A Methodology for Real-Time Implementation of Nonlinear Observers via Convex Optimization

This paper is concerned with the design and real-time implementation of nonlinear observers via convex optimization techniques. The well-known problem of unmeasurable premises is overcome by employing only known signals for the observer while the difficulty of obtaining the error system is solved by algebraic rearrangements and Taylor series. The resulting error dynamics are exactly rewritten as a tensor-product polytope which enables solving for the observer gains via linear matrix inequalities. Both simulation and real-time examples are provided.