Active Disturbance Rejection Control and Embedded Model Control: A case study comparison

Abstract

The paper aims to compare Active Disturbance Rejection Control (ADRC) and Embedded Model Control (EMC). Both control algorithms are designed and implemented around an internal model which includes a disturbance dynamics, capable of encoding external disturbances and model discrepancies. Both schemes appear to be an extension of the internal model control (IMC). The disturbance dynamics is driven by the so-called uncertainty input signals (briefly noise) that are real-time estimated by an observer feedback (uncertainty estimator) driven by the model error: plant output minus model output. The main advantage of ADRC and EMC is that the estimated disturbance contains the past uncertainty that now enters both the internal model and the control law as a signal, and allows the rest of the control law (state feedback) to be designed model-based, not depending on the uncertainty. The main difference between ADRC and EMC is that the former assumes that model errors can be treated like input disturbances, whereas EMC shows that high-frequency neglected dynamics cannot be treated as such. The former standpoint does not place any limitation to the control bandwidth (BW), unlike the latter one which is compelled to find out an optimal BW in the presence of uncertainty. The different design has been enhanced by treating the same case study with two different internal models. The simpler model adopted by EMC is affected by the neglected dynamics. The different models do not impede performance comparison with the aid of suitable scale factors.