Willow Garage Blog

Victoria Groom of Stanford University has been performing teleoperation and autonomy experiments at Willow Garage, looking for ways to improve usability, safety, and effectiveness in human-robot interactions. While some robots are driven by co-located operators (e.g., surgical teleoperated robots), other robots are teleoperated remotely (e.g., search and rescue robots) or are even sent to act autonomously (e.g., drones). These different types of interaction affect how people think and feel about these robots. While remote teleoperation may be appropriate in one instance, the same type of interaction could be ineffective or even dangerous in another.

Victoria just finished running the last of approximately 84 study participants, and has just disassembled her experimental set-up.

For the last few months, one room at Willow Garage has been dedicated to Victoria's human-robot interaction experiment investigating the effects of autonomy vs. teleoperation, direct vs. mediated view, and real vs. simulated robot presence, on the user experience of self-extension into robots.  Self-extension refers to the feeling that an object is a part of oneself-- more like a limb or familiar tool than a separate entity.

To accomplish this, she set up a floor-to-ceiling curtain with a monitor-size rectangle cut into it. The curtain separates a small workspace with desk, chair and computer, from the robot and prop area.  The robot area is outfitted with pictures of such items as a flashlight, tarp, rope, and water. These objects are part of the desert survival task, an exercise that asks participants to imagine that they are stranded in desert and must select a limited number of survival-critical objects to salvage from their luggage.  In Victoria's experiment, participants used the surveillance robot Rovio, and were told to select their choices with the assistance of the robot.

Half of the participants were told that Rovio autonomously approached the selections they made, while other participants used a small mobile device to drive the robot to their item choices. Volunteers were also put into different viewpoint conditions. While some were able to see the robot directly through the rectangular window in the curtain, others watched the robot on a large monitor with the help of a webcam mounted in the robot area. Additionally, some participants carried out the study with an actual Rovio, while others were shown the entire setup in a simulation format.

This study aims to help uncover important insights on how robot autonomy, mediation, and virtuality affect people's behaviors and feelings about robots. With this increased understanding, we will be able to design and build robots that can be used more easily and effectively.

ROS 0.11 has been released! ROS 0.11 includes many of the final features and API modifications that we wish to have in place for ROS 1.0. Our focus with this release was on improving consistency, better performance, and bug fixes. We are also removing much of the deprecated functionality with this release so that ROS 1.0 will have minimal deprecated functionality.

Some of the major updates in this release include new "wait for message" API routines in roscpp and rospy, bash tab-completion for most command-line tools, improvement in Python-related startup latency, and streamlining of the service API in rospy.

There were many other changes with this release as we work to improve and finalize the ROS feature set. You can find a complete list in the changelist.

You can download a tarball of the release, though we recommend that you start with the ROS installation instructions instead.

Notes on Updating from 0.10 or earlier in SVN: There are a number of previously installed programs which are now versioned scripts. SVN will error when updating and not delete the installed programs to add the scripts in their place. You will need to delete the program and svn up again. The error will read:

svn: Failed to add file 'FILENAME': an unversioned file of the same name already exists

The solution is to rm FILENAME && svn up again.

Peter Pastor, a PhD student at USC, spent the past three months developing software that allows the PR2 to learn new motor skills from human demonstration. In particular, the robot learned how to grasp, pour, and place beverage containers after just a single demonstration. Peter focused on tasks like turning a door handle or grasping a cup -- tasks that personal robots like PR2 will perform over and over again. Instead of requiring new trajectory planning each time a common task is encountered, the presented approach enables the robot to build up a library of movements that can be used to execute these common goals.  For this library to be useful, learned movements must be generalizable to new goal poses. In real life, the robot will never face the exact same situation twice. Therefore, the learned movements must be encoded in such a way that they can be adapted to different start and goal positions.

Peter used Dynamic Movement Primitives (DMPs), which allow the robot to encode movement plans. The parameters of these DMPs can be learned efficiently from a single demonstration, allowing a user to teach the PR2 new movements within seconds. Thus, the presented imitation learning set-up allows a user to teach discrete movements, like a grasping, placing, and releasing movement, and then apply these motions to manipulate several objects on a table. This obviates the need to plan a new trajectory every single time a motion is reused. Furthermore, the DMPs allow the robot to complete its task even when the goal is changed on-the-fly.

You can find out more about the open source code for Peter's work here, and check out his presentation slides below (download PDF). For more about Peter's research with DMPs and learning from demonstration, see "Learning and Generalization of Motor Skills by Learning from Demonstration", ICRA 2009.

Our online animation study has gone live! We invite you to weigh in with your perspective on our PR2 animations. You got a sneak preview of these animations as they were under development, but now you can check out the full study.

The above video shows an example of the type of clip you'll find in the full-length study. We're investigating how to make personal robot behaviors more human-readable by identifying important principles that can inform the design of more effective human-robot interactions. To do this, we need your feedback on how you interpret certain robot behaviors. Please follow the link up top, and help us make PR2 more human-readable!