Date of Award
Master of Science in Biomedical Engineering - (M.S.)
Richard A. Foulds
In theory, admittance control offers a very effective method of implementing smooth human-robot interaction. It allows the user’s applied force to control the movement of a powerful robot as if the robot were a small, passive mass. However, the real-world application of admittance control faces limitation posed by the dynamics of servo motors, the accuracy of the force sensors, and the computation speed of processors.
This research investigates the limitations on achieving compliant passive behavior when using state-of-the-art actuators, sensors and processors. The work involves characterizing the dynamic behavior of the servo motors, development of improved differential equations representing admittance control, and testing to determine the ability of a robotic system to represent the behavior of passivity. A method has been developed for experimentally determining the inertial, and dissipative (damping and friction) characteristics of three different models of Dynamixel motors. These parameters are optimized using data from a pendulum drop test with mass at various distances from the center of rotation. With these parameters, we assess the ability of our motor model to generate an ideal motion based upon a torque input from the user. The aim is to understand the limitations of our control paradigm to allow users to be unable to feel any difference between the performances of the passive and motor joints.
Rao, Ahmad Zahid, "Realization of dynamixel servo plant parameters to improve admittance control for a compliant human-robot interaction" (2015). Theses. 263.