Automatic robot kinematic modeling with a modular artificial skin

Abstract

In this paper, we present a new approach to automatically acquire the kinematic model of a robot equipped with a multi-modal, modular artificial skin. Here, we make use of the tri-axial accelerometer built into every cell of our artificial skin, along with a set of open loop test motions of the robot. With a circle point analysis we extract the relevant vectors to assemble a forward kinematic model. In comparison to existing approaches, our method builds and calibrates a kinematic model with minimal motions around the initial workspace, does not require any external metrology system and only requires low-grade motion sensors. We demonstrate the robustness of our approach by estimating the kinematic parameters of a UR-5 robotic arm mounted on an inertial dual arm platform.