See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/254057457

Moveit![ROS topics]

Article  in  IEEE Robotics & Automation Magazine · March 2012

DOI: 10.1109/MRA.2011.2181749

CITATIONS READS

138 4,282

3 authors, including:

Sachin Chitta Ioan Alexandru Sucan

SRI International Google Inc.
67 PUBLICATIONS   3,113 CITATIONS    26 PUBLICATIONS   1,454 CITATIONS   

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Sachin Chitta on 28 November 2015.

The user has requested enhancement of the downloaded file.

 ROS TOPICS •

MoveIt!
By Sachin Chitta, Ioan Sucan, and Steve Cousins

R
obots are increasingly finding motion planning, trajectory generation, robot. A custom analytical kinematics
applications in domains where and environment monitoring for the solver for the PR2 allowed for fast inverse
they have to work in close robot arms of the PR2 robot. Sensor kinematics solutions. An execution com-
proximity to humans. Indus- data from the laser and stereo sensors ponent called move_arm formed the
trial robotic applications are starting to on the PR2 robot was fused to generate primary interface to this set of capabil-
examine the possibility of robots and a consistent model of the environment. ities, exposing an interface that allowed
humans as coworkers, sharing tasks The environment model can handle users to specify Cartesian and joint space
and workspace. Autonomous robotic occlusions from parts of the robot’s goals that the robot needed to reach. A
cars operating on crowded streets and body and also the significant noise in trajectory monitoring component kept
freeways have to share space with 3-D sensor data. The environment was an eye on the motion of the robot,
pedestrians and cyclists in addition to represented as a mixture of two formats: actively servoing the sensors if necessary,
other vehicles. Domestic robots, in 1) a voxel grid that represented most of to track the motion of the arm, and
particular mobile manipulation sys- the obstacles in the environment and respond to changes in the environment.
tems, will be confronted with cluttered, 2) geometric primitives and mesh mod- The arm_navigation software was a key
messy environments where obstacles els to represent objects that had been component of the grasping pipeline (sub-
exist at every corner, and people are recognized and registered in the envi- ject of a previous “ROS Topics” column),
continuously moving in and out of the ronment by object detection routines. allowing the generation of collision-free
workspace of the robots. The environment model was built upon trajectories to execute, pick, and place
Robots working in human environ- the Octomap package [1], an octree- actions, obeying the geometric con-
ments clearly have to be aware of their based probabilistic representation, straints imposed by the task.
surroundings and must actively attempt allowing for efficient representation of
to avoid collisions with humans and large cluttered scenes. The environment Evolution
other obstacles. MoveIt! is a set of soft- model served as the primary input to The evolution of the arm_navigation
ware packages integrated with the fast motion planners that could quickly software to MoveIt! is the result of an
Robot Operating System (ROS) and generate collision-free motion plans in effort to package the core algorithmic
designed specifically to provide such fairly cluttered environments. A generic capabilities in the arm_navigation
capabilities, especially for mobile ROS interface for geometric planning software separately from the middle-
manipulation. MoveIt! will allow robots allowed multiple motion planners to be ware (ROS), allowing for easier code
to build up a representation of their integrated, including randomized plan- reuse. Lightweight ROS bindings and
environment using data fused from ners from the Open Motion Planning wrappers will allow ROS users to
three-dimensional (3-D) and other sen- Library (OMPL) [2], search-based plan- easily configure and interface with the
sors, generate motion plans that effec- ners from the Search-Based Planning components of MoveIt!. Advanced
tively and safely move the robot around Library (SBPL) [3], and trajectory opti- users and application developers will
in the environment, and execute the mization libraries including CHOMP be able to directly incorporate the core
motion plans while constantly monitor- [4] and stochastic trajectory optimiza- capabilities in MoveIt! through libra-
ing the environment for changes. tion for motion planning (STOMP) [5]. ries without needing to depend on a
Some of the motion planners were capa- large part of ROS. MoveIt! builds on
Ground Work ble of dealing with geometric con- new features in ROS, including the
MoveIt! is an evolution of the arm_ nav- straints, e.g., an orientation constraint ability to use the ROS message gener-
igation software in ROS. The arm_navi- specifying that a glass of water should ation infrastructure as a standalone
gation packages were designed for stay upright. Path smoothers and trajec- capability. MoveIt! also integrates the
tory generators were used to parameter- Fast Collision Library (FCL) [6],
Digital Object Identifier 10.1109/MRA.2011.2181749 ize the planned paths so that smooth which provides distance-checking
Date of publication: 26 March 2012 trajectories could be executed on the capabilities that should allow for the

