In social insect swarms, such as ants and bees, behaviors of the swarm as a whole are called swarm intelligence. Without any centralized control, the individual behaviors result in the necessary functions, structures, and behaviors of the swarm, such as foraging, nest building, and role assignment. The individuals have only limited abilities, and they completely autonomously behave.
Similar phenomena are observed in various fields, such as chemical reactions and human society, and this is called self-organization. Self-organization is defined as "In a system consisting of multiple components, each component operates and interacts based on simple rules derived from local information, resulting in the emergence of more sophisticated patterns in the system as a whole.".
Self-organization is a fully autonomous and decentralized mechanism that has been used in various applications of information networks. This is because the information networks are collections of nodes, and self-organization avoids the management overhead associated with centralized control. For example, foraging ants take the shortest path between the nest and the food sources. This mechanism of foraging is modeled as a heuristic approach for the traveling salesman problem, called ant colony optimization (ACO). ACO has many applications for path control in information networks.
In the Wakamiya Lab, we have also developed information and communication technologies based on various self-organizing mechanisms of life:
Not only swarm intelligence, we also study adaptive control based on metabolic response mechanisms and so on.
These studies are based on mathematical models (e.g., pulse-coupled oscillator model, reaction-diffusion model, and reaction threshold model) rather than mere mimicry of apparent behaviors. The mathematical models allow theoretical verification of the scalability, adaptability, and robustness of the applied technology. We evaluate the effectiveness of the proposed method not only by numerical analysis and simulation, but also by demonstration experiments using prototypes.
In parallel, we also promote theoretical research to better understand self-organization itself. The patterns in self-organization are merely a consequence of the interaction between the components, and do not always emerge in a desirable way. In some cases, convergence or its speed is inadequate, and pattern formation fails or takes time. To maintain the merit of self-organization, we have developed the Guided/Controlled/Managed Self-Organization (GS). GS achieves desirable self-organization by adding gradual control or interference from the outside while maintaining the autonomy of the components. For example, we found that we can change the period of synchronized firefly emission, by controlling fireflies that have less interaction with other fireflies. We are also working on establishing a theoretical basis for hierarchical self-organization for large-scale information and communication systems.