Particle filtering mode estimation of hybrid systems based on a novel weighted mode activation record approach

This paper presents a novel weighted mode activation record approach for mode estimation in hybrid systems. Discrete mode estimation is crucial to achievement of state estimation in hybrid systems. While, mode transition probability plays a decisive role in system mode estimation. Empirical setting of this probability and relying the mode decision on the current system information are major drawbacks of the currently used estimation strategies. The proposed weighted mode activation record is incorporated in a particle filtering algorithm to introduce the past dynamic knowledge of the system at the edge of vision in the estimation procedure. This provides a more comprehensive view, leading to a more wise and dynamic decision about the system operational mode. A series of simulation experiments is conducted to comparatively demonstrate the capability of the proposed approach to enhance the mode estimation accuracy of a particle filtering algorithm.