Computing the positioning error of an upper-arm robotic prosthesis from the observation of its wearer's posture (bibtex)
by , , ,
Bibtex Entry:
@InProceedings{2021ACTIalexisicra,
  author    = {Poignant, A and Legrand, M and Jarrassé, N and Morel, G},
  title     = {{Computing the positioning error of an upper-arm robotic prosthesis from the observation of its wearer's posture}},
  booktitle = {{2021 IEEE International Conference on Robotics and Automation (ICRA)}},
  year      = {2021},
  pages     = {xx},
  address   = {Xi'an, China},
  month     = May,
  abstract  = {When the arm prosthesis worn by an amputated Human being is not adequately configured with respect to the end-effector task, {\it body compensations} are often observed. Namely, to compensate for a wrong joint positioning on the robotic distal side, a subject trying to reach a desired position/orientation of his/her hand mobilizes his/her proximal joints, thus exploiting the redundancy of the human+robot kinematic chain. In this paper, we explore the possibility of exploiting this well-known behavior to reverse the causality: if we observe the posture of an amputated subject wearing a prosthesis during a hand positioning task, to what extent can we infer the positioning error of the prosthesis? To answer this question, we make the assumption that the adequate, or natural posture for a given task is one that optimizes a postural score. The proposed approach then consists in i) measuring the joint posture of the subject fitted with the prosthesis; ii) search for an alternative posture that optimizes a postural score within the null space of the human+robot kinematic chain and iii) compute the position error for the robot joints between the initial and the optimized posture. An experimental evaluation is provided with non amputated subjects who emulate erratic positioning of their distal joints during hand positioning tasks. Results show that joint errors are estimated with a precision that seems compatible with the implementation of a real time control algorithm.},
  category  = {ACTIS},
  crac      = {n},
  hal       = {y},
  owner     = {nath},
  timestamp = {2021.07.20},
}
Powered by bibtexbrowser