Optimal input design for identification of systems with quantized measurements

This paper addresses system identification of FIR models with quantized measurements in a worst-case setting. It is assumed that measurements are collected through a multi-threshold sensor and that the system output is corrupted by unknown but bounded noise. The main contribution of the paper consists in the solution of the optimal input design problem for identification of a scalar gain. This result allows one to design a suboptimal input for a FIR model of arbitrary order. Moreover, for a selected configuration of the sensor thresholds, an upper bound on the time complexity of the identification problem is derived.