Real-time vision-based navigation for nonholonomic mobile robots

Abstract

Mobile manipulators are one of the major pillars in the thriving field of service robotics, perhaps mainly due to the wide workspace provided by the mobile base. Exploitation of this vast workspace entails accurate positioning of the robot while overcoming several complexities, such as nonholonomic constraints and actuator limitations. In this paper, we analyze the integration of a vision and path-following controller for the accurate localization and tracking of known objects. Due to the broad workspace of mobile manipulators, the improved repeatability of the mobile base positioning can significantly improve overall system performance, especially near the robot's workstation. Our method provides an accurate synchronization and real-time fusion of data that together with the closed-form solution of the given path-following controller manages to accurately navigate the robot to perform a gripping task. The achieved repeatability of the system is below 3 cm, which is demonstrated by experimental results.