Abstract
Many control applications in real-world processes require accurate models for the active system. In particular, hybrid systems which are defined as an interaction of continuous dynamics, usually described by differential equations, and discrete dynamics, described through switching sequences. Note that the sub-models of a hybrid system are activated alternatively by a switching rule which indicates the active submodel at each time instant. Nowadays, the estimation of both the time-interval in which a sub-model is active and the parameters of such sub-model is an important issue. In fact, it allows suitable choice of the operating modes in a real process. Hence, the hybrid identification problem is a challenging task due to the inherent nonconvexity of the prediction-error function according to the parameters to be identified. In this paper, the Particle Swarm Optimization (PSO) technique is exploited to locate the switching instants of Autonomous Switched Linear Systems (ASLS) and to estimate the parameters of the submodels only by using measurements from the real process. Then, statistical validations are proposed to show the efficiency of the framework through a literature benchmark.