Task-dependent impedance and implications for upper-limb prosthesis control

Modern-day prosthetic limbs are currently unable to imitate the versatile interaction behaviors of real human arms. Although humans can vary the impedance of their arms, commercially available prosthetic limbs have impedance properties that cannot be directly controlled by users. We investigate the hypothesis that user-selectable prosthesis impedance properties could improve the user's ability to interact effectively with a variety of environments. We report the results of a series of human subject studies exploring this hypothesis using either a virtual prosthesis or a robot arm as a prosthesis proxy. We observed human performance with different stiffness and damping levels in the prosthesis proxy in two one-degree-of-freedom tasks: (1) a force minimization task and (2) a trajectory tracking task. The virtual prosthesis studies focus on human performance in an ideal simulated system to avoid limitations of a physical implementation, whereas the robot arm study focuses on performance changes that result from limitations of physical robotic hardware. The virtual prosthesis results showed that task-dependent impedance can improve user performance and that users can evaluate the effects of changing impedance. The robot arm results showed similar performance benefits of task-dependent impedance in a physical robotic system. These studies identified areas in which non-ideal characteristics of the physical system limited users' performance; most notably, the physical system could not achieve the low damping levels that helped subjects reduce contact forces in the virtual prosthesis studies. Thus, we identify some design considerations for prostheses with user-selectable impedance that can achieve useful impedance ranges for improving user performance.