Assist-as-Needed Framework for Robotic Rehabilitation: Adaptive Admittance Control with Passivity-Based Safety Features
L Pezeshki, H Sadeghian, X Chen, M Keshmiri, S Haddadin, A Mohebbi
L Pezeshki, H Sadeghian, X Chen, M Keshmiri, S Haddadin, A Mohebbi
Supervisor
Department
Robotics
Embargo End Date
Type
Journal article
Date
2025
License
Language
English
Collections
Research Projects
Organizational Units
Journal Issue
Abstract
This paper presents an adaptive admittance control scheme that integrates an adaptive neural network (NN) algorithm as a shared autonomy framework to achieve the Assist-as-needed (AAN) property in robotic rehabilitation. The proposed algorithm enables real-time adjustment of control parameters based on human performance, without requiring extensive offline training. An energy-based performance index dynamically balances tracking accuracy with minimal robotic intervention to encourage active human participation. Furthermore, a modified virtual energy tank approach is introduced to preserve system passivity, preventing unsafe behaviors. Experimental results underscore the algorithm’s adaptability, ensuring compliant behavior as evidenced by a notable 83% reduction in average stiffness, reflecting a corresponding decrease in robotic intervention, due to detection of active human participation. Moreover, the algorithm ensures safe interaction and effective task completion. These findings highlight the framework’s potential for improving robotic rehabilitation by intelligently adapting to user needs and providing safety-aware control.
Citation
L. Pezeshki, H. Sadeghian, X. Chen, M. Keshmiri, S. Haddadin and A. Mohebbi, "Assist-as-Needed Framework for Robotic Rehabilitation: Adaptive Admittance Control with Passivity-Based Safety Features," in IEEE Transactions on Medical Robotics and Bionics, doi: 10.1109/TMRB.2025.3643943
Source
IEEE Transactions on Medical Robotics and Bionics
Conference
Keywords
Rehabilitation robotics, Assist-as-needed control, Adaptive Neural Network, Passivity preservation
Subjects
Source
Publisher
IEEE