model predictive control method is that the predicted states/outputs are constructed by an exact nominal model. Its accuracy varies if uncertainties exist, which will ultimately deteriorate the closed-loop control performances. To this end, we propose a generalized dynamic predictive control method for a class of lower-triangular systems subjected to nonparametric uncertainties. Instead of relying on the inherent robustness property of the standard predictive controller or on-/off-line parameter identification, a dual-layer adaptive law is designed to estimate the lumped effect of system uncertainties. As another main contribution, under a less ambitious but more practical control objective, namely semi-global stability, various non-linearity growth constraints utilized in the existing related methods could be essentially relaxed. Numerical simulation and illustrative experimental tests of a series elastic actuator system are provided to demonstrate both simplicity and effectiveness of the proposed method.