Deployment of an autonomous mobile manipulator at MBZIRC

Jan Carius,Martin Wermelinger,Balasubramanian Rajasekaran,Kai Holtmann,Marco Hutter

DOI: https://doi.org/10.1002/rob.21825

IF: 6.385

2018-10-27

Journal of Field Robotics

Abstract:Abstract This paper introduces a custom mobile manipulator and presents the planning and control architecture to execute autonomous missions including navigation and manipulation. The developed system combines a commercially available ground vehicle with a custom‐designed robot arm. A typical mission requires the robot to find and navigate to its site of action, identify the suitable manipulation tools, and manipulate the correct objects according to the task specification. Our contribution includes a complete software pipeline for autonomous map building, localization, and navigation. We demonstrate stable localization on vast outdoor areas without GPS support. Our exploration strategy coupled with a path planner reliably scans the environment while avoiding obstacles until it finds the manipulation site. The position of the target location is estimated from laser data within a 5 cm radius, and the subsequent positioning procedure of the mobile base achieves a precision of ±4 cm. The developed manipulator uses series elastic actuation, making the system mechanically robust and compliant. Thanks to the precise joint torque controllability, a low end‐effector impedance can be realized to passively overcome positioning inaccuracies during manipulation. Our solution incorporates a support vector machine to detect the manipulation object as required for visual servoing. We show how a valve stem and standard wrenches can be identified with a detection rate higher than 90%. The paper finally discusses the performance and lessons learned while participating at the Mohamed Bin Zayed International Robotics Challenge (MBZIRC) 2017, where we showed the successful autonomous operation in a partially unknown environment.

robotics