Willow Garage Blog
Ethan Dreyfuss, who recently received a master's degree from Stanford University, is continuing his work here on autonomous person-following and dataset collection and annotation. The former project provides a useful building block for a wide variety of tasks. Consider a robot that helps you carry groceries. This robot is vastly more useful if it can carry your bags to the house without requiring teleoperation; the robot can simply track you and follow behind. At a high level, person-following comprises two principal tasks: person tracking and navigation.
The approach developed by Ethan and Caroline Pantofaru fuses a face detector with two weak person trackers: one for legs, and one for 3D blobs at person-height. None of these approaches is individually effective enough to provide robust tracking, but their
strengths are complementary. The face detector is effective when the person is close to, and directly facing the robot. While the leg tracker provides high accuracy
when multiple people are present, it is often confused by non-human
obstacles and can therefore not work reliably from afar. Conversely, the height-based blob tracker can effectively track from further away, yet it is
easily confused by groups of people. By combining techniques, Ethan and Caroline were able to develop a more robust person-tracking tool.
Once the robot can track a designated person, the information is passed on to the navigation stack. This same navigation software was used to complete Milestone 2, with some improvements made to help deal more quickly and robustly with dynamically-moving obstacles such as people.
In addition to the person-following project, Ethan is contributing to the collection and labeling of a large dataset of people in an indoor office environment. One of the major drivers of computer vision research is the availability of high-quality labeled data. The bulk of existing person datasets exclude indoor environments, and instead focus on outdoor pedestrians. Indoor environments present numerous challenges for person detection, including poor lighting and environmental clutter. By automating as much as possible, the process of both collecting (using the robot) and labeling (using Amazon's Mechanical Turk and Alex Sorokin's CV Web Annotation Toolkit), Ethan's team will be able to provide a large, compelling dataset to encourage other researchers to tackle these challenging problems.
Ethan also picked up a number of side projects including rapid neighborhood computation on point clouds, and implementing a package that uses the open-source video codec Theora to allow low-bandwidth video streaming within ROS.
The engineering and production team is making significant progress on the production of Beta PR2 robots. After completing the grippers, attention has been focused on the various other assemblies. Most of the other assemblies -- head, laser sled, forearm, upper arm, shoulder, caster, spine, base, and turret -- are in process, with one head already in testing, and additional heads in burn-in. There are already twenty-four full casters in production, which is enough for six PR2 robots. Most of these casters have already undergone qualification tests, with six units also run through our rigorous burn-in process. After all of the casters are complete, they will be set aside and later mated with the base and spine.
Of the over six thousand parts needed to build each PR2, we are only waiting on the final twenty-four. Our stockroom is filled nearly to capacity, and as we continue to combine one sub-assembly with another, the PR2 robot is beginning to take shape.
-- The PR2 Builders
If you like robots as much as we do, we hope you'll enjoy this montage of clips from IREX 2009 in Tokyo, Japan. It's amazing to see the diversity of robots all on display in one exhibit hall.
Steve Cousins (CEO) and Keenan Wyrobek (Co-Director, Personal Robots Program) gave a talk at the IREX 2009 Robolink Forum. This talk, "A Platform for Personal Robotics", gives our perspective on the future of the personal robotics industry, discusses the importance of open platforms, and provides an introduction to the ROS and PR2 platforms.
We've uploaded the talk for you to view and have divided it into four sections. You can also download a PDF version of the slides.
Part 1: Lessons from the Personal Computer Industry
Part 2: ROS
Part 3: PR2
Part 4: Community for Open Platforms
Along with our Texas Robot experimentation, we continue our endeavors to improve the usability of our software and hardware components. By focusing outwards on the ROS and PR2 communities, we've made significant progress towards improving user experience, and we continue to iterate on these changes with the generous help of numerous ROS community members. We've been running software and hardware component tutorials through rigorous user testing, and along the way, discover opportunities to develop new teaching tools.
In an effort to simplify ROS adoption for new users, we developed turtlesim, a LOGO-inspired tool that provides a hands-on approach to learning ROS basics. This tool functions as an entry-level tutorial that takes in velocity commands, and "drives" a turtle according to the input. Turtlesim offers a simple simulator that allows new users to more readily visualize their commands and work with a simulated "robot." Once comfortable with some of the more basic ROS commands, users can try simulators, like Stage and Gazebo, for more advanced experimentation and tutorials.
These tutorials, along with many others, can be found at ros.org. There are currently 19 tutorials available for the core ROS system, and 186 tutorials (and counting) in total, covering much of the functionality available on ROS. To learn how to document a package, check here, and to learn how to write a tutorial for a package, click here. Ros.org has seen great expansion and improved organization. We strongly encourage you to upload your work and share your documentation and tutorials with the ROS community!
Although Milestone 3 keeps us busy with PR2 beta production and software documentation, we continue to have fun with other experiments around the office. Our Texas 1 telepresence robot was been such a hit, that we couldn't help but tinker with its design and capabilites. Dallas Goecker still regularly uses Texas 1 to telecommute, but his surrogate at Willow Garage was frequently incapacitated by our eager experimentation. Instead of sharing, we built a couple more.
Texas 4 is our high-speed prototype. A touchscreen allows developers to interact directly with the robot, as opposed to doing so via the teleoperator. This version has more powerful motors and a differential-drive configuration. With the drive wheels up front, we were able to more creatively experiment with Texas 4's capabilities.
Two of our controls engineers, Wim Meeussen and Melonee Wise, spent a day writing a dynamically-balancing controller that allows Texas 4 to balance on its two front wheels, and move about like a Segway. The robot can switch between 4-wheel drive and 2-wheel drive on the fly. With an IMU (Inertial Measurement Unit) taped to the base, the robot self-adjusts to a pre-designated forward tilt angle whenever stability is compromised. The result was a fun sight to see around the office as well as a good experiment into the capabilities of the ROS/PR2 software platform. Wim and Melonee were able to use the same controller infrastructure they are developing for the PR2. This infrastructure allowed them to write and compile their new controller for Texas 4 without having to restart the entire system. They were also able to access many other libraries for ROS, like the filters package, which enables runtime-configurable data filtering.
ROS 0.10 has been released!
The focus of this release is preparing ROS for a 1.0 release. We have been making final reviews of the ROS APIs to address any essential feature gaps, API inconsistencies, and deprecations. Documentation is also being improved and reviewed. Final deprecations prior to 1.0 are described on the roadmap and are either being made in this release or 0.11.
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.
After ARSO 2009, Willow Garage will head over to the International Robot Exhibition (IREX) 2009 in Tokyo, November 25-28. IREX provides a venue for exhibiting and promoting new technologies in the robotics field. CEO Steve Cousins and Personal Robotics Program Co-Director Keenan Wyrobek will discuss ROS and the PR2 platform at the Robot Business Forum on November 25. We'll also be there to share hardware and software demos. Check out the video below for a demo teaser, and stop by our IREX booth to learn more about Willow Garage!
We'll be set up at SR2-21 East 3 Hall, Tokyo Big Sight.
Willow Garage is getting ready to head over to Japan at the end of this month. Our first stop will be in Tokyo at ARSO 2009, November 23-25. ARSO is the IEEE Workshop on Advanced Robotics and its Social Impacts. Our CEO, Steve Cousins, will give a plenary talk on November 23 titled, "Building an Open Source Robotics Platform." This talk will provide insight on why we're working on ROS, and why we're committed to releasing our code as open source. Check here for detailed information on the conference schedule and program. If you're going to be at ARSO, and want to learn more about Willow Garage, please stop on by. See you there!