Important Dates
Call for Papers/WS
Paper Submission
Program Committee
Author Instructions
Invited Talks
Industrial Day
Accepted Papers
Travel Information

Invited Talks

Bernhard Nebel
Institut für Informatik
Universität Freiburg, Germany

Cooperative Sensing and Acting in Robotic Soccer

Robotic soccer is a challenging research problem because problems in robotics, artificial intelligence, multi-agent systems and real-time reasoning have to be solved in order to create a successful team of robotic soccer players. In the talk, I will concentrate on cooperative sensing methods and on team play approaches that have been developed for the CS Freiburg team, which has become RoboCup world champion the third time recently.

Hans Kamp
Institut für Maschinelle Sprachverarbeitung
University of Stuttgart, Germany

Representations of temporal Information in natural Language and their Logic

Most of our more mundane uses of natural language involve temporal information: utterances locate the events of which they speak somewhere in time and/or they present those events as standing in certain temporal relations to each other. Semantic representations of natural language discourse must make these temporal properties and relations explicit and they must do this in a "logically" transparent form, so that inference engines can operate on the resulting representations and draw from those the conclusions that human interpreters draw from them (unconsciously as well as consciously).

The history of the representation of temporal information expressible in natural language discourse is one which by now spans a period of more than 40 years. In the early days there was a strong emphasis on systems of "tense logic", a variety of modal logic, developed by Arthur Prior and others. Results from the sixties established the considerable expressive power of syntactically simple systems of this sort, as well as their (moderate) "tractability". (In spite of these tense logics being fragments of second order logic, the systems in question are decidable.)

It became gradually clear, however, that to the representation of temporal information in natural language these systems are not particularly well suited, as they do too much violence to the form in which such information is represented in ordinary language. So in their stead formalisms were designed in which temporal information can be represented by explicit reference to (and quantification over) times. Unfortunately, representations of this latter kind come with complexity properties familiar from general first order and second order predicate logic: Logical consequence is recursively enumerable at best and at worst it is beyond hyperarithmetic. Moreover, it is not obvious how the purely temporal information contained in such representations can be separated from the remainder to yield proof-theoretically tractable representations of just this temporal information. Yet, for the automated processing of natural language extracting such representations of the temporal information is of the utmost importance.

In the paper I will rehearse in some detail the history of the subject hinted at above; I will then present, by way of examples, more recent approaches towards the representation of natural language sentences and discourses in which temporal information is relevant and I will address the problems of designing special purpose mechanisms that detect and evaluate the temporal information stored in those representations.

Michael Kearns
Syntek Capital

Computational Game Theory and AI

There has been growing interest in AI and related disciplines in the emerging field of computational game theory. This area revisits the problems and solutions of classical game theory with an explicit emphasis on computational efficiency and scalability. The interest from the AI community arises from several sources, including models and algorithms for multi-agent systems, design of electronic commerce agents, and the study of compact representations for complex environments that permit efficient learning and planning algorithms.

In this talk, I will survey some recent results in computational game theory by myself and others, and highlight similarities with algorithms, representations and motivation in the AI and machine learning literature. The topics examined will include a simple study of gradient algorithms in general games, the application of reinforcement learning algorithms and their generalizations to stochastic games, and the introduction of compact graphical models for multi-player games. Interesting directions for further work will be discussed.


Raymond Reiter
Department of Computer Science
University of Toronto, Canada

Cognitive Robotics

In the past decade, the intensive study of deliberative autonomous robots has emerged from within the knowledge representation community. The single most important reason for this has been the development of principled solutions to the frame problem by researchers in theories of actions, and by now, there are a number of well developed action theories, including the situation calculus, features and fluents, the event calculus, the fluent calculus, and the family of A-languages. Many of these languages provide representations for time, concurrency, processes, probability and decision theory, and perhaps most importantly, logic-based programming languages for writing robot control programs at a very high level of abstraction.

This talk will describe how these ideas have been realized for the situation calculus in the Cognitive Robotics Project at the University of Toronto. I shall also describe our implementation experiences with an RWI B21 autonomous robot.

V.S. Subrahmanian
Department of Computer Science
University of Maryland, USA

IMPACT: Interactive Maryland Platform for Agents Collaborating Together

IMPACT (Interactive Maryland Platform for Agents Collaborating Together) provides a platform for the creation and deployment of distributed, multiagent applications by building on top of legacy and/or specialized codebases. In this talk, I will describe the overall architecture of the IMPACT system, and outline how this architecture (i) allows agents to be developed either from scratch, or by extending legacy code-bases, (ii) allows agents to interact with one another, (iii) allows agents to have a variety of capabilities (reactive, autonomous, intelligent, mobile, replicating) and behaviors, and (iv) how IMPACT provides a variety of infrastructural services that may be used by agents to interact with one another. I will describe specific methods to scale IMPACT agents to handle large numbers of concurrent service requests.