Biological systems appear to emerge from the simultaneous interaction of three fundamental processes: self-organization, highly constrained developmental programs, and the continuous activity of the organism within its environment.
Conventional connectionist approaches generally fail to capture the full range of mechanisms involved in the formation of such networks. In contrast, adaptive multi-agent systems provide a technological framework capable of generating open, autonomous, and complex systems that can adapt to dynamic environments.
Our work combines the functional granularity of artificial neurons with the ontogenetic potential of self-organizing multi-agent systems. This hybridization relies on cooperation as a local adjustment criterion governing interactions between agents.
Each cooperative neuro-agent performs not only information processing and transformation but also exhibits behaviors related to proliferation, connectivity regulation, and activity control. Two complementary dynamics can therefore be distinguished:
• Connectionist learning, corresponding to synaptic regulation
• System adaptation, corresponding to ontogenetic modifications of the network structure through self-organization
The neuro-agent network has been tested in four main applications:
• learning logical functions from randomly activated inputs
• simulation of the migratory behavior of the leatherback sea turtle
• flood prediction using radar and rainfall data
• simulation of simple trophic learning processes using virtual robots
This approach is distinctive because it reproduces cognitive functions within an artificial system without relying on semantic or symbolic models to interpret perceived information.
Our objective is to study the emergence of properties in a self-organized system and to demonstrate that cooperative neuro-agents can generate functional systems whose complexity approaches that of biological systems.
To support this objective, we have developed analytical tools capable of identifying, from a functional perspective, the emergent structures and the dynamic processes that enable the transition toward higher levels of organization.