Phase-based Dynamic Movement Primitives for Powered Knee-Ankle Prosthesis Control

Guoqing Zhang,Honglei An,Yi Liu,Yongshan Huang,Hongxu Ma,Qing Wei
DOI: https://doi.org/10.1109/icras55217.2022.9842244
2022-01-01
Abstract:This paper presents a phase-based dynamic movement primitives method applied to the motion control of a knee-ankle active prosthesis. The phase-based control method uses continuous, user-synchronised phasing to unify the gait cycle of a powered knee-ankle prosthesis. The residual thigh motion is used to calculate the phase variables, which are then filtered to eliminate noise interference. Dynamical movement primitives (DMPs) are popular methods of reproducing trajectory for learning control. We use phase as a 'bridge' to obtain the forcing term by means of didactic learning, and then use DMPs to reproduce the learned joint trajectory and generate the desired joint angle for the control of the powered prosthesis. Finally, simulations of the method were run to test the stability and resilience of the phase-based DMPs, indicating our method has a potential to facilitating prosthesis motion control.
What problem does this paper attempt to address?