Home

From rss2012

Revision as of 21:56, 1 August 2012 by MihailPivtoraiko (Talk | contribs)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

Banner.jpg


The RSS 2012 Mobile Manipulation workshop in Sydney, Australia

The aim of this proposed workshop is to discuss the state of the art in mobile manipulation research. Robust, reliable mobile manipulation is critical for robotics applications in the home, health care and retail industries. The workshop will focus on research at the intersection of motion generation and manipulation contact.

The workshop's schedule will feature invited talks and contributed papers.


Hope to see you in Sydney!


Mihail Pivtoraiko, Dov Katz, Oliver Brock


Discussion Results

Here are some notes from the discussion at the workshop.


Aspects of Contact

Perception: use environment for accurate relative sensing and to improve localization of robot and objects.

Compensate for actuation limitations of robot: use the environment for support.

Specialized strategies that exploit contact: e.g. push grasping or other special grasping strategies.


Difficulties of Contact

The whole world is the robot: because the robot can move objects, the motion must be determined in the combined configuration space.

Contact constraints lead to accessible configuration space of “measure zero”.

How to deal with uncertainty becomes crucial: minimal displacements are the difference between zero and infinite forces.


Characteristic Dimensions of Solutions

Uncertainty:

What do we mean by it? Representation of uncertainty or internal representation of the world? Prediction of uncertain outcome of actions? Both?

System building:

Plan-Sense-Act? Pipeline? Couple planning, sensing, action tightly? Don't even distinguish between them? Should hardware be considered explicitly, i.e. hard- and software co-development?

Planning/Control:

Is planning for full uncertainty with POMDPs feasible? How should planning and control be combined? What role can motion primitives play?

Learning/Hand-tuning:

Where is the boundary between motion capabilities that are best hand-tuned or learned or planned?


Instances of contact

Grasping

Are special cases of the grasping problem easy?

Dynamics

Sliding contact appears beneficial in some applications, such as opening spring-loaded doors. It may be useful to consider this kind of contact during motion generation.