Overview

Speaking with unconstrained natural language is an intuitive and flexible way for humans to command robots to perform spatial tasks such as navigation or manipulation. However, command-and-control interfaces are problematic because the human instructor may give ill-specified commands which cannot be grounded in the robot's perceptions, and which could refer to anything from small modifications of an existing plan to the creation of entirely new plans. Conversely, a robot might need to provide guidance to a human user in order to perform physical tasks. Dialog interfaces could address these problems by establishing common ground between the person and the robot, enabling natural and flexible interaction. Improved sensing algorithms are enablers for more powerful dialog interfaces, providing essential contextual information required to ground the language.

This workshop specifically aims to bring together communities studying perception and natural language understanding to build powerful spatial dialog systems for robots. We aim to bridge the gap between the theoretical understanding and practical application of spatial language understanding, create and discuss shared datasets and problems, and define key research problems and challenges. The following topics are of potential interest:

Call for Papers

We invite submissions of 4 pages in RSS 2011 format on research towards building spatial dialog interfaces for robots. Topics include, but are not limited to the following:

Dialog interfaces for spatial reasoning tasks, such as navigation, object finding, object manipulation and delivery.
Challenges inherent to multi-modal interaction for commanding robots with spatial semantics including speech, gesture and sketch.
Interpreting and generating route instructions for spatial tasks.
Learning meanings for spatial words such as “past, “on," “right," etc.
Vision and other sensing modalities for grounding linguistic elements such as places, objects, spatial relations, and verbs.
Knowledge representations that support spatial reasoning.
Challenges for effective human-robot interaction with spatial dialog.
Submit your manuscript by email to sl-workshop@lists.csail.mit.edu with “[RSS 2011 WORKSHOP]” in the subject line.