Experiment Designs for Control-Oriented MIMO Identification

In this paper, guidelines for designing inputs for control-oriented identification of multivariable systems are given. The information needed for design of inputs such as double rectangular pulses, pseudo-random sequences (PRBS), and multi-sinusoidal signals can be obtained from a simple step test. Such a step test also gives accurate information about the directionality properties of the system. It is shown how this information can be used to design input signals as well as directional inputs that explicitly excite the various gain directions of the system. By such excitations, small gains and fast dynamics can be better identified than by standard, non-directional, methods. If the identified model is to be used for control design, it is important that such properties are included in the model. The design techniques are illustrated by simulations on an ill-conditioned, 3-input, 3-output system. In addition to output noise, low-frequency input disturbances are added to make the identification task realistic with regard to real-world applications.