In this study, we assess the robustness of four distinct control approaches: pole placement; the command generator tracker (CGT) approach to model reference control; model reference adaptive control (MRAC); and MRAC using a fixed feedback gain. We use a second order, single-input single-output (SISO) plant to examine the performance and stability of each method. This evaluation spans a broad range of design goals and uncertainty in models of the plant.

Pole placement and CGT designs are linear and relatively easy to implement, but require explicit knowledge of the plant. Although MRAC schemes require little knowledge of the plant's dynamic characteristics, such algorithms are non-linear and involve design variables whose effects are not readily apparent. Currently, there are no general design procedures for MRAC. In this study, we propose a method for designing an MRAC controller applied to a second order SIS a plant. This method does not require the controller to be tuned for different closed-loop performance goals. This procedure also creates a consistent basis for comparing the robustness of all four algorithms.

Pole placement and the CGT control perform as designed if the plant is modeled correctly. Under this circumstance, the adaptive controllers also perform at levels equivalent to the linear algorithms. However, conditions with plant modeling error highlight enormous differences among the four algorithms. Pole placement suffers the largest response error and for extreme testing conditions, instability. The CGT controller exhibits better performance than pole placement and remains stable over all testing variables. NIRAC maintains a high performance level under severe testing conditions. MRAC requires minimal plant knowledge to guarantee stability and good performance.