W3: RSS Workshop on Robots in Clutter: Manipulation, Perception and Navigation in Human Environments
Robots operating in our homes will inevitably be confronted with scenes that are simultaneously congested, unorganized, diverse and complex - or, simply put, cluttered. Clutter is a universal problem and severely affects all robot operations: manipulation, perception, navigation, and sensing. This makes it extremely difficult for a single approach to effectively handle clutter, perhaps explaining why robots (and robotics researchers) often shy away from it.
This workshop aims to bring researchers from different domains together and promote a discussion about clutter. This will contribute to robotics research in at least two ways. First, it will be a venue for the exchange of strategies, ideas, and algorithms used by individual domains. Second, it will provide an opportunity to discuss system-level approaches where manipulation, perception, and navigation work together. We hope to explore directions that will accelerate the deployment of robots into real human settings performing useful tasks even in the presence of clutter.
- Manipulation: Prof. Charlie Kemp, Georgia Institute of Technology: "Manipulation in Clutter with Whole-Arm Tactile Sensing"
Abstract: Clutter creates challenges for robot manipulation, including a lack of non-contact trajectories and reduced visibility for line-of-sight sensors. By using whole-arm tactile sensing, a robot can take advantage of contact across its entire arm to better perceive clutter and maneuver within it. In this talk, I will present research from my lab that has enabled robots to successfully reach to locations in high clutter while keeping contact forces low. Our main contribution is a novel controller that assumes low-force contact is benign. At each time step, it minimizes the predicted distance to a goal location subject to constraints on the predicted contact forces. Using this controller, our robots have performed complex maneuvers that involve bending objects, compressing objects, sliding objects, and pivoting around objects. The robots have reached target locations in a variety of environments, including artificial foliage, a cinder block, shelves, and randomly generated clutter fields. In simulation, whole-arm tactile sensing outperformed force-torque sensing at the joints, with the relative benefits increasing with the amount of clutter. Most recently, a person with quadriplegia commanded a robot to perform tasks that involved contact with his body while he was in bed. Based on our promising results to date, I will speculatively discuss the possibility for this approach to serve as a new foundation for robots that manipulate in clutter.
- Perception: Prof. Darius Burschka, Tecnische Universitat Munchen: "Spatial, Transient, and Temporal Clutter in Human Environments"
Abstract: Knowledge about the structure and motion parameters of objects in the scene is essential for safety and planning purposes in robotic applications. While there exist multiple solutions for perception in mobile and manipulation systems in static scenes with well separated objects, the dynamic aspect of the scene and the possible clutter with occlusions and inherent stability problems are usually neglected.
I will address the problem of visual detection of independent motion parameters for objects in the scene. The approaches are based on monocular and binocular camera systems detecting the independent motion parameters in real-time at up to 120Hz. I will give an overview over some current object detection and localization techniques in cluttered scenes that can be used for manipulation purposes. These techniques allow to complete the sensor information from the 2.5D in a camera or laser view to a complete 3D information with additional information about object properties. I will motivate how to use perception to parse a scene and to provide additional Atlas information, which is acquired a-priori.
Perception in medical environments adds additional challenges to the dynamic aspect of the scene, which is also of interest for difficult environments with poor structure and/or illumination. On one side, the observed scene in some parts of the human body, like e.g. abdomen, is highly dynamic and deformable during a surgical procedure. Additionally, the perception system faces the problem of a very poor image quality due to insufficient lighting, small lens aperture, and sparse texture. I will motivate how additional information from the image formation process, like e.g. shading and motion blur, can be used to enhance the perception of structure and motion in challenging environments.
- Navigation: Prof. John Leonard, Massachusetts Institute of Technology: "Navigation and Mapping in Clutter: Challenges and Opportunities"
Abstract: This talk will discuss the challenges and opportunities posed by navigation and mapping in cluttered environments. Operating a mobile robot in a cluttered environment poses a myriad of difficult questions for research. Questions include: data association, dynamic environments, robustness, and safety. Key goals for future research include the ability to achieve robust performance in the face of changing environments, and the ability to reason about complex physical interactions with the environment. An underlying issue relates to the resolution of environmental representations (maps and models) -- can we develop perception algorithms that achieve sufficient precision to localize objects in cluttered scenes, yet still provide robust, real-time performance? Despite these difficulties, clutter can provide a richness of input and a new perspective that, to some extent, can actually provide benefits to research progress in 3D perception. This talk will discuss these issues using examples from a variety of recent MIT research projects including long-term visual SLAM, Kinect Monte Carlo Localization, dense mapping with RGB-D cameras, and navigation and mapping in complex marine environments.
Joint work with: Maurice Fallon, Ross Finman, Franz Hover, Hordur Johannsson, Michael Kaess, John McDonald, David Rosen, Mark Van Middlesworth, and Tom Whelan.
- submission deadline: Wednesday, May 9th, 2012 --> extended to May 20th, 2012
- notification of acceptance: on or before Wednesday, May 30th, 2012
- workshop: mornings of Monday, July 9th and Tuesday, July 10th, 2012
Call for Papers
The organizers invite you to submit an approximately 4-6 page paper for review to the RSS Workshop on Robots in Clutter: Manipulation, Perception, and Navigation in Human Environments. This workshop will span two mornings during the main RSS conference, and will consist of invited talks, contributed paper talks, and a contributed paper poster session. All accepted work will be published in a citable digital archive of the proceedings.
Contact the workshop organizers via email at firstname.lastname@example.org.