AN IMPROVED EXTENDED KALMAN FILTER BASED ON PIECEWISE SELF-ADJUSTING WEIGHTED NONLINEAR PREDICTIVE FILTERING ALGORITHM FOR MOBILE ROBOT POSITIONING AND NAVIGATION

Qiyuan Fan and Kin Sam Yen

Keywords

Kalman filter, SLAM, mobile robot, localisation and navigation, nonlinear filtering

Abstract

This paper introduces piecewise self-adjusting weighted nonlinear predictive filtering (PSAWNPF) algorithm, an enhanced extended Kalman filter (EKF) approach tailored for optimal mobile robot positioning and navigation. The algorithm uniquely incorporates quaternion-based attitude representation and a self-adjusting weighting mechanism into a conventional EKF framework. The quaternion-based attitude representation facilitates linear matrix operations, while the self-adjusting weighting mechanism dynamically adapts based on real-time error estimation, enhancing both computational efficiency and position estimation accuracy. Experimental validation on a mobile robot equipped with LiDAR and gyroscope demonstrates the superior performance of PSAWNPF- EKF, recording a 3.1% uncertainty interval for position estimation, compared to 5.3% for the unscented Kalman filter (UKF) and 6.5% for traditional EKF algorithms. The PSAWNPF-EKF algorithm also achieves a substantial 24% reduction in the average operation time for predefined routes, surpassing other algorithms. The findings affirm PSAWNPF-EKF as the most accurate and efficient option among the three algorithms for mobile robot positioning and navigation. Looking forward, future research can explore the adaptability and applicability of PSAWNPF-EKF in diverse scenarios, such as autonomous vehicle navigation and aerial drone mapping, further expanding its impact in the realm of robotics and autonomous systems.

Important Links:



Go Back