Willow Garage Blog

October 9, 2012

Jia Pan is a PhD student from the University of North Carolina Chapel Hill. This summer he worked on the FCL library, which is used in the MoveIt! project.

First, he improved the efficiency of collision checking and distance computation. FCL now includes specialized algorithms for collision between geometric primitives and additional broadphase collision algorithms that are amenable for both collision checking and distance computation between many objects. 

Secondly, for broadphase collision, Jia focused on the fast construction of broadphase structure. This is very important for collision with sensor data because slow construction of broadphase structures from sensor data received at high frame rates can severely reduce system performance. To reduce the construction overhead, a lazy construction strategy is used to balance the construction cost and collision speed. 

Finally, FCL can now perform collision checking with octree representations of the world. In the previous arm_navigation pipeline, regions that are unknown to the robot or are obstructed, are considered as free. This could cause problems when the robot performs actions in the unknown space. To address this issue, FCL now includes the ability to perform collision checking directly with an octree representation of the world.

The octomap library is used to construct an octree representation of the robot's environment. Each cell specified in the octree includes a probability of occupancy. FCL uses the octree structure directly, without the need for an expensive broadphase structure construction step. During collision checking with an octree, FCL determines both whether a collision occurs and a set of cost volumes. A cost volume is represented by the probability of collision, given by the probability of occupancy, and by the volume of the octree cell that incurs that probability of collision.

This capability enables robot to reason about unknown space as well. In particular, when a motion plan is computed, the cost of this plan is evaluated by looking at the intersection of a plan's swept volume and the unknown space. Using this information the robot can decide whether to execute the plan or to perform sensing operations: for example, pointing the head towards the unknown areas of the environment. This loop of planning and sensing additional parts of the environment can be performed multiple times, until the computed motion plan can be executed.

For more information please visit visit FCL library, and Moveit!

October 5, 2012

Clearpath Robotics aims to progress robotics development by providing universities with robust platforms for prototyping and research. To give robotics research an extra boost, Clearpath Robotics introduced the PartnerBot Grant Program this summer.


The PartnerBot Grant Program is a one-year commitment during which a prestigious research team will use the Clearpath Robotics Husky A200 to pursue its research goal, publish for public review, and add code to the rapidly growing open source ROS (Robot Operating System) community. Initially aimed at donating $25,000 worth of equipment, the sheer number of applicants and outstanding quality of the submissions prompted Clearpath Robotics to increase the total value to $100,000.

The Husky A200 is a rugged, all-terrain robotic platform for rapid prototyping and research applications. In the past we've seen it used for prototyping planetary rovers, researching autonomous navigation, environment mapping, and countless other applications. The Husky A200 integrates seamlessly with the free and open source ROS robotics framework that offers full control of the Husky, including (but definitely not limited to) autonomous navigation, perception, and mapping. Using the standardized platform and open software allows researchers to cooperate, share findings, and repeat experiments.


More than 150 universities world wide showed interest in the PartnerBot Grant Program. Clearpath Robotics has received submissions from every continent on the planet (except from Antarctica) with applications ranging from mine clearing, to agricultural robots, to planetary rovers, and everything in between. With such a great turnout and impressive applications it seemed a shame to support only one research project, so Clearpath Robotics has worked feverishly over the past month to scrape together enough money to support the following 10 distinguished research teams.

  • University of California, Santa Cruz
  • University of Coimbra
  • Orebro University
  • Universidad de Chile
  • Drexel University
  • Federal University of Minas Gerais
  • University of Hohenheim
  • University of Michigan
  • Queensland University of Technology
  • University of Bremen

Clearpath Robotics would like to thank everyone who invested time and energy to submit a proposal. The judging process was tough, selecting only one would have been impossible - even 10 was difficult. To read more about the outstanding research projects and PartnerBot recipients, visit the PartnerBot site.

September 30, 2012


During his internship at Willow Garage, Aaron Blasdel, a recent graduate from The University of Tokyo, developed software to model and analyze robotic grippers. 

