Robust controller design via neural network computing

摘要

This paper deals with the problem of robust characteristic polynomial assignment (RCPA) for process systems having parametric uncertainties. A novel computation technique, which is mainly based on the idea of converting the RCPA problem into an equivalent Kennedy and Chua neural network model, is proposed. A measure of uncertain constraint violation is introduced in the solution scheme, which reflects the degree of violation of the uncertain constraints and thus makes the resulting neural model having no uncertain terms and easy to solve. With the Lyapunov-based analysis, the new solution model for the RCPA problem is proven to be stable and without oscillation. To demonstrate the effectiveness and applicability of the presented scheme, two illustrative examples are provided. Extensive simulation results show that the presented approach to the robust controller design is simple, effective and applicable. [References: 18]