Three-dimensional vehicle attitude estimation using modified invasive weed optimized particle filter

Abstract

In this paper, a novel attitude estimation approach for the vehicle using a modified invasive weed optimized particle filter is studied, and it can be applied to provide the vehicle angle accurately. A modified invasive weed optimization (MIWO), being added a control envelope in the standard deviation of the invasive weed optimization (IWO) to improve the convergence speed and avoid the conventional IWO plunging in the local optimal solution, is introduced in the sampling process of the particle filter. The proposed particle filter makes particles reproduced dynamically by their fitness in a nearby space, and optimizes the particle population of optimal weights to let particles move towards the regions where they have the great values of the posterior probability density. Within the framework of the proposed particle filter, the modified Rodrigues parameters (MRPs) associated with the angular velocity of a gyro are included in a state vector, and the outputs of the accelerometer and magnetometer are regarded as the observation vector which is used for correcting the errors of the angular velocity measurement. The simulation results show that the suggested particle filter can enhance the accuracy of the state estimation compared with other improved particle filters. The experimental results prove that the attitude errors using the presented method are significantly reduced by comparing to that obtained from the extended Kalman filter and the classic particle filter.