Self-organization in large populations of mobile robots

A homogeneous population of robots described as an Army-ant swarm is to be realized for material transportation. Robots envisioned in the Army-ant scenario are relatively small, independent autonomous mobile robots that can cooperatively carry palletized loads. In this thesis, the agents are treated as a self-organizing system of moving points. This characteristic makes the Army-ant swarm a modular, adaptive, and dynamic system.

Several algorithms for the spatial self-organization of the robots are given. Self-organizing agents can arrange themselves geometrically in two- and three-dimensional space using only local information about teammates. The method is a distributed one: each agent uses only the information obtained by its own sensors. Algorithms are based on feasible assumptions. It is also shown possible to divide such a population into different groups around goals by communicating minimal data. Data transfer has a broadcast characteristic.

Behavioral self-organization in the Army-ant scenario is also investigated. Activation and inhibition relations between robots determine the behavior (position in a behavioral space) of the agents, while in spatial self-organization force fields are in effect. Several problems which may be encountered and the solution to some of these problems are outlined. Methods for communication and cooperative decision systems — such as coupled van der Pol oscillators — in finding and carrying the pallets are proposed. Sensors and communication systems that may be used in the Army-ant scenario are also briefly discussed.