Comparison of estimation performance between single and multiple disturbance observers for LTI systems

This paper compares estimation performance of two optimal disturbance observers which estimate more than one unknown and unmeasurable disturbances acting on a linear time-invariant system. Both observers are designed in O´Reilly´s framework, but they are constructed in a slightly different manner. The one is constructed so that maximum estimation errors are minimized in the sense of Pareto optimality. The other is constructed by combining the single-output optimal observers which minimize maximum estimation errors for each disturbance. Through a numerical example, estimation performances and computational costs of the two observers are compared.