Concurrent multiple cameras calibration and robot localization from visual and 3D inertial measurements

Abstract

Robots will be capable of navigating in the environment autonomously and efficiently with accurate pose estimates. In this paper, we develop an inertial sensor module that provides accurate absolute 3D velocity and orientation estimates. The module fuses together data from accelerometer, gyroscope and digital compass. A client-server infrastructure comprises of concurrent multiple cameras calibration and robot localization is also presented. The presented infrastructure integrates information from 3D inertial sensor module, odometry data from robot and images from overhead cameras in the environment to localize the robot and concurrently cameras can then be calibrated. Experiment results show the presented approach could calibrate multiple cameras and estimate robot poses iteratively with average errors around 20cm in world reference plane.