ICRA 2011 Workshop on Manipulation Under Uncertainty

Quick jump: [Schedule] [Papers] [Invited talks] [Description] [Contact] [Organizers]

Schedule

Time: Friday, May 13th, full-day. Location: Room 5H.

8:45 Welcome
9:00 Invited talk: Prof. Kamal Gupta, Simon Fraser University. "Motion Planning with Localization and Map Uncertainties for Mobile Manipulation"
9:40 Invited talk: Prof. Jeff Trinkle, Rensselear Polytechnic Institute. "The Application of Particle Filtering to Grasping Acquisition with Visual Occlusion and Tactile Sensing"
10:20  Invited talk: Prof. Matt Mason, Carnegie Mellon University. "Grasp Invariance"
11:00 Podium poster introductions for Session 1
11:15 Contributed papers: Poster Session 1
12:00 Lunch break
1:30 Invited talk: Prof. Joris De Schutter, K.U. Leuven. "A Model-Based Approach for Manipulation under Uncertainty"
2:10 Invited talk: Prof. Aaron Dollar, Yale University. "Underactuated Hand Mechanisms Allow for Passive Adaptability to Uncertainty"
2:50 Invited talk: Prof. Pieter Abbeel, U.C. Berkeley. "Towards Robotic Laundry"
3:30 Podium poster instroductions for Session 2
3:45 Contributed papers: Poster Session 2
4:30 Plenary discussion
5:00 Wrap-up

Contributed Papers

Invited Talks

  • Prof. Aaron Dollar, Yale University. "Underactuated Hand Mechanisms Allow for Passive Adaptability to Uncertainty"

    Abstract: The uncertainty associated with operating in unstructured, “real-world” environments, where object properties are not known a priori and sensing is prone to error, makes precise positioning and controlling contact forces difficult. As a result, unintended contact happens frequently, resulting in large impact forces that can damage target objects and/or robot hardware. In this talk I will detail our recent and ongoing work related to the development of robotic hands for operation in these environments. These efforts seek to effectively implement adaptive underactuation and passive compliance the mechanical structure of the robot hands. I will present basic studies related to understanding and optimizing the performance of these mechanisms, as well as implementation for a wide range of applications, including robotic and prosthetic grasping, dexterous within-hand manipulation, as well as manipulation from small aerial vehicles.

  • Prof. Kamal Gupta, Simon Fraser University. "Motion Planning with Localization and Map Uncertainties for Mobile Manipulation"

    Abstract: We address the motion planning problem in robotic exploration, inspection and manipulation tasks, where robot localization and mapping uncertainties have to be incorporated into the planned motions. The robot considered in this work is a mobile-manipulator system, which combines mobility of the base with dexterousness of the manipulator. We consider a specific sub-problem where the manipulator is required to move to a given goal configuration (say, to carry out an inspection or a manipulation task), with the base staying stationary. We explicitly consider the uncertainty in the base pose that arises due to localization error in its estimation, and call this problem Manipulator Path Planning with Base Pose Uncertainty. We extend the sampling based planning framework, specifically the probabilistic roadmap method (PRM) to address this problem. Because of the uncertainty, a path for the manipulator is associated with a collision probability, which fundamentally changes the nature of the PRM's query phase. For a given roadmap, we search for a shortest path such that the collision probability of the manipulator is lower than a user defined threshold, were the manipulator to follow the path. We call our path query problem collision probability constrained shortest path problem (CPC-SPP). Using a particle based representation for the uncertainty, we show that CPC-SPP is NP-hard w.r.t. the number of the particles. We then present a lazy query algorithm, called CPC Lazy PRM (collision probability constrained Lazy PRM), based on a k-shortest path algorithm in conjunction with a label setting algorithm. CPC Lazy PRM exploits the definition of domination in the label setting algorithm to prune out those portion paths that cannot be the solution path. This leads to significant efficiency gains in practice. Although, worst case complexity is exponential in the number of particles, we empirically show the effectiveness of our query algorithm for simulated on-board manipulators of 3-dof and 6-dof, in planar and 3D environments, respectively.

  • Prof. Pieter Abbeel, U.C. Berkeley. "Towards Robotic Laundry"

    Abstract: Due to their inherently high-dimensional configuration spaces, non-rigid objects pose a number of difficult challenges. This difficulty is exemplified by the state of the art in robotic laundry folding. In this talk I will present our work which has enabled for a range of articles to have a robot start from a crumpled, unknown article and, through manipulation and perception, first get the article into a known configuration, and then fold it.

