Triplet loss-based metric learning for haptic-only robot localization
[ 1 ] Wydział Automatyki, Robotyki i Elektrotechniki, Politechnika Poznańska | [ 2 ] Instytut Robotyki i Inteligencji Maszynowej, Wydział Automatyki, Robotyki i Elektrotechniki, Politechnika Poznańska | [ SzD ] doctoral school student | [ P ] employee
[2.2] Automation, electronics, electrical engineering and space technologies
2024
chapter in monograph / paper
english
- localization
- haptic sensing
- triplet loss
- transformers
EN This study investigates an approach to haptic localization for legged robots, employing triplet loss within a transformer-based neural network. Through experimentation, we evaluate diverse triplet loss variations and their impact on localization accuracy, shedding light on latent space structures. Our findings highlight the superiority of TL-BA triplet loss for haptic-only robot localization, surpassing alternative loss methods. This research not only enhances understanding of machine learning optimization for practical robotics but also identifies effective strategies for haptic localization implementation. Our insights pave the way for more refined methodologies in the development of robotic systems reliant on sparse sensory data.
338 - 345
copyright
publisher's website
final published version
at the time of publication
20