Amplifying Human Strength Through a Virtual Sprung Inertia

While better known for their ability to work autonomously in structured tasks, robots have the potential to aid human workers in unstructured tasks via force control. More specifically, cobots and exoskeletons that physically interact with human operators can amplify human strength by feeding back measurements of their interaction forces, agnostic to the task the operator is performing. However, due to limited control bandwidth and coupled human-robot dynamics, naive forcefeedback can easily result in instability similar to pilot-induced oscillation in planes. Controllers that amplify strength while also guaranteeing some stability properties have become the standard approach in the space, often at the cost of the user's perception of a transparent system. This paper 1) extends a previously theorized virtual-mass-based exoskeleton controller to a strength-amplifying robot arm, 2) theoretically and experimentally demonstrate the improved stability properties that result from the inclusion of the virtual mass term, 3) identifies the robot's end-effector compliance and self-deflection transfer functions, and 4) determines the extent to which the robot can reliably measure the stiffnesses of a spring-box apparatus in order to better understand its ability to identify human mechanical impedance.