Extrinsic Calibration of a Multiple Radar System for Proximity Perception in Robotics

Abstract

The simultaneous use of multiple small and low-cost radars has recently become feasible due to their increasing availability and functionality. Concerning the data fusion, computing the extrinsic parameters (rotation and translation) is a well known problem. However, the calibration of the sensor system is particularly challenging when dealing with devices having low number of antennas and therefore limited angular resolution, and there is currently no standard procedure for a setup consisting exclusively of such radars. This setup is though beneficial for collaborative and safety-oriented applications in robotics; therefore, we present an extrinsic calibration method for a multi-radar system deployed in a robotic cell. The calibration procedure only requires to move a single radar-signal reflector within the perceived area, without the need for additional sensing technology. The method is based on data sequential collection and pre-processing, combined with the closed-form registration of 3D point clouds. Furthermore, we include uncertainty information with the use of a custom MATLAB Toolbox <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sup> <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sup> https://www.aau.at/en/smart-systems-technologies/sensors-and-actuators/downloads/an-uncertainty-toolbox-2/. Two different data collection procedures, inspired by a state of the art Motion Capture system, are presented and evaluated.