Continuous-time Identification for a class of Switched Linear Systems

Direct identification of continuous-time linear switched state-space models

“…To the best of our knowledge, direct CT identification of switched linear models has been addressed very recently only in (Goudjil et al, 2020;Kersting and Buss, 2019;Du et al, 2021). These approaches, however, rely on strong assumptions imposed on the system's signals.…”

“…In particular, the CT identification method proposed in Goudjil et al (2020) requires that the input signal exciting the system is sinusoidal. Then, by exploiting the linearity of subsystems, outputs of individual subsystems are estimated using a DT switched IO method.…”

“…To position our work w.r.t. these contributions, we do not impose any of the aforementioned assumptions required in (Goudjil et al, 2020;Kersting and Buss, 2019;Du et al, 2021), which are quite restrictive in practice. In particular, in our contribution the input signals used to excite the system are not restricted to sinusoidal inputs.…”

“…All computations are carried out on an i7 1.9-GHz Intel core processor with 32 GB of RAM running MATLAB R2019a. We consider a continuous-time linear switched system governed by (1) having K = 2 modes with subsystem matrices given as follows (Goudjil et al, 2020): The system is excited with a zero-mean Gaussian input signal having unit variance, u(t) ∼ N (0, 1). The dynamics switches between the two subsystems with a Markov switching signal, such that the true mode s t k has π = 10% probability of being different from s t k−1 , starting from s t0 = 1.…”