We describe an autonomous robotic system capable of navigating through an office environment, opening doors along the way, and plugging itself into electrical outlets to recharge as needed. We demonstrate through extensive experimentation that our robot executes these tasks reliably, without requiring any modification to the environment. We present robust detection algorithms for doors, door handles, and electrical plugs and sockets, combining vision and laser sensors. We show how to overcome the unavoidable shortcoming of perception by integrating compliant control into manipulation motions. We present a visual-differencing approach to high-precision plug-insertion that avoids the need for high-precision hand-eye calibration.

PR2 is a sophisticated mobile manipulation platform designed for operation in dynamic and unstructured indoor environments. In this paper we describe an experiment using TREX, a hierarchical control framework based on constraintbased temporal planning, to coordinate PR2 behavior. The experiment was part of a fully integrated demonstration of PR2 capabilities involving autonomous navigation, dooropening, and recharging using standard electrical outlets. The goal of this experiment was to evaluate the applicability of a model-based, planning centric approach for practical robotics on a large scale. The results were encouraging. Not only did TREX play an important role in accomplishing the milestone, which was in itself a significant achievement in autonomous robotics, but it did so with modest computatational overhead and system complexity. In this paper we outline the details of the milestone and how TREX was used to achieve it, providing a quantitative and qualititative evaluation of TREX performance. We believe this presents a promising pathway for deep integration of declarative models, and automated planning as a paradigm for practical robot programming. All software described in this paper, including TREX, is available under an Open Source license, via http://ros.sourceforge.net.

The ability to follow man-made paths and roads is an important capability for a number of robotic tasks. To operate in outdoor environments designed for humans, autonomous robots must identify footpaths, and drive along them. In this paper, we describe a computationally efficient approach to identifying and following footpaths, using only a single camera. Our technique takes the robot’s kinematics into consideration when planning the best trajectory to follow the path that it is on. We show that our approach is highly robust to visual artifacts such as shadows, lighting changes, ground texture changes, and occlusions.