| Summary: | 碩士 === 國立中正大學 === 電機工程研究所 === 105 === Extended Kalman Filter (EKF) is well known as a popular solution to the Simultaneous Localization and Mapping (SLAM) problem for mobile robot platforms or
vehicles. Since the performance of EKF significantly depends on the priori knowledge
of process noise and measurement noise covariance matrix Q and R respectively, the
improper setting of these parameters can cause significant performance degradation.
To overcome this weakness of EKF, this thesis presents a development of Neuro-Fuzzy
based adaptive EKF for the SLAM problem with the aim of estimating the proper
values for the elements of R matrix at each running step. In other words, the adaptive neuro fuzzy EKF (ANFEKF) is designed to reduce the mismatch between the
theoretical and actual covariance of the innovation consequence. Then the particle
swarm optimization is employed to train the free parameter of ANFEKF offline. By
employing Particle Swarm Optimization (PSO) we can exploit the advantages of the
high-dimensional search space algorithm for the more effective of training phrase of
ANFEKF. The performance of the proposed approach is evaluated by doing experiment on the mobile robot platform under two benchmarks of environment situation
with varying number of landmarks. Additionally, the real implementation on ARIA
mobile platform is also proposed to evaluate our ANFEKF ability for the SLAM
problem. The results have shown that the improvement of the proposed ANFEKF
method in comparison with conventional EKF method in term of computational cost,
performance efficiency and real time implementation.
|
|---|
