Brown ROS package: Reproducibility for shared experimentation and learning from demonstration

Abstract

We describe our efforts toward developing infrastructure for shared remote robotics laboratories that allow reproducible robotics experimentation and large scale learning from demonstration (LfD). Our goal is to create a facility where users demonstrate desired tasks by teleoperating robot platforms through video-game style web-based interfaces, providing data that LfD approaches will turn into robot policies. We describe a selection of open source tools, provided by the Brown robotics lab, created in service of building such a remote lab. Supported Platforms Our work has focused on off-the-shelf systems that are available for purchase by other research groups. While much of the work in the Brown ROS Package can be used on any platform running ROS, we also provide drivers for two widely-available platforms: the iRobot Create and the Aldebaran Nao (pictured in Figure 1).