zur Startseite

ALLOW Ensembles EU Specific Targeted Research Project

Adaptable Pervasive Flow Ensembles
ProjekttypEU Specific Targeted Research Project
Gefördert durch EU
Beginn 2013/02/01
Leiter Prof. Dr. rer. nat. Dr. h. c. Kurt Rothermel
Mitarbeiter Tariq, Muhammad Adnan
Koldehofe, Boris
Schäfer, David Richard
Bach, Thomas
Ansprechpartner Rothermel, Kurt
Tariq, Muhammad Adnan
Kooperationspartner University of Stuttgart, Germany (Coordinator)
German Research Center for Artificial Intelligence, Germany
Bruno Kessler Foundation, Italy
Imperial College London, UK
University of Crete, Greece

The recent advances in pervasive technologies enable construction of large-scale socio-technical systems which tightly interweave humans and their social structures with the technology. These systems are realized as a collective of diverse heterogeneous actors situated both in the physical world such as people, objects etc., and in the backend computer systems such as control processes. The objective of ALLOW Ensembles is to develop a new design principle for large-scale collective systems (CAS) based on the concepts of cells and ensembles. Cells are basic building blocks representing the different components of the system and ensembles are collections of cells collaborating together to accomplish certain goal in a given context. We use Adaptive Pervasive Flows – a programming paradigm based on workflow technology for pervasive systems – to model the behavior of cells as a set of interrelated tasks. This enables the salient principle of cell specialization. It allows for changing the behavior of the individual cell (tasks and order of execution) to fit into an ensemble and to achieve a given goal with high utility in collaboration with other cells of the ensemble.

Following the principle of cell specialization, we develop methodologies for the evolution of cells and ensembles to meet arbitrary system goals, autonomously improving the utility of the system under changing contexts. Another major goal of the project is to develop models, theories and algorithm to ensure robustness and security so that ensembles can survive wide range of hardware/software failures and can protect sensitive data. Furthermore, we do novel research on the controllability of emergent properties of complex ensemble systems.

The ensemble concept challenges current thinking as it represents a new type of systems that evolve over multiple generations to adapt to contextual changes and constantly improve utility. Evolutionary data is collected and analyzed to learn from the characteristics of past ensembles executions. This knowledge forms the foundation of evolution, leading to robust and high utility systems.

The resulting fundamental concepts will be tested based on visionary application scenarios such as integrated urban transport and smart production chains, to evaluate their applicability.

externer Link  Project Website