EUVIMEDEUVIMED entdecken
Crossref

Embodied Terrain Recognition via Sliding Sensing on Closed-Chain Legged Mechanisms

Haoyu Sun, Zihan Zhang, Yu Du, Jianxu Wu

IEEE Sensors Journal · 01.06.2026 · 26 (11) · 16730-16741

DOI: 10.1109/jsen.2026.3684034 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Sensors Journal, erschienen am 01.06.2026. Verfasst von Haoyu Sun und 3 weiteren Autor:innen. Die Arbeit nennt 45 Literaturangaben. Gefördert von: National Natural Science Foundation of China.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The accurate and robust terrain perception is essential for legged robots operating in unstructured natural environments. Tactile sensing strategies based on open-chain leg mechanisms have contributed valuable insights into foot–ground interaction. To augment their sensitivity and reliability, we introduce a terrain recognition framework employing sliding-driven tactile sensing with a closed-chain legged mechanism, supported by a high-fidelity testbed for systematic triaxial force acquisition across terrains. A dedicated preprocessing pipeline converts the raw tactile signals into standardized single-cycle samples. We publish a dataset tailored for closed-chain legged systems, preserving periodic structure and distinct energy distribution characteristics. The dataset encompasses 11 terrain classes with over 1200 labeled sliding sequences collected at two motor speeds to capture speed-dependent tactile variation. To support high-frequency and low-latency perception in real-time outdoor scenarios, we design a time-series classification model based on deep learning that integrates a convolutional neural network (CNN), bidirectional long short-term memory (Bi-LSTM), and an attention mechanism. Experimental results demonstrate that the proposed model achieves an overall accuracy of 86.18%, requiring only 0.25 million parameters and achieving an inference time of just 1.7 ms/sample. These findings validate the effectiveness and robustness of sliding-driven tactile sensing in terrain recognition, laying a solid foundation for embodied environmental understanding and navigation in future closed-chain legged robotic systems.

Förderung

  • National Natural Science Foundation of China

Zitieren

APA

Sun, H., Zhang, Z., Du, Y., & Wu, J. (2026). Embodied Terrain Recognition via Sliding Sensing on Closed-Chain Legged Mechanisms. IEEE Sensors Journal, 26(11), 16730-16741. https://doi.org/10.1109/jsen.2026.3684034

Vancouver

Sun H, Zhang Z, Du Y, Wu J. Embodied Terrain Recognition via Sliding Sensing on Closed-Chain Legged Mechanisms. IEEE Sensors Journal. 2026;26(11):16730-16741. doi:10.1109/jsen.2026.3684034

BibTeX
@article{Sun2026,
  author = {Sun, Haoyu and Zhang, Zihan and Du, Yu and Wu, Jianxu},
  title = {Embodied Terrain Recognition via Sliding Sensing on Closed-Chain Legged Mechanisms},
  journal = {IEEE Sensors Journal},
  year = {2026},
  volume = {26},
  number = {11},
  pages = {16730--16741},
  doi = {10.1109/jsen.2026.3684034}
}
Kostenlos starten →