New results on the inverse-system/deconvolution problem for linear dynamical systems (Q1581814)

This paper addresses the problem to construct an inverse system \(S^{-1}\) if a finite-dimensional, linearized system \(S\) is given. The solution is derived assuming uncertain (vector) input signals \(s\) having the so-called waveform structure \[ s(t)=C_1f_1(t)+\dots+C_mf_m(t); \] (\(m\) finite; \(C_i\) stepwise constant unknown coefficients with jumps occurring sparsely in time). The algorithm is derived based on the state space representation of linearized systems which is adapted to the special type of input signals having the waveform structure given above. It is presented for continuous time and discrete time systems as well and will be useful especially in cases where the pole/zero structure of linear systems is difficult to handle by conventional methods. An example is given in detail.