IEEE Robotics & Automation Magazine - September 2011 - 16

ROS TOPICS

Robot Developer Kits
By Brian Gerkey and Ken Conley

t's a great time to be interested in
robotics. While robots have been
around for over half a century, they
have generally fallen into three categories: 1) application-specific devices
(e.g., welding, medication delivery, and
surgical assistance), 2) research instruments (e.g., Asimo, PR2, and HRP-2),
and 3) toys (too many to name). We're
now witnessing the debut of an entirely
new class: the robot developer kit.
The robot developer kit falls in the
middle of the three traditional robot
types: it's more flexible than an application-specific device, less expensive than
a research instrument, and more capable
than a toy. The goal of a robot developer
kit is to provide creative people with an
integrated hardware and software platform on which they can prototype their
ideas for revolutionary robotic applications. These kits provide developers with
key robotic capabilities such as mapping,
navigation, two-dimensional (2-D) and
three-dimensional (3-D) perception algorithms, and diagnostics. They also
provide developers with low-cost hardware that is modular and extensible.
The class of robot developer kits is
filling out quickly. Willow Garage's
TurtleBot combines an iRobot Create,
a Microsoft Kinect, and a netbook in a
low-cost mobile-sensing platform. The
Bilibot Project starts with similar parts
and adds a simple gripper. Tim Payne's
POLYRO goes further, starting with a
TurtleBot base and adding two servo
arms, a torso, and a sensor head with
two cameras. Many in the community,
from hobbyists to educators, have built
their own variations on these robots.
These kits make up for the lack of polish with extensibility for developers who
are looking for the VisiCalc of personal

Digital Object Identifier 10.1109/MRA.2011.942483
Date of publication: 13 September 2011

IEEE ROBOTICS & AUTOMATION MAGAZINE

robotics. Robotic vacuum cleaners have
proven themselves to be useful application-specific robots for the home, but we
have not yet made the transition to a general-purpose robot platform that can be
upgraded with new applications in the
future. Although these general-purpose
devices are harder to design, it has been
a successful model for providing increasing value to consumers over time.

The TurtleBot combines perception,
mobility, and computation in a low-cost
platform, designed for developers. The
TurtleBot hardware is available under an
open-source license, free for anyone to
reproduce, modify, and sell.

There remains the question of what
is the right hardware configuration for
these general-purpose devices, but the
extensibility of the robot developer kits
provides developers with the freedom
to address this challenge. Developers
already have a vast array of affordable
consumer electronic devices they can
integrate, such as picoprojectors, smartphones, tablets, and fourth-generation
(4-G) modems. These devices are also
driving down the cost of underlying
components, such as cameras and
touch screens. All of this makes it easy
for the developers to extend hardware
capabilities with new sensing, display,
connectivity, and computation options.
A key enabler of these new robot
developer kits is the availability of low-

SEPTEMBER 2011

cost 3-D sensors, such as the Microsoft
Kinect with PrimeSense technology.
Developers have long had access to inexpensive mobility, but just a year ago,
adding 3-D perception would increase
costs by an order of magnitude. Now that
the mobile 3-D perception kits can be
easily and inexpensively assembled, we
expect to see a new surge in applications.
Costwise, the hardware for sophisticated
manipulation remains a challenge, but
we believe that these mobile 3-D perception kits will bridge the gap to future platforms that incorporate a forthcoming
generation of low-cost arms.
Another key enabler is cloud computing. Battery life and computation, which
are cheap in the cloud, remain the key
cost concerns in developing affordable
robot hardware. The cloud also offers the
opportunity to seamlessly upgrade robot
platforms with new capabilities over time.
Google has demonstrated cloud-based,
crowd-sourcedobjectrecognitionforboth
TurtleBot and Hasbro Phondox platforms. This approach offers a cloud-sized
database of knowledge that is constantly
adapting to new objects in the real world.
A common element of all the robot
developer kits described in this article is
that they are integrated with robot operating systems (ROS) and have opensource software libraries. No matter
which one you choose-or build yourself-the shared capabilities of mobility
and 3-D perception enable the software
to be easily shared across multiple platforms. This means that the community
can work together to solve key platform challenges of performance,
robustness, power management, application infrastructure, and more, while
developers are free to focus on innovative new applications for robotics.
If you've been dreaming of an app
store for robotics, now is the time to
get involved.

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - September 2011