Willow Garage Selects Eleven Recipients of PR2 Beta Robots Valued at Over $400,000 Apiece!

Willow Garage

Today Willow Garage announced that eleven (rather than the original ten anticipated) PR2 Beta robots, with a total value of over $4.4M, will be loaned out to academic and research institutions worldwide to develop a slew of impressive capabilities over the next two years.  The recipients include 7 US-based institutions, 3 European, and 1 Asian.  The final list is a panoply of robotics specialists:  University of Freiburg (Germany), Bosch, Georgia Tech, KU Leuven (Belgium), MIT, Stanford, TU Munich (Germany), UC Berkeley, U Penn, USC, and University of Tokyo (Japan) -- full details can be found in the Willow Garage press release.  It is difficult to overstate the importance of this event in the grand history of robotics...  Let me try to explain.

First, an introduction to the PR2 Beta robot:

PR2 Robot by Willow Garage  PR2 Robot by Willow Garage  PR2 Robot by Willow Garage


In just a few short years, with funding from early Google employee Scott Hassan, Willow Garage has developed an impressive piece of robot hardware already capable of autonomous indoor navigation, opening doors, plugging itself in, and folding towels.  By giving away eleven identical high-end hardware platforms, Willow Garage has given the community a common platform for development, rigorous testing, and common base-line comparisons. 

Perhaps more significantly, Willow Garage has shepherded the creation of a hardware-agnostic Robot Operating System called ROS that has usurped most other "robot operating systems" and sensor visualization packages that preceded it.  Literally, ROS is in use by dozens of high-end heterogeneous robots worldwide:

Robots Using Willow Garage's ROS (Robot Operating System)

Robots Using Willow Garage's ROS (Robot Operating System)

Robots Using Willow Garage's ROS (Robot Operating System)


Oh, and did I mention that Willow has revived development on the ubiquitous OpenCV computer vision libraries and extensively publishes their work (more than 10 papers being presented this week at ICRA 2010 alone)?  All of this seems to fit nicely with Willow's stated vision and commitment to open source:

We see personal robots as the next paradigm-shifting personal productivity tool.  By investing in open source and open platform adoption models, we aim to lay the groundwork for the use of personal robotics applications in everyday life.

We are committed to open source robotics software and the furtherance of the open source personal robotics community. We helped found, and continue to contribute heavily to, the robot operating system, ROS.  The ROS software we contribute is BSD-licensed, making it completely free for anyone to use and change, and free for other companies to commercialize on. We see this open-source approach enabling robotics innovation, and helping to ensure that the adoption of robotic technologies is a transparent process with positive societal impact.


Of course, even with the best software in the world, $400,000 is a little steep for a consumer-priced personal robot.  This is where advances in ultra low-cost planar laser rangefinders and ultra low-cost depth cameras (aka ranging cameras or RGB-D cameras) come to play. Devices that used to cost $10,000+ are coming on the market in the next 6-9 months for two orders of magnitude less (around $100 retail).

Ultra Low-Cost Planar Laser Rangefinder from Neato Robotics  Ultra Low-Cost Depth Camera (RGB-D Camera)


Coupled with inexpensive computing, pervasive cellular data networks, low-cost LCD screens, and the reduced cost of arms / manipulators through mass production... are we approaching a knee in the personal robot hockey stick curve?  

Hockey Stick Curve for Robot Adoption

I'd like to think so.  Regardless, I'm optimistic and excited.  Oh, and did I mention that my lab (Georgia Tech's Healthcare Robotics Lab, led by my advisor Dr. Charlie Kemp) is receiving one of the PR2 Betas?!?  I could be playing with a PR2 later this summer!



As if Stanford, Tokyo, USC or GT or really any of the others need more robots... Way to spread the wealth around Willow. I expect some of these bots to sit in corners while the rest of us make due with less

@ Anonymous,

I can certainly understand your criticism.  A few things are worth noting:

  • Willow Garage is a private company and will generally make decisions in their own best interests. 
  • We are not privy to Willow's internal decision making process; I'm sure it was a very complex joint optimization.  There are numerous top robotics institutions not on the list (notably: CMU, Cornell, Brown, etc).
  • An alternative, the DARPA ARM program for autonomous robotic manipulation (discussed here), will actually benefit an even smaller number of institutions and remain (presumably) more closed-source.
  • Willow is going to great lengths to make their robots available.  With eleven additional robots being distributed, odds are that there is one not far away.  I recommend reaching out -- most labs would welcome collaboration and / or interested students.  It is never too late to get involved.
  • ROS is platform agnostic.  We use ROS on almost all of our robots, and I've even heard that it works on the $150 Beagle Board.  This software suite is easily worth as much as the PR2's and is available to everyone.  Try it; use it; contribute.

—Travis Deyle


The best I can do to address your concerns is point to this post:


This is as much about open source software as it is robots, and the contributions of each of these instiutions will benefit the community as a whole.


$400k a robot is still the main problem. A few freebies is just a nice marketting ploy, unless you happen to be one of the 11 institutions that get one. Whilst the software is free, there is little motivation to use a simulated PR2 in research when you know you can never afford the hardware.

What would be really good right now would be a collective open project to put together a platform with similar abilities (eg supporting the same sofware stacks), but with a more grant-friendly price tag.

 Out of interest, does anyone know who the robot top left in the block of 14 photos is made by?


I agree with your point about running simulations for the PR2.  However, our lab (Healthcare Robotics)  controls everything from Roombas to our 7 dof mobile manipulator Cody with ROS and we find it very useful.  

I think one of the major problems is that hardware that is as good as the PR2 or Meka robotics (which is where my lab's 7 dof arms come from) is still not cheap.  Sensors are coming down in price but good robotic arms or omnidirectional mobile bases for example aren't getting any cheaper yet. 

In answer to your question, I believe the robot in the upper left is from TUM,

—Marc Killpack

Yes, the upper left is TUM's "Rosie":


I think we all want cheaper robot hardware. First, though, you have to prove the hardware is capable of doing valuable tasks. ROS enables research groups to coordinate their efforts on this front. 11 PR2 enables 11 groups to coordinate even more tightly.

"PR2" code was used to help port the Care-O-bot to ROS, including simulator-related code. The simulator is actually Gazebo, which is a robot-generic simulation environment. Regardless of whether or not you will have a PR2, having good simulations of mobile manipulation platforms enables groups to do research on advanced hardware while the market works towards making the hardware more affordable.