Direkt zum Inhalt springen
Intelligent Autonomous Systems
TUM School of Computation, Information and Technology
Technical University of Munich

Technical University of Munich



IROS 2011 Workshop: Knowledge Representation for Autonomous Robots

Over the past years, computers have made significant progress in mining knowledge from the web, aggregating information from different sources and using this knowledge to answer complex queries. Watson's recent victory over human jeopardy players impressively shows that very complex tasks in terms of knowledge acquisition, question answering, and natural language understanding can be solved nowadays.

So far, this technology has not been applied much in the robotics domain. We do, however, see a strong trend towards using semantic information, for instance in form of semantic maps, and believe that semantics will become more and more important. Robots will soon have to perform household tasks like cleaning up, setting a table or cooking simple meals. These tasks are extremely knowledge-intensive: To competently perform them, a robot needs a large amount of knowledge about properties of objects, actions required for a task, or execution problems that can arise.

In this workshop, we invite researchers from both robotics and knowledge representation to discuss the state of the art in robot knowledge processing, examine how formally represented knowledge can help robots in performing their tasks, and identify major research challenges that need to be addressed.

Program and Venue

The workshop takes place on Sunday, September 25, 2011, at the IROS conference site at the Hilton San Francisco Union Square, Continental Ball and Parlor room #7. More information can be found on the IROS conference website.

The workshop program and proceedings can be found here.

Intended audience

The primary audience are researchers working on knowledge-based systems in a robotics context. In addition, we want to bring together researchers in perception, manipulation, planning, and human-robot interaction who would like to include more semantic information into their systems.

List of Topics

  • Knowledge representation for robots: Which kinds of knowledge are required? Which representation formalisms are suitable for being used on autonomous robots? Which aspects (e.g. spatio-temporal reasoning, changes in objects and the environment over time) need to be modeled, and how can they be expressed in the chosen formalism?
  • Grounding and anchoring: Practical methods for grounding abstract symbols in percepts and actions, including the selection of the right object to be used for a task among multiple alternatives.
  • Hybrid Reasoning: AI methods typically focus on discrete, symbolic knowledge, but a robot also needs to reason about continuous, non-symbolic entities like time, geometry, and resources. How should the two types of reasoning be combined? Examples include hybrid task and motion planning, and combined planning and scheduling.
  • Knowledge acquisition: How to acquire the large amounts of knowledge required to competently act in human environments? Can web resources help robots with knowledge acquisition?
  • Knowledge exchange between robots: Knowledge acquisition and learning are complex and time-consuming tasks – can robots profit from sharing knowledge? Which kinds of knowledge can be exchanged, and how do they have to be processed to be used by a different robot?
  • Human-robot interaction: How can knowledge help robots to communicate with humans and understand dialogs? Which methods are required for disambiguating dialog situations and grounding words in actions and perceived objects?
  • Knowledge and perception: How can robots relate percepts to their background knowledge, and how can they use knowledge to improve perception?
  • Living with inconsistent knowledge: A robot's knowledge about the environment may become wrong or inconsistent due to sensor errors, outdated information or inappropriate knowledge exchange. How can the knowledge base handle this situation and still derive useful results from the remaining knowledge? Can these inconsistencies be detected and resolved?


We invite papers of 4-6 pages in the standard IROS format. Submissions should describe clearly the problems to which knowledge processing techniques are applied, explain the methods that are being used, and give an outlook on challenges that need to be solved in the future.

Besides technical quality, the submissions will be judged by their novelty, their potential to generate discussion, and their ability to foster collaboration within the community.

Questions and submissions should be sent to Moritz Tenorth, tenorth-at-cs.tum.edu

Important Dates

  • Paper submission deadline: 25 May 2011
  • Notification of acceptance: 20 June 2011 (delayed to 22 June)
  • Submission of final papers: 3 July 2011 (extended due to late notification, HARD DEADLINE)

Organizing Committee

  • Michael Beetz, TU Munich
  • Rachid Alami, LAAS-CNRS, Toulouse
  • Joachim Hertzberg, Universitaet Osnabrueck
  • Alessandro Saffiotti, Orebro University
  • Moritz Tenorth, TU Munich

Rechte Seite