In the classical approach to optimal filtering, it is assumed that the stochastic model of the physical process is fully known. With uncertain models, the natural solution is to optimize over both the original objective and the model uncertainty, thereby arriving at optimal robust operators, the topic of this book.