  • Prof. Jeff Trinkle, Rensselear Polytechnic Institute. "The Application of Particle Filtering to Grasping Acquisition with Visual Occlusion and Tactile Sensing"

    Abstract: Advanced grasp control algorithms could benefit greatly from accurate tracking of the object as well as an accurate all-around knowledge of the system when the robot attempts a grasp. This motivates our study of the G-SL(AM)2 problem, in which two goals are simultaneously pursued: object tracking relative to the hand and estimation of parameters of the dynamic model. We view G-SL(AM)2 problem as a filtering problem. Because of stick-slip friction and collisions between the object and hand, suitable dynamic models exhibit strong nonlinearities and jump discontinuities. This fact makes Kalman filters (which assume linearity) and extended Kalman filters (which assume differentiability) inapplicable, and leads us develop a particle filter. An important practical problem that arises during grasping is occlusion of the view of the object by the robot’s hand. To combat the resulting loss of visual tracking fidelity, we designed a particle filter that incorporates tactile sensor data. The filter is evaluated off-line with data gathered in advance from grasp acquisition experiments conducted with a planar test rig. The results show that our particle filter performs quite well, especially during periods of visual occlusion, in which it is much better than the same filter without tactile data.

  • Prof. Matt Mason, Carnegie Mellon University (joint work with Alberto Rodriguez). "Grasp Invariance"

    Abstract: This talk explores the role of finger shape in adapting to variations in object shape and pose. For very complex hands, numerous actuated joints are used to adapt to varying object shape and pose. But for simpler hands, with fewer fingers, fewer freedoms, and fewer actuators, gracefully adapting to shape and pose variations may fall on the finger form. In this work we explore grasp invariance over shape and/or pose variation as a principle for finger form design. We show how, under certain conditions, the problem can be mathematically formulated and admits a unique solution as an integral curve of a vector field. Under specific circumstances, the principle gives rise to spiral shaped fingers, including logarithmic spirals and straight lines as special cases. We apply the technique to derive scale-invariant and pose-invariant fingers to grasp disks, and we also explore the principle's application to derive optimal shapes for many common devices from jar wrenches to rock-climbing cams.

  • Prof. Joris De Schutter, K.U. Leuven, "A Model-Based Approach for Manipulation under Uncertainty"

    Abstract: This talk reviews a systematic approach to specify tasks of general sensor-based robot systems consisting of rigid links and joints that has been introduced in [1]. The approach integrates constraint-based task specification with instanta- neous control and an explicit model of geometric uncertainty in a unified framework. The approach uses feature coordinates, defined with respect to object and feature frames, to facili- tate the constraint-based task specification, and uncertainty coordinates to model geometric uncertainty. From the task specification controllers are derived to execute the task. Both velocity-resolved and acceleration-resolved controllers may be designed. In addition, the explicit uncertainty modeling allows us to configure estimators, such as Kalman filters or particle filters, to obtain the unknown uncertainty coordinates and their time derivatives. While the estimates of the uncertainty coordinates are used to adapt the geometric task model, their first or second time derivatives are added to the output of the velocity- or acceleration-based control scheme, respectively, to compensate for time-varying geometric uncertainty. The approach is illustrated with several examples. Finally, a short overview of recent work is given, including extension to inequality constraints [2], optimization-based esti- mation of uncertain parameters [3], and an approach for global task optimization.

Description

This workshop will focus on robot manipulation strategies that are robust to the uncertainty inherent in unstructured environments. Recently, the problem of uncertainty has been addressed in different ways, including: identifying manipulation strategies that are robust in task-specific situations, designing manipulator hardware that makes classes of manipulation activities more robust, and proposing planning algorithms that explicitly model uncertain variables. Our goal is to better understand all of these approaches by fostering discussion amongst the involved researchers.

If you have original research relevant to the area of manipulation under uncertainty, we would be very interested in receiving your submission. Our interests include, but are not limited to, the following topics in the context of manipulation:

  • explicitly modeling and/or planning under uncertainty (including general planning algorithms that could be used in the context of manipulation, even if they are not explicitly designed for such)
  • strategies for implicitly reducing uncertainty in specific, manipulation scenarios
  • manipulator designs that effectively handle uncertainty
  • new sensor technologies and their applications
  • other methods and techniques for manipulation in real-world scenarios affected by variability and uncertainty

Call for Papers

The organizers invite you to submit a paper for review to the ICRA 2011 Workshop on Manipulation Under Uncertainty. This full-day workshop will consist of six invited speakers and two contributed paper sessions. In order to encourage discussion and exchange of ideas between particpants, all contributed papers will be presented in a poster session format, prefaced by short overviews from the podium. All accepted work will be published in a citable digital archive of the proceedings.

Important Dates

  • submission deadline: March 1st, 2011
  • notification of acceptance: on or before March 20th, 2011
  • workshop: May 13th, 2011

Contact

Contact the workshop organizers via email at muu11@willowgarage.com 

Organizers