HAPTR2: Improved Haptic Transformer for legged robots’ terrain classification

The haptic terrain classification is an essential component of a mobile walking robot control system, ensuring proper gait adaptation to the changing environmental conditions. In practice, such components are a part of an autonomous system and thus have to be lightweight, provide fast inference time...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Robotics and autonomous systems Jg. 158; S. 104236
Hauptverfasser: Bednarek, Michał, Nowicki, Michał R., Walas, Krzysztof
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Elsevier B.V 01.12.2022
Schlagworte:
Data sets for robot learning
Deep learning methods
Legged robots
Data sets for robot learning
Legged robots
Deep learning methods
ISSN:0921-8890, 1872-793X
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:The haptic terrain classification is an essential component of a mobile walking robot control system, ensuring proper gait adaptation to the changing environmental conditions. In practice, such components are a part of an autonomous system and thus have to be lightweight, provide fast inference time, and guarantee robustness to minor changes in recorded sensory data. We propose transformer-based HAPTR and HAPTR2 terrain classification methods that use force and torque measurements from feet to meet these requirements. For reliable comparison of the proposed solutions, we adapt two classical machine learning algorithms (DTW-KNN and ROCKET), one temporal convolution network (TCN), and use the state-of-the-art CNN-RNN. The experiments are performed on publicly available PUTany and QCAT datasets. We show that the proposed HAPTR and HAPTR2 methods achieve accuracy on par or better than state-of-the-art approaches with a lower number of parameters, faster inference time, and improved robustness to input signal distortions. These features make HAPTR and HAPTR2 excel in terrain recognition tasks when considering real-world requirements. •Fast and lightweight transformer-based approach for the terrain classification.•Modality Attention Layer improves the system’s robustness in real-world scenarios.•Autonomy requirements inspired our evaluation of terrain classification algorithms.•The benchmark on two open source datasets creates a baseline for further research.
ISSN:0921-8890
1872-793X
DOI:10.1016/j.robot.2022.104236