KI 2010 - Home


Call for Papers

Call for Workshop Proposals and Tutorials

Call for Special Session Situation, Intention and Action Recognition Papers


Important Dates

Author Area








Special Session at KI 2010 on
Situation, Intention, and Action Recognition

KI 2010

33rd Annual German Conference on Artificial Intelligence (KI 2010)

September 21-24, 2010
Karlsruhe, Germany

General Co-Chairs

Rüdiger Dillmann
Jürgen Beyerer


Program Chairs

Tanja Schultz
Uwe D. Hanebeck


Publicity Chairs

Alex Waibel
Rudi Studer


Exhibition Chair

Heinz Wörn


Local Chair

Fernando Puente León


Workshop and Tutorial Chairs

Marius Zöllner
Rainer Stiefelhagen



KI 2010 - Team



Almost all technical devices and software agents surrounding humans attempt to recognize the behavior of the user. With the embedding of an increasing number of sensors and the spread of mobile sensor devices, e.g., Humanoid Robots, the need for theoretically well-grounded and systematic approaches becomes eminent.
Situation, Intention, and Action Recognition relates to the process of inferring the user’s behavior on differing scales and scopes. The situation, defined as a set of conditions, limits the user’s current and future behavior. Within the scope of the prevailing situation, the user will have various coarse intentions that manifest in fine-grained actions, i.e., manipulations of the world. Recognizing the user’s behavior with respect to these differing levels of abstraction may be used for the facilitation of human-machine-interactions or the recognition of exceptional behavior. Potential applications range from software usability to human-robot-cooperation for household-robots.
The main challenges for the recognition tasks are the fusion of online incoming data (e.g., video streams) with appropriate domain knowledge, e.g., relational spatio-temporal information (i.e., object relations, human motion models or ontologies). The inference is most often based on uncertain and hardly sufficient data (e.g., human action recognition based on video signals) increasing the need for appropriate models of the human and the world. The models necessary for non-trivial tasks entail complex hierarchical models, spanning atomic actions to multilevel action sequences. Finally, for many realistic scenarios, the situation can only be assessed when taking the multi-agent context into account, i.e., the relation of the human/user surrounded by other humans/users as well as robots.

Contributions are invited but not limited to the following research fields and applications:

  • Representations of user/human behavior
  • Recognition of Situation
  • Recognition of Intention
  • Recognition of Action
  • Machine learning for probabilistic inference
  • Modeling and monitoring (multiple) agents
  • User Modeling and its Applications
  • Behavior Recognition for Human-machine-interaction
  • Action Recognition for Programming-by-Demonstration and Imitation Learning

Submission Guidelines:

Submitted papers, which have to be in English, should regularly not exceed 8 pages in Springer LNCS style, two additional pages will be permitted for an extra charge. Conference submission is electronic, in pdf format. The paper will be treated as a regular contribution to the conference, i.e., will be part of the regular review process and will be published as part of conference proceedings in the Springer Lecture Notes in AI (LNAI) series. At least one author per accepted paper must register for the conference and present the contribution. Papers have to be submitted before 16 April 2010 and should be marked as “special session papers” during submission.
The conference is supported by the "Gesellschaft für Informatik" (GI).