Vision-Guided Inch-Perfect Docking of an Autonomous Robot скачать в хорошем качестве

Vision-Guided Inch-Perfect Docking of an Autonomous Robot

In this video, we show several docking tests of our Scitos G3 robot as part of longterm trials in the CompanionAble project. The robot autonomously drives to a place in front of the docking station. Afterwards, its navigation software switches seamlessly into docking mode, where the robot localizes itself in front of the docking station using a checkerboard pattern that is tracked by a webcam in the rear of the robot. While driving backward, the robot follows an alignment trajectory that perfectly aligns the robot's charging connector with a docking plug that is mounted on the docking station. Although the docking plug is slightly flexible, the precision of the docking alignment must be within 2 centimeters, which is quite challenging for a robot of the given size and weight and with restricted free space in front of the docking station. The navigation software is able to detect if the robot cannot reach the station with the required precision, e.g. when the initial pose of the robot is too far away from the docking station due to localization errors. In this case it performs a correction movement to achieve a better alignment as shown in the test runs #2 and #3. Using this failsafe docking technique the robot is able to operate fully autonomously around the clock without need of intervention in our longterm trials. The Scitos G3 robot platform is manufactured by the MetraLabs Robotics GmbH (http://www.metralabs.com). The navigation software is a joint development of MetraLabs and Neuroinformatics and the Cognitive Robotics Lab at the Ilmenau University of Technology (http://www.tu-ilmenau.de/neurob).