Generalized momentum based-observer for robot impact detection — Insights and guidelines under characterized uncertainties
Abstract
In the context of human-robot interaction, detecting an impact between a serial robot manipulator and the human operator or its environment efficiently and at the earliest is essential for safe and efficient operations. This work relies on the generalized momentum based-observer for robot impact detection using only motor/joint information and characterizes the uncertainties to quantify the expected sensitivity of detection. Indeed, modeling uncertainties affect the external torque estimation in the same structural way as external disturbances and no structural decoupling is reachable between both. For this purpose in this study, modeling uncertainties are divided into parametric and numerical differentiation errors. Their contribution in process and measurement noises is detailed and then approximated by a white noise of characterized variance for the observer design. High-level tuning guidelines are provided for the design parameters depending on the expected speed and sensitivity of detection. This approach is applied experimentally on the CEA robot arm manipulator.