Learning-Aided Wheel-Encoder Reliability Assessment and Adaptive Sensor Fusion in Ground-Robot Odometry
IEEE Sensors Journal · 2026-11 · 26 (21) · 33103-33110
DOI: 10.1109/jsen.2026.3735115 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Sensors Journal, 2026-11. Verfasst von Jialu Li und 4 weiteren Autor:innen. Die Arbeit nennt 30 Literaturangaben.
Abstract (Originalsprache)
Wheel encoders provide low-cost, high-rate motion sensing for ground robots, but their reliability can degrade sharply under slip, low-traction surfaces, snow, grass, rough terrain, or floating-wheel conditions. Direct fusion of the resulting anomalous forward velocity can corrupt odometry. This article presents a learning-aided wheel-encoder reliability assessment and adaptive sensor-fusion method. A short-horizon model driven by an inertial measurement unit (IMU) provides an encoder-independent consistency reference rather than a replacement odometry source. Wheel–reference disagreement is converted into reliable, uncertain, and unreliable modes, which, respectively, enable a normal update, covariance-inflated update, or rejection of the encoder observation. The scalar forward-velocity cue is incorporated through a reliability-aware update with adaptive variance. Eight held-out sequences from three public datasets cover nominal driving, snow, grass, floating-wheel, and unpaved-orchard conditions. Controlled same-backbone results show selective average improvement rather than per-sequence dominance, while direct instrumentation confirms real-time operation and modest mapper overhead.
Zitieren
APA
Li, J., Shi, Y., Liu, Y., Liang, B., & Lu, W. (2026). Learning-Aided Wheel-Encoder Reliability Assessment and Adaptive Sensor Fusion in Ground-Robot Odometry. IEEE Sensors Journal, 26(21), 33103-33110. https://doi.org/10.1109/jsen.2026.3735115
Vancouver
Li J, Shi Y, Liu Y, Liang B, Lu W. Learning-Aided Wheel-Encoder Reliability Assessment and Adaptive Sensor Fusion in Ground-Robot Odometry. IEEE Sensors Journal. 2026;26(21):33103-33110. doi:10.1109/jsen.2026.3735115
BibTeX
@article{Li2026,
author = {Li, Jialu and Shi, Yongliang and Liu, Yu and Liang, Bin and Lu, Weining},
title = {Learning-Aided Wheel-Encoder Reliability Assessment and Adaptive Sensor Fusion in Ground-Robot Odometry},
journal = {IEEE Sensors Journal},
year = {2026},
volume = {26},
number = {21},
pages = {33103--33110},
doi = {10.1109/jsen.2026.3735115}
}
EUVIMED