Designing an Adaptive Fuzzy Hybrid Unscented Particle Filterented Particle Filter

Abstract

With the rapid developments in computer technology, the particle filter is becoming more desirable in navigation applications as it is able to handle

 nonlinear systems and non-Gaussian measurement noise. However, its computational cost still limits its widespread use. Unlike the Unscented Kalman filter, although it is computationally inexpensive and gives a high accuracy of displaying the system's state, it imposes Gaussian restrictions on measurement noise. It is also sensitive to any sudden change in the system dynamics. One way to reduce the computational cost of PF without any degradation of the system estimation accuracy is to combine the particle filter with other filters. In this paper, a new algorithm of an adaptive fuzzy hybrid filter between a developed particle filter and an adaptive UKF is proposed, in which the advantages of both algorithms are taken, and obtaining a robust filter against non-Gaussian, time-varying noise, and being adaptive for dynamic nonlinear systems with the lowest computational cost. The results showed that the proposed algorithm succeeded in adapting to the dynamics of the nonlinear system and dealing with the non-Gaussian noise of the measurements by providing high accuracy and robustness while estimating the state of the system.