Organic Computing


The Vision of Organic Computing Systems

Organic Computing (short: OC) has emerged recently as a challenging vision for future information processing systems. It is based on the insight that we will soon be surrounded by systems with massive numbers of processing elements, sensors, and actuators. Many of those systems will be autonomous and due to the complexity it will be infeasible to monitor and control them entirely from external observations. Instead they must monitor, control, and adapt themselves. To do so, these systems must be aware of themselves and their environment, communicate, and organize themselves in order to perform the actions and services required.

The presence of networks of intelligent systems in our environment opens up fascinating application areas but, at the same time, bears the problem of their controllability. Hence, we have to construct these systems – which we increasingly depend on – as robust, safe, flexible, and trustworthy as possible. In order to achieve these goals, our intelligent technical systems must act more autonomously and they must exhibit life-like (organic) properties. Hence, an OC system is a technical system that adapts dynamically to the current conditions of its environment. It is self-organizing, self-configuring, self-healing, self-protecting, self-explaining, and situation-aware.

Central aspects of OC systems are inspired by an analysis of information processing in biological systems. Within short time, OC became a major research activity in Germany and worldwide.

Knowledge Exchange Within Intelligent Distributed Systems

In our work in the field of OC we focus on intelligent distributed systems such as teams of robots, smart sensor networks, or multi-agent systems. Often, the nodes of such a system have to perform the same or similar tasks, or they even have to cooperate to solve a given problem. Typically, these nodes know how to observe their local environment and how to react on certain observations, for instance, and this knowledge is represented by (symbolic) rules. However, many environments are dynamic. That is, new rules become necessary, old rules become obsolete, or rules change slightly over time (concept drift). That implies that really intelligent nodes (robots, smart sensors, agents, etc.) should adapt on-line to their environment by means of certain machine learning techniques.

Typically, nodes exchange information about what they observe in their environment in order to collaborate. We refer to this kind of knowledge as descriptive knowledge. We claim that the rules that are adapted or learned on-line (we call this functional knowledge) are more abstract and often more valuable than descriptive knowledge. Furthermore, in the case of a dynamic environment descriptive knowledge may be inadequate or even wrong to describe novel or changing phenomena. That is, an observation in the input space of one node might be misinterpreted by another node when the models that represent functional knowledge are different, for instance. Therefore, organic nodes should exchange functional knowledge instead of or in addition to descriptive knowledge. The advantages are obvious:

  • Techniques and ontologies needed for functional knowledge exchange are independent from a particular application domain,
  • The communication effort needed for functional knowledge exchange may be significantly lower than the effort needed for descriptive knowledge exchange, and
  • Organic nodes may behave proactively: Before certain situations come up in their local environment, they will already be enabled to handle them.



Robots exchange observations (descriptive knowledge that describes what is seen in their local environment).



Robots exchange learned rules (functional knowledge that describes how to interpret these observations).

To realize rule exchange in intelligent distributed systems, we address, amongst others, the following research issues:

  • Representation of uncertain (semi-symbolic) rules in a form which enables rule exchange,
  • Partially unsupervised training of classifiers (rule systems) from sample data,
  • Detection of the need to adapt a classifier (situation-awareness), e.g., novelty and obsoleteness detection mechanisms,
  • Numerical assessment of the interestingness of learned knowledge (e.g., uniqueness, informativeness, usefulness, importance),
  • Numerical assessment of the uncertainty concerning the parameters of learned rules,
  • Mechanisms for interestingness-based rule exchange and rule integration into a classifier,
  • Sample applications in the fields of robotics, driver assistance systems, and traffic control.

Together with partners from other universities we also address the following issues in the field of OC:

  • Evolvable hardware for electromyographic prosthetic hand control (with M. Platzner, Paderborn, Germany),
  • Emergence and self-organization in OC systems (with C. Müller-Schloer, Hanover, Germany),
  • Relations between various functionally equivalent but semantically different classifier paradigms (with S. J. Ovaska, Espoo/Helsinki, Finland).



Further Information

Staff:

Funding:

Publications:

D. Fisch, M. Jänicke, B. Sick, C. Müller-Schloer; Quantitative Emergence – A Refined Approach Based on Divergence Measures; Fourth IEEE International Conference on Self-Adaptive and Self-Organizing Systems; pp. 94-103; Budapest, 2010


D. Fisch, F. Kastl, B. Sick; Novelty-Aware Attack recognition - Intrusion Detection With Organic Computing Techniques; 3rd IFIP Conference on Biologically-Inspired Collaborative Computing (BICC 2010) at the World Computer Congress (WCC 2010); pp. 242-253; Brisbane, 2010


D. Fisch, T. Gruber, B. Sick; SwiftRule: Mining Comprehensible Classification Rules for Time Series Analysis; in: IEEE Transactions on Knowledge and Data Engineering; (accepted)


D. Fisch, B. Kühbeck, B. Sick, S. J. Ovaska; So Near And Yet So Far: New Insight Into Properties Of Some Well-Known Classifier Paradigms; in: Information Sciences; vol. 180, no. 18, pp. 3381-3401; 2010


D. Fisch, B. Sick; Training of Radial Basis Function Classifiers With Resilient Propagation and Variational Bayesian Inference; Proceedings of the ”International Joint Conference on Neural Networks (IJCNN 2009)”; pp. 838-847; Atlanta, 2009


D. Fisch, M. Jänicke, E. Kalkowski, B. Sick; Learning by Teaching Versus Learning by Doing: Knowledge Exchange in Organic Agent Systems; in Proceedings of the “IEEE Symposium on Intelligent Agents (IA 2009)”; pp. 31-38; Nashville, 2009


C. Müller-Schloer, B. Sick; Controlled Emergence and Self-Organization; in: R. P. Würtz (Ed.): Organic Computing; ch. 4, pp. 81-104; Series on Understanding Complex Systems; Springer Verlag, Berlin, Heidelberg, New York; 2008


D. Andrade, T. Horeis, B. Sick; Knowledge Fusion Using Dempster-Shafer Theory and the Imprecise Dirichlet Model; in: Proceedings of the ”2008 IEEE Conference on Soft Computing in Industrial Applications (SMCia/08)”; pp. 142-148; Muroran, 2008


K. Glette, J. Torresen, T. Gruber, B. Sick, P. Kaufmann, M. Platzner; Comparing Evolvable Hardware to Conventional Classifiers for Electromyographic Prosthetic Hand Control; in: Proceedings of the ”3 rd NASA/ESA Conference on Adaptive Hardware and Systems (AHS-2008) ”; pp. 32-39; Noordwijk, 2008


and others...