Nonlinear filters for linear signal models

The paper explores the benefits of using nonlinear filters with linear signal models. In particular, we show how switching and known bounds on states or disturbances can be used to advantage in filter design. The resulting nonlinear filters incorporate either switching between fast and slow modes or projection to constrain filter states. Both mechanisms can be used in the same filter. It is shown that these modifications retain filter stability properties whilst enhancing transient performance.