| Pub. Date | : March, 2020 |
|---|---|
| Product Name | : The IUP Journal of Information Technology |
| Product Type | : Article |
| Product Code | : IJIT10320 |
| Author Name | : Ashish Kumar, Gurjit Singh Walia, Kapil Sharma |
| Availability | : YES |
| Subject/Domain | : Engineering |
| Download Format | : PDF Format |
| No. of Pages | : 12 |
Generic Particle Filter (GPF) is extensively used in the area of computer vision for nonlinear and non-Gaussian state estimation. However, GPF suffers from the problem of sample impoverishment and particle degeneracy. The aim of the paper is to propose a method using Chaotic Crow Search Algorithm as resampling method to overcome these problems of GPF. The proposed method has been simulated on benchmark 1D and 2D state estimation problems. The simulation results of the proposed method are compared with GPF, particle filter- particle swarm optimization and particle filter-backtracking search optimization. On average of the outcome, we have achieved RMSE value of 2.0214 for 1D problem and value of 0.0281 for 2D problem for the proposed method. The results demonstrate that the method not only outperforms other methods but also achieves high accuracy with minimum computational requirement.
Generic Particle Filter (GPF) is based on Sequential Monte-Carlo framework. GPF has been widely explored in the fields of science, artificial intelligence, Robot intelligence, military, target detection and object tracking (Gurjit and Rajiv, 2015). However, GPF suffers from two fundamental problems of particle degeneracy and sample impoverishment (Miodrag et al., 2004; and Tiancheng et al., 2014). Resampling techniques like sequential resampling and partial resampling were explored with GPF to address these problems, but there is still scope of improvement.
Particle filter, CCSA, Sample impoverishment, Particle degeneracy