Self-organized systems

Description

Self-organizing arises from the systems ability to efficiently respond to external stimulations, without being managed by an outside source. Self-organizing systems often benefit from internal emergent structures resulting from complex interactions between their components. These structures adapt in order to fit the environment needs.

Research topics

The current existing approaches to problem solving, using emergence and self-organization, are generally based on inspirations and metaphors of natural systems like ant foraging (ACO) or flocking birds (PSO), etc. The mechanisms used in these approaches are a simple transposing of natural models to computational ones, without a deep understanding of the mechanisms that makes these metaphors work. The main concern of our research is to consider the complex characteristics of systems, in order to manage complex emergent structures and get appropriated responses to resolve problems.

The self-organizing study involves other topics :

• Situated Multi-Agents
• Swarm Intelligence
• Emergence of language, Emergence of meaning and representations
• Positive/negative feedback
• Local/global relation in complex environment
  
• Balance of exploration and exploitation

Self-organized systems can address the following industrial applications :

• Critical resource sharing problems (CESNA project and Governance project)
• Dynamic staff  management
• Supply chain management
• Cities' building for content creators (City building project)
• etc.

Long-range contributions

To increase semantic expressiveness is still an open problem. The more advanced ontologies doesn't lead programs to appropriate the data they manage. Process are executed blindly by the way of static structures. In general, we can't expect more from programs that what we explicitly define for them. On the contrary, if we refer to natural and self-organized systems, the properties we are looking for emerge from local interactions. That is the case for the human brain : the meaning emerges from the complex nerve cells interactions. This is the case for artificial neuronal networks that can learn and recognize patterns (applied for instance to digits). In these cases, the appropriations performed by systems, based on self-organized data, allow to make emerge results beyond the initial system knowledge.

In our point of view, the understanding of the rooting mechanisms of these complex behaviours will lead to artificial intelligence improvements in the following domains :

• Semantic expressiveness
• Collaboration between humans and programs
  
• Learning process
• etc.