Rotation-Traction manipulation bionic training robot based on visual servo and impedance control

Abstract

Chinese Traditional Medicine Rotation-Traction manipulation is an effective treatment of Cervical Spondylosis. However, incidence of medical malpractice is relatively high in the rotation-traction therapy when the unfledged internship physicians give the therapy. But the conventional pedagogy is inefficient for lack of experimental platform. So it is crucial to develop an experimental platform to train the manipulation. The Rotation-Traction manipulation can be divided into two steps. The first step is patient's self-positioning by tracking clinician's hand, and the second step is the clinician giving therapy. By utilizing robotic and computer vision, a training robot was proposed in this paper to simulate the patient. A 2 DOF platform was designed to mimic the human neck and a variable stiffness structure was adopted to simulate human spine. During the self-positioning, a visual servo control system was proposed to track clinician hand. The robustness of hand tracking algorithm was enhanced by utilizing an improved Camshift. The tracking result was transformed into robot position command by Image Jacobian. The conventional position-based impedance control was utilized during the second step of manipulation to mimic the bio-mechanical characteristics of human neck. Experiments of visual servo and impedance control were carried out in this paper. The results suggested that the robot can track hand well and satisfy the requirement of clinical application.