February 16 – 21 , 2014, Dagstuhl Seminar 14081

Robots Learning from Experiences


Anthony Cohn (University of Leeds, GB)
Bernd Neumann (Universität Hamburg, DE)
Alessandro Saffiotti (University of Örebro, SE)
Markus Vincze (TU Wien, AT)

For support, please contact

Dagstuhl Service Team


Dagstuhl Report, Volume 4, Issue 2 Dagstuhl Report
Aims & Scope
List of Participants
Dagstuhl Seminar Wiki
Dagstuhl Seminar Schedule [pdf]

(Use seminar number and access code to log in)


The ability to exploit experiences is an important asset of intelligent beings. Experiences provide a rich resource for learning, solving problems, avoiding difficulties, predicting the effects of activities, and obtaining commonsense insights. Current robots do not in general possess this ability, and this is a decisive reason for the often perceived "lack of intelligence" of current robotic systems: they repeat mistakes, do not learn to anticipate happenings in their environment, and need detailed instructions for each specific task.

Consider an everyday task of a service robot, such as grasping a cup from a cupboard and bringing it to a person sitting at a table. This task may occur in many variations and under unpredictable circumstances. For example, persons may sit at different sides of a table, a direct path to the table may be blocked, the table may be cluttered with various objects, hot water may be ready or not, the cup on the shelf may be upside-down, etc. It is clearly infeasible to provide the robot with precise instructions for all contingencies at design time or to specify tasks with highly detailed instructions for each particular concrete situation which may arise. Hence without such knowledge, robot behaviour is bound to lack robustness if the robot cannot autonomously adapt to new situations.

How would the robot, for example, avoid pouring coffee into an upside-down cup? Based on experiences with multiple pouring actions, the robot will have formed a conceptualisation of all concomitant circumstances of successful pouring, for example to pour into a "container". The robot may not know the name of this conceptualisation but will know that it must be open on top, hollow, empty, etc. Similarly, the robot may have encountered upside-down objects before and hence be able to conceptualise the corrective action of turning an object to make it a usable container.

This seminar has brought together experts and scholars from the robotics, learning, and knowledge representation communities to discuss current approaches to make robots learn from experiences. Emphasis was on the representation of real-world experiences and on exploiting experiences for autonomous acting in a changing or partially unknown environment.

Summary text license
  Creative Commons BY 3.0 Unported license
  Anthony Cohn, Bernd Neumann, Alessandro Saffiotti, and Markus Vincze


  • Artificial Intelligence / Robotics


  • Learning
  • Experiences
  • Cognitive systems


In the series Dagstuhl Reports each Dagstuhl Seminar and Dagstuhl Perspectives Workshop is documented. The seminar organizers, in cooperation with the collector, prepare a report that includes contributions from the participants' talks together with a summary of the seminar.


Download overview leaflet (PDF).

Dagstuhl's Impact

Please inform us when a publication was published as a result from your seminar. These publications are listed in the category Dagstuhl's Impact and are presented on a special shelf on the ground floor of the library.


Furthermore, a comprehensive peer-reviewed collection of research papers can be published in the series Dagstuhl Follow-Ups.