Multitask Learning for Multiple Recognition Tasks in Lower-limb Exoskeleton robot
Goal : This study shows that Gait Phase Recognition (GRP) and Terrain Classification (TC), the most conventional Recognition Tasks of Lower-limb Exoskeleton robots, can be effectively solved by introducing a Multi-task Learning.
Summary Multi-task learning, (or Transfer learning), is a concept derived from the field of computer vision, and is a learning method that enables better performances and data-efficient learning when learning is conducted by sharing a Feature Network (encoder) for related tasks.
Researcher Joon-hyun Kim developed a model that predict Gait Phase Recognition (GRP) and Terrain Classification (TC), the most conventional Recognition Tasks of Lower-limb Exoskeleton robots, based on theses Multitask Learning Framework idea.
He first created a high-performing GPR model that achieved a Root mean square error (RMSE) value of 2.345 ±0.08 and then utilized its knowledge-sharing backbone feature network to learn a TC model with an extremely limited dataset. Using a limited dataset for the TC model allows us to validate the data efficiency of our proposed Multitask learning approach. He compared the accuracy of the proposed TC model against other TC baseline models. The proposed model achieved 99.5 ±0.044% accuracy with a limited dataset, outperforming other baseline models, demonstrating its effectiveness in terms of data efficiency.
Topic
- Recognition Problems in Lower-limb Exoskeleton Robot
- Few-Shot Learning
- Multitask Learning
Conference
- kspe (Korean Society for Precision Engineering) Conference
- IEEE ROMAN 2023 (Acceted, Announcement scheduled for August 28)
Video
Multitask Model Online Test Video to simultaneously Re ognize Gait Phase and Terrrain
related project