Simulated Kalman Filter with Randomized Q and R Parameters

Nor Hidayati Abdul Aziz, Nor Azlina Ab. Aziz, Zuwairie Ibrahim, Saifudin Razali, Mohd Falfazli Mat Jusof, Khairul Hamimah Abas, Mohd Saberi Mohamad, Norrima Mokhtar

Proceedings of International Conference on Artificial Life and Robotics · 2017 · 13 citations · 3 references

DOIFull text

Open access

Concepts

Abstract

Inspired by Kalman filtering, simulated Kalman filter (SKF) has been introduced as a new population-based optimization algorithm. The SKF is not a parameter-less algorithm. Three parameter values should be assigned to P, Q, and R, which denotes error covariance, process noise, and measurement noise, respectively. While analysis of P has been studied, this paper emphasizes on Q and R parameters. Instead of using constant values for Q and R, random values are used in this study. Experimental result shows that the use of randomized Q and R values did not degrade the performance of SKF and hence, one step closer to the realization of a parameter-less SKF.

References

3