Due to the broad and varied nature of the robot grasping and manipulation problem, new and interesting grippers are regularly constructed. Aaron constructed models of the Robotiq 3-Fingered Adaptive Gripper and various configurations of the Sandia Hand for the publicly available GraspIt! simulator. In the case of the Robotiq hand, the challenge was to model in simulation its mechanical adaptation capabilities. For the modular Sandia Hand, which uses fingers magnetically attached to a palm structure to form a virtually limitless number of possible designs, Aaron analyzed and compared a number of different configurations. For both hands, he then used the simulation engine to pre-compute grasps for a large set of common household objects. 

Simulation and optimization can help us better use existing hand models, and also find new points in the design space which can in turn lead to functional improvements in the future.

For more information see the Graspit! package, the Sandia Hand and the Robotiq 3-Fingered Adaptive Robot Gripper.


September 28, 2012

During his internship at Willow Garage Tobias Kunz, a PhD student from Georgia Tech, worked on a project to enable the PR2 to quickly transport objects on a tray without them falling off. This effort in conjunction with the sushi challenge aim to give robots the ability to perfrom tasks like efficient transport of every day objects and dynamic place setting. 

Dynamic trajectories are important for robots to execute delivery or butler tasks at home in a timely manner. The problem also appears in industrial robotics when objects are transported on tray-like end-effectors at high speeds. An identical problem is that of carrying an open container of liquid. Just keeping the tray horizontal is not enough to keep the object from sliding off when moving fast.

Instead, the tray needs to be tilted in a way that minimizes lateral force on the object.

Tobias implemented two different methods. The first method explicitly calculates the time-optimal trajectory for a free-floating tray along a straight line in workspace given constraints on the linear and angular accelerations of the tray. The second method considers arbitrary paths and the kinematic limitations of the robot arm. STOMP, a local trajectory optimization method, is used to iteratively minimize the lateral force acting on the object while also avoiding obstacles. The video shows the PR2 executing trajectories generated by this this method using different types of objects with different start and end locations for the tray.

Tobias’ work demonstrates the use of motion planning to enable robots to execute tasks with complex dynamic constraints. His goal is to get robots to do even more complex tasks in the future, just like humans can. 

You can download and playback these dynamic butler trajectories on your own PR2 robot. The necessary instructions are available here.


September 25, 2012

The business model at Willow Garage is built on the premise that we will spin off businesses when we see near-term market potential and when we see the opportunity to make a difference in people's lives.  We now have seven spin-offs, and while we love all our children equally, today belongs to Suitable Technologies.

Suitable today introduced Beam Remote Presence System, a mobile videoconferencing platform.  Beam lets people be in two places at once – literally.  The Beam is driven by its pilot from anywhere with an Internet connection and uses Beam as their remote presence at a different location.  The video below provides a good overview of how Beam will be used.


Beam is no robot.  To paraphrase Senator Lloyd Bentsen, we know robots, and Beam is no robot.  For example, there is no autonomy whatsoever with Beam.  Everything is controlled by the Beam pilot working remotely. 

We have, however, conducted a great deal of research on the implications of robots someday working alongside and physically sharing space with humans.  That research has been incorporated into the design of Beam in order to produce a solution that will effortlessly slip into the workplace.  For example, our research, (along with our personal experience) shows that it only takes about twenty minutes on average before a person forgets they are interacting with someone via Beam and begins to take for granted that they are simply having a conversation with a colleague. 

By now, the story of Suitable's beginnings are familiar to anyone that follows Willow Garage.  Dallas Goecker, a Willow Garage electrical engineer working remotely from Indiana, grew tired of being the lone voice on the Polycom during company meetings.  Skype also fell short because Dallas needed mobility to truly interact with his co-workers.  A DIY initiative known as Project Texai was created to fashion a mobile device that allowed Dallas to remotely control where and when he wanted to go within Willow's office.  Texai touched a nerve in the market, and even made a guest appearance as Shel-bot on the Big Bang Theory.   Ultimately, "Skype on a stick" fell short with respect to networking, navigation, and audio.  The team at Suitable then went back to the drawing board. 

Following 18 months of R&D, Beam offers state-of-the-art technology in a sleek RPD.  We wish them the best of luck as Beam hits the market.