18 • IEEE ROBOTICS & AUTOMATION MAGAZINE • MARCH 2012

•
Manipulation [Online]. Available: http://
octomap.sf.net/
[2] I. A. Sucan, M. Moll, and L. Kavraki. (2010).
The Open Motion Planning Library (OMPL),
[Online]. Available: http://ompl.kavrakilab.org
[3] B. Cohen and M. Likhachev. (2009). The
Search Based Planning Library (SBPL) [Online].
Available: http://www.ros.org/wiki/sbpl
[4] N. Ratliff, M. Zucker, J. A. D. Bagnell, and
S. Srinivasa, “CHOMP: Gradient optimization
techniques for efficient motion planning,” in
Proc. Int. Conf. Robotics and Automation, May
2009, pp. 489–494.
[5] M. Kalakrishnan, S. Chitta, E. Theodorou,
P. Pastor, and S. Schaal, “STOMP: Stochastic
A snapshot of the PR2 robot using the motion planning capabilities in ROS to execute a trajectory optimization for motion planning,”
constrained motion plan as part of a pick-and-place task while keeping the object (a in Proc. Int. Conf. Robotics and Automation,
stapler) horizontal. The visualization to the right shows the environment representation, Shanghai, China, May 2011.
built from onboard sensors, used by the robot for motion planning.
[6] J. Pan and D. Manocha. (2011). The Fast
Collision Library (FCL) [Online]. Available:
generation of better-quality motion implementation required from the http://www.ros.org/wiki/fcl_ros.
plans. MoveIt! is also expected to con- user on the robot side. This allows
tribute in other application areas, nonmotion planning experts to easily
including automating the calibration configure the motion planning and
process for robotic arms with 3-D and associated components in ROS for
other visual sensors. The arm_naviga- their own robots.
tion software stack will retain its func-
tionality, but its implementation will The Future
consist primarily of arm-specific con- We expect MoveIt!, in concert with
figurations for the MoveIt! software. the motion planning capabilities in
OMPL and other packages, to form
Portability the basis for new and exciting applica-
The arm_navigation software was tions with robots. As MoveIt! evolves
primarily designed for the PR2 robot and finds more uses in the com-
but has been used on several other munity, new features and capabilities
robotic systems. In MoveIt!, we make will find their way into the system,
an additional effort to extract all the e.g., active sensing, anomalous event
needed information into the existing detection and response, and the
URDF (the standard robot description generation of more natural-looking
format in ROS) and newly defined paths. We also aim to develop a set of
Semantic Robot Description Format benchmarking tests around MoveIt!,
(SRDF) to avoid dependencies on spe- allowing for the comparison of mo-
cific robots. Furthermore, new tools tion planners and other components
such as the Arm Navigation Wizard, for tasks in real-world situations.
specifically designed to enable config- MoveIt! will be crucial in our goal of
uration of the arm_navigation soft- enabling safer, more capable robots
ware for other robotic manipulators that can function effectively in human
and systems, are being ported to environments.
MoveIt!. The Arm Navigation Wizard
uses a combination of URDF and References
SRDF to autoconfigure the arm_navi- [1] K. M. Wurm, A. Hornung, M. Bennewitz,
gation software for a robot. An C. Stachniss, and W. Burgard. (2010).
interactive graphical user interface OctoMap: A probabilistic flexible compact 3D
allows users to specify motion-plan map representation for robotic systems. In
requests for their new robots, with Proc. ICRA 2010 Workshop on Best Practice
a minimal amount of interface in 3D Perception and Modeling for Mobile

MARCH 2012 • IEEE ROBOTICS & AUTOMATION MAGAZINE • 19

View publication stats