Concurrent learning based finite time parameter estimation in adaptive control of uncertain switched systems

In this paper, We propose concurrent learning adaptive controller, which uses recorded and current data concurrently for adaptation, to model reference adaptive control (MRAC) of uncertain switched systems. In standard MRAC architecture for switched systems, the adaptive update laws are derived based on the gradient descent scheme, but here we developed two novel parameter estimation schemes by using modification terms in adaptation laws in which recorded data is used simultaneously with current data and a triggering time is considered in which a sufficient condition on the linear independence of the recorded data is obtained to guarantee the exponential convergence of tracking error and parameter estimation error to zero for the uncertain switched system under all admissible switching strategy. The convergence of the parameters to the ideal values makes an on-line learned model of the system available. This sufficient condition is easily verifiable in comparison to the restrictive persistence of excitation (PE) condition of the standard MRAC structures in practical applications. Finally a simulation example is given to illustrate the efficacy of the proposed method.