September 24, 2012

During his internship at Willow Garage, Adam Zimmerman from the University of Illinois, created a version of Rviz for Android tablets. Rviz is a ROS visualization program used to help understand robot systems by providing a 3D environment that lets users see the world from a robot's perspective. This project was started in response to community demand for a mobile robot visualizer that ran on a tablet.

Rviz for Android uses a custom-made OpenGL rendering engine to display information received in ROS messages. The application was written in Java using the ROSJava API to publish and subscribe to ROS topics and transform between TF frames. The Rviz for Android interface is similar to the desktop Rviz interface, featuring a view panel and a list of enabled display types. The 3D display panel is manipulated using intuitive touch gestures. Many of the standard Rviz display types, such as robot model, point cloud, map, and interactive markers, are included. Instructions for adding new display types and camera controls are available online on the source code repository. A Web server, written in Python, runs alongside the ROS core to supply the tablet with meshes and textures referenced by URDF files and marker messages. This server can also create ROS nodes upon request from the tablet. These created nodes can be any ROS application, including message throttlers or compressors to more easily move large amounts of data to the tablet.

Because very little of the desktop Rviz code was portable to the Android platform, Rviz for Android is written almost entirely from scratch. Porting all of the desktop C++ Rviz code to Android using the Android NDK wasn’t a feasible solution for this project. The desktop Rviz code makes use of the Ogre3D graphics engine which has tenuous Android support. Even if Ogre3D was well supported on Android, a good deal of the Rviz code would need to be rewritten in Java to interface with ROSJava. Compiling roscpp natively on an Android device is outside the scope of the Android NDK and requires a good deal of functionality that it doesn't currently include. Another alternative considered was to use an HTML5 Web visualizer like wviz, but WebGL and Web sockets have extremely limited support on current iOS and Android devices. Rviz for Android lets users visualize and debug robot applications in environments where using a desktop or laptop computer is impossible.  It provides an intuitive touch interface and an easily extensible platform to meet the needs of any robot application.  The application supports any Android tablet running Android 3.2 or higher, and the source code, compiled application package, and tutorials for creating new display types are available online at https://bitbucket.org/zimmrmn3/rviz-for-android

September 13, 2012

Robots operating in our homes will inevitably be confronted with scenes that are congested, disorganized, and complex - or, simply put, cluttered. Consider, for example, a case where the robot must acquire an object from the back of a cluttered bookcase or refrigerator shelf. Reaching the target object from the top is impossible due to the constrained space inside the shelf; various obstacles block approaches from the front or side. Some of the obstacles are too large for the robot to grasp. The robot may even be uncertain about the poses of some objects.  How can the robot complete the task?

During his internship at Willow Garage, Mehmet Dogar from Carnegie Mellon University worked on extending a method called "push-grasping" for operation in cluttered environments. When push-grasping, rather than directly closing its gripper, the robot can first "push" the target object along a linear path and into the gripper, an action primitive that is very robust to uncertainty in the object's pose.

In our novel implementation, the robot can make simultaneous contact with multiple objects in a deliberate and controlled fashion. This enables the robot to reach for and grasp the target while simultaneously moving aside obstacles in order to clear a desired path. We use a physics-based analysis of pushing to compute the motion of each object in the scene in response to a set of possible robot motions.

Rather than shying away from complex and sustained interactions with the world while grasping, robots can use these interactions to their advantage. Our planner utilizes pushing interactions to address clutter and uncertainty. This means robots can work more effectively in highly cluttered environments and succeed even if there is uncertainty about the poses of objects.

For more information about the planner, head to the push_grasp_planner at ROS.org
September 11, 2012

Irene Rae from the University of Wisconsin-Madison spent the summer of 2012 exploring how the appearance of an embodied system affects the interaction between people using that system to communicate.

While embodied systems like the Texai can provide greater access to informal and opportunistic interactions to remote users, their physical traits such as height, color, or proportion may mediate the way that people who are local to the system might treat them.

