From Multicellularity to Cell Based Optimization:
Studying the Cooperative Capabilities of Evolvable Cells

As computer systems become increasingly complex and distributed, there is a growing need to build decentralized systems using a bottom-up design strategy. However, humans often have a centralized, hierarchical mindset. This makes it difficult for us to design decentralized systems. Slime Molds (Dictyostelium discoideum) provide an example of emergent behavior from nature where relatively simple cells occasionally cooperate to form multi-celled organisms. This occurs without a leader or central organization. In the same way that groups of cooperating cells form a complex organism in nature, computer programs or agents are parts of information systems. Inspiration from these natural systems may help us to better understand and build decentralized information systems.

The Cell Based Optimization (CBO) algorithm is proposed consisting of an artificial ecosystem and evolvable Slime Mold-like cells. It is used to study the ability of simple cells to evolve collective behavior. CBO applied to a resource allocation problem in an information ecosystem domain. In this problem, simulated users make requests to databases that are distributed across a network. The cells are used as a metaphor to describe agents that handle these requests. A multiagent platform, DIET (Decentralized Information Ecosystem Technologies), was used to build this model and carry out simulations. The objective was that populations of cells would migrate to collectively minimize user wait time. The CBO algorithm proved to be able to evolve stable, optimal solutions.

Download

Thesis: Download in pdf format (850 KB)

Paper: Presented as a poster at the UK Workshop on Computational Intelligence (UKCI 02). UKCI'02 Poster: ppt (550 KB).