Prof. Kok Lay Teo, gave a lecture titled “Switching Time and System Parameter Optimization in Nonlinear Switched Time-Delay Systems” at AMSS on 5, April 2018.
This talk discussed the problem of estimating unknown time-delays (both state- and input-delays) and unknown system parameters in a general nonlinear time-delay system. They proposed a unified computational approach that involved solving a dynamic optimization problem, whose cost function measures the discrepancy between predicted and observed system output, to determine optimal values for the unknown quantities. They discussed two computational procedures for computing the partial derivatives of the cost function in this problem: one involved solving a set of auxiliary time-delay systems forward in time; the other involved solving an auxiliary "costate" system backwards in time. By combining either of these computational procedures with a standard gradient-based optimization algorithm, the optimal values for the unknown delays and system parameters could be obtained. Furthermore, this approach would be applied to the situation when the observed data was noisy. Finally, some real practical examples would be solved using the methods proposed.
Prof. Kok Lay Teo is from Curtin University, Australia.