Robust filtering under nonrandom disturbances: The invariant ellipsoid approach

We present a simple and universal observer-based approach to solving the problem of robust filtering of unknown-but-bounded exogenous disturbances. The heart of this approach is the method of invariant ellipsoids. Application of this technique allows for a reformulation of the original problem in terms of linear matrix inequalities with reduction to semidefinite programming and one-dimensional optimization, which are easy to solve numerically. Continuous-time and discrete-time cases are studied in equal detail. The efficacy of the approach is demonstrated via the double pendulum example.