Multi-Source Sensor Fusion for Robust Localization and Perception in Contactless Delivery Vehicles: A Hybrid Approach Combining Extended Kalman Filtering and Transformer Networks

View at publisher Summarise with VIRAT
Authors: Wenji Zuo
Cited by: 0
References: 9

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

This study proposes a multi-sensor fusion framework to enhance localization and environmental perception accuracy in unmanned delivery vehicles. The front-end module of the system employs the extended Kalman filter for real-time state estimation, and the rear-end achieves multi-modal feature fusion through the transformer architecture. This design, which combines traditional algorithms with deep learning, enables robust navigation and perception across urban roads, industrial parks, and indoor environments. The test platform is equipped with four types of sensors—satellite positioning, inertial navigation, monocular vision, and lidar—to collect data in complex urban areas, suburban roads, and warehouse environments, respectively. Based on high-precision RTK positioning data, experiments show that this scheme reduces the root mean square error (RMSE) in positioning by 30% and improves obstacle recognition accuracy by 15%. The system's processing time is constrained to under 200 milliseconds to meet the real-time requirements of logistics transportation. These findings validate the technical advantages of hybrid sensor fusion in dynamic environments, and offer a reliable solution for automating last-mile delivery systems.

Fields of Study & Subjects

Robotics and Sensor-Based Localization Autonomous Vehicle Technology and Safety Inertial Sensor and Navigation

Publication details

Journal / source
IEEE International Conference on Power, Electronics and Green Energy (ICPEGE)
Pages
666–670
Publication date
28 January 2026
Language
EN