The study looked at how changes as simple as modifying the height of the system may cause subtle shifts in the way that the local user behaves toward the remote user, possibly supporting or undermining the remote user’s authority and ability to persuade others.

Look for the final results in our upcoming publications!

September 6, 2012

If you have a robot that needs to move, MoveIt is the planning framework for you. MoveIt is an open source project that includes various planning techniques, kinematics, dynamics, collision checking, constraints evaluation and sampling, visualization, and more.  To easily enable MoveIt to run on your robot, Dave Coleman of the University of Colorado Boulder and Willow Garage intern developed the MoveIt Setup Assistant to get your robot quickly up and running.

MoveIt is a framework that connects a wide variety of planning techniques and collision-checking algorithms to useful tools such as benchmarking and visualization. The MoveIt Setup Assistant helps you generate all needed configuration and launch files, with an easy to use visual front.

Additionally, Dave contributed to the Open Motion Planning Library (OMPL) by implementing a new planning algorithm, Transition-based RRT, that keeps random tree search in low cost regions of the configuration state. This is convenient for minimizing objectives such as mechanical work over a trajectory, or taking probability of occupancy into account in an occupancy grid.

Finally, to aid OMPL development, Dave developed the ompl_rviz_viewer. The visualizer shows cost maps in RViz along with all searched states, nodes and edges from a planner. It also serves as a good code example for anyone hoping to get started with OMPL.

Dave’s work with MoveIt and OMPL this summer at Willow Garage contribute to the overall progress of robotic planning and its adoption by roboticists everywhere. 

For more information visit:

moveit.ros.org and ompl_rviz_viewer


August 29, 2012

hiDOF logo

Today marks another significant milestone in Willow Garage's mission to create a personal robotics industry.  Willow Garage is proud to announce the creation of its latest spin-out: hiDOF, Inc., a software consulting company. The name is a play on the fact that the founding team has extensive experience with high degree of freedom systems.

hiDOF will be applying robotics technology to commercial applications, and is founded by Willow Garage engineers who believe in the power of using open source software to provide increased value to customers. Over the last few years, hiDOF’s founders have authored and maintained many of the core ROS tools, and programmed the PR2 to do everything from recharging itself to fetching beer.  Now, with hiDOF, they are excited to leverage these tools to solve real-world problems, while providing the community with battle-tested code.

The hiDOF team is available for a variety of consulting and contracting services. If you have a difficult automation problem that requires greater flexibility, hiDOF can help. hiDOF also provides extensive technology transfer services. They can harden an algorithm, optimize an approach, or create an amazing technology demonstration.  

Redefining Automation

hiDOF is ready to take on difficult automation challenges that other's can't solve. hiDOF believes that industrial automation is on the cusp of a revolution in flexibility, robustness, and safety.

Leveraging ROS

The hiDOF team leverages ROS to solve complex problems quickly.  The founding team authored many of the core components of ROS, giving hiDOF a wealth of deep expertise and knowledge across the entire ROS ecosystem.  They can build everything from drivers and controllers all the way through high-level capabilities like navigation and manipulation.  

Harnessing the Cloud

hiDOF has experience building large, scalable cloud solutions. 

For more information, or if are interested in talking through a potential project, please contact the team at info@hidof.com. You can also visit their Web site at http://www.hidof.com.

Creating an Industry

Since Willow Garage was founded in 2006, there have now been seven spin-outs.  Given that our goal is to create a new robotics industry, we think seven new companies in six years is a pretty good start. Those companies are:

·Industrial Perception Inc. - IPI is working on integrating perception technology with industrial robots.

·OpenCV - An open source computer vision and machine learning software library built to provide a common infrastructure for computer vision applications and to accelerate the use of machine perception in the commercial products.

·Open Perception Foundation – Their mission is to advance the development and adoption of open source software for 2D/3D processing of sensory data, for the benefit of the industrial and research communities.

·Open Source Robotics Foundation - OSRF is an independent non-profit formed to support the development, distribution, and adoption of open source software for use in robotics research, education, and product development.

·Redwood Robotics - A joint venture between Meka Robotics, Willow Garage, and SRI.

·Suitable Technologies - Creating an innovative new remote presence product.