Member Login

JCal Pro Mini-calendar

January 2020 February 2020 March 2020
Mo Tu We Th Fr Sa Su
Week 5 1 2
Week 6 3 4 5 6 7 8 9
Week 7 10 11 12 13 14 15 16
Week 8 17 18 19 20 21 22 23
Week 9 24 25 26 27 28 29

Current Time

The time is 01:12:10.
Workshop: New Perspectives for Intelligent Goal-Directed Behaviour in the Real-World Robotics Domain Print E-mail
Date and Time   Tuesday, 09 September 2008
09.00 - 13.00h (to be confirmed)

Location  
University of Hamburg,
Edmund-Siemers-Allee 1,
ESA 1W, Room 121
(On the first floor of the west wing)

Keynote   Prof. PhD. Alessandro Saffiotti , AASS Mobile Robotics Lab, Örebro University, Sweden
Big Challenge of Physical Embedded Intelligent Systems
Or: How to Integrate Cognitive Abilities And Sensori-motor Abilities

  Dr. Eiji Uchibe , Neural Computation Unit, Okinawa Institute of Science and Technology, Japan
icon Learning and Evolution for Real Robots

Topics   Martin Weser , Hamburg University
icon Multisensory Memory Representations of Robot Actions

  Sascha Jockel , Hamburg University
iconSparse Distributed Memory for Robot Action Manipulation and Prediction

Contacts
  Martin Weser
E-mail : weser(at) informatik.uni-hamburg.de
Phone : +49 . 40 . 428 83 - 2362

Sascha Jockel
E-mail : jockel(at) informatik.uni-hamburg.de
Phone : +49 . 40 . 428 83 - 2361

Workshop Abstract


One of the main goals of robotic research is the autonomous production of goal-directed behaviour in real world environments. The robotic community almost agrees on the fact that both, symbolic deliberative planning as well as analogue reactive execution of actions is necessary to achieve intelligent behaviour. Without symbolic planning an agent may never be able to reason about consequences of future actions and to determine consecutive sequences of actions that lead to a desired goal. On the other hand the executions of actions have to be reactive due to dynamic changes in the world and uncertainty about the current state of the world during the planning phase. Representative for the analogue layer are reactive systems [Braitenberg, 1984; Brooks, 1986] while symbolic planning is the major subject of the traditional AI community. The workshop will present current models of reactive and symbolic systems and make an endeavor to develop a concept for combining the two approaches.

We will discuss possibilities for an integration of reactive behavior execution with symbolic planning by first identifying existing overlaps and potential interfaces between the two approaches. We will then try to derive general principles for interactions between them, and assess potential benefits for robot behaviour. Specific questions that will guide the discussion include:

  • How can symbolic representations be generated if robot control structures for robot behaviour are learned autonomously?

  • Can the results of hierarchical learning algorithms serve as symbolic descriptions of the properties of the environment?

  • How to make effective use of communication in a population of robots?

  • How can robot actions and action sequences be represented with respect to temporal aspects of multiple sensor and actuator trajectories? Is a mathematical formulation, like sparse distributed memory, suitable for representing, learning and regenerating reactive robot behaviour?

  • Which atomic skills are necessary for future service robot applications?

  • How to design an interface of atomic robot skills for deliberative planning components? What are important parameters to allow symbolic planning while encapsulating the details of a reactive execution of robot action?

For further information on the speakers' research background, the workshop topic, and the workshop format, please refer to the Detailed Workshop Description.

Last Updated ( Monday, 15 September 2008 )