On the infinite horizon active fault detection problem for linear Gaussian multiple models

The paper deals with the design of an active fault detector that consists of a given detector and an input signal generator whose output is fed back into the system to improve the quality of detection. The problem is considered for a system described by linear Gaussian multiple models and a discounted criterion on the infinite horizon is used to evaluate the quality of the active fault detector. The problem is solved by reformulating the original problem as a perfect state information problem and then the approximate closed loop input signal generator is designed off-line by applying an approximate dynamic programming technique.