[ros-users] Fuerte SIG 2.10: Efficient Simulation/Control of…

Top Page
Attachments:
Message as email
+ (text/plain)
Delete this message
Reply to this message
Author: User discussions
Date:  
To: ros-users
CC: Liu, C. Karen
Old-Topics: [ros-users] Fuerte SIG 2.15: Efficient IPC for Robust Modularity
Subject: [ros-users] Fuerte SIG 2.10: Efficient Simulation/Control of Articulated Robots
Dear ROS Users,

In a continued effort to contribute robust and useful tools to the ROS
community, please check out ROS Fuerte Planning SIG 2.10. and consider
joining the effort if you agree that it will be valuable to the community.

SIG 2.10: Efficient Simulation and Control of Articulated Robots
----------------------------------------------------------------
The goal is to develop and integrate a unified dynamic simulation/control
infrastructure with efficient representation, efficient computation of
articulated body dynamics in generalized coordinates, contact handling and
its link to planning, trajectory generation and rapid motion.

Currently ROS users link numerous data structures and packages to achieve
dynamic simulation for control and controller validation. For instance,
URDF data structures are copied to KDL ones to use kinematic solvers and
then again to ODE or Bullet for dynamic simulation and contact resolution.
This can be computationally inefficient and difficult to master for
end-users. Furthermore, the use of ODE/Bullet for articulated body
simulation can also cause instability and inefficiency. These simulators
are designed to handle individual bodies with iterative (penalty)
resolution of constraints. This SIG would focus on the development of a
single data structure that can handle arbitrary rigid and articulated
bodies through featherstone-like dynamics solvers for handling articulated
dynamics in generalized (minimal) coordinates.

We also aim to have the SIG consider broadening the library with a suite
of algorithms based on forward and inverse kinematics and dynamics such as
robot control, animation and planning.

Prof. Karen Liu and I of the Graphics and Humanoid Robotics Labs at GT are
already well into the development of a data structure and simulator that
satisfies these needs. The missing link was collision detection and
conveniently, Jia Pan, Sachin and Dinesh Manocha have just created an
efficient BSD collision detection library in ROS! We are now using this
library to implement LCP-based contact handling which will complete the
first beta of the new simulator to be released open source (BSD) to the
ROS community in a few weeks.

The goal of this SIG would be to jointly develop this library, integrate
it with Gazebo simulation as well as motion planners, controllers and
optimal motion generation algorithms. We think this will be a fantastic
resource that will spur collaboration between different groups that
require a single highly efficient solution to robot kinematics and
dynamics including contact between robots, objects and enviroments.

If you are interested, please shoot us an email:
(, )

And sign up for SIG 2.10 at:
http://www.ros.org/wiki/fuerte/Planning

Best,

Mike

---------------------------------
Mike Stilman
Assistant Professor
Director: Humanoid Robotics Lab
Robotics and Intelligent Machines
Georgia Institute of Technology
---------------------------------
email:
cell: (650)283-4284