Estimation of mass parameters for cooperative human and soft-robots as basis for assistive control of rehabilitation devices

Abstract

For safe human-robot interaction, the robots, especially exoskeletons, should be transparent for the user, in a way that the user does not feel their weight. In robot-assisted orthopedic rehabilitation, where patients are often sensitive to pain and often not able to lift affected limbs by themselves, this requirement is of crucial importance. To allow patients directly after surgical intervention to participate actively, not only the weight of the robot, but also the weight of the patient's extremity must be compensated. Several assistive acting soft robotic rehabilitation devices have been developed for upper and lower extremities based on pneumatic actuators with rotary elastic chambers (REC-actuators), well suitable for human-robot interaction because of natural inherent compliance. The essential basis for developed adaptive model-based assistive control strategies, in order to provide robot transparency as well as take into account patient's behavior, effort and abilities, and support patients with high voluntary effort, are separated models of soft-robots and of individual patient's extremities. This paper presents an approach to estimate unknown mass parameters of soft robotic rehabilitation devices as well as of individual human's extremities, using a minimal number of experimental measurements as well as specific characteristics of pneumatic REC-actuators without the use of torque sensors. An experimental study shows the influence of using the models based on estimated parameters during control, using a four DoF soft-robot as well as a prototype of an assistive elbow trainer with an unimpaired subject, as first part of a upper extremity rehabilitation robot.