Xu, Yongyue; Chen, Jia; Su, Jinya; Yan, Yunda; Li, Shihua; (2025) Mobile Robot Localization and Prediction by UKF with Neural Network Aided Modeling. In: Proceedings of the 30th International Conference on Automation and Computing (ICAC 2025). (pp. pp. 1-6). IEEE: Loughborough, UK. (In press).

Preview

Text Mobile Robot Localization and Prediction by UKF with Neural Network Aided Modeling.pdf - Accepted Version Download (2MB) | Preview

Abstract

State estimation is crucial for mobile robot navigation, enabling tasks like motion planning and obstacle avoidance. Global Positioning System (GPS) signals are unreliable indoors due to obstructions, Ultra-Wideband (UWB) systems offer a robust alternative for indoor positioning. However, traditional IMU-UWB integration methods suffer rapid degradation in accuracy when low-frequency UWB measurement data is unreliable, primarily due to their dependence on the IMU kinematic model. In this paper, we propose an enhanced localization framework that synergizes kinematic modeling with neural network-aided techniques to enhance the IMU process model under GPS-denied and low-frequency measurement conditions. By embedding the improved model within an Unscented Kalman Filter (UKF), our approach markedly improves both single-step and multi- step prediction accuracy. Comparatively experimental results using a Mecanum-wheeled mobile robot demonstrate that the proposed system not only improves localization precision but also enhances robustness, providing a reliable solution for indoor 2D mobile robot navigation. The dataset and code are available at: https://github.com/YyX-ssr/MLP-Aided-UKF.

Download activity - last month

Export as CSVXMLJSON

Download activity - last 12 months

Export as CSVXMLJSON

Downloads by country - last 12 months

Export as JSONCSVXML

Archive Staff Only

View Item