Date of Award

5-2010

Degree Name

Doctor of Philosophy

Department

Electrical and Computer Engineering

First Advisor

Dr. Frank Severance

Second Advisor

Dr. Gerald L. Sievers

Third Advisor

Dr. Liang Dong

Abstract

Inspired by the collective behavior observed in biological colonies, swarm robotics is a new approach to design a distributed control algorithm in order to coordinate a group of simple robots performing a complex group task. With only limited computation and communication ability of individual robots, one of the challenges in designing such multi-robotic systems is to understand the effect of individual robots behavior on the group performance. This thesis dedicates the research to designing a set of local interaction and adaptation rule for individual robots so that optimized collective foraging performance can be achieved at group level.

The research starts with designing a computer simulation program to simulate collective foraging of multiple robots. A behavior based controller was used to design a robot that performs basic foraging behavior.

Inspired by the widely observed division of labor phenomenon in biological systems, the desired group behavior in collective foraging is to have an optimal division of labor between active foraging and resting among robots. Two variables, foraging threshold and foraging stimulus, were used to calculate the foraging probability for each robot. A set of local adaptation rules are designed to adapt these two variables in a self-organized manner through the multiple interactions between robot and food source in the environment, as well as interactions between foraging robots. The results of the experiments show that the robot group with adaptation mechanisms not only achieve optimal group foraging performance, but also show robustness and flexibility of robot group to environment changes.

The study then extended to that a group of heterogeneous robots foraging in a more complex environment with two types of food. Desired robot group behavior requires an optimal task allocation of foraging robots on two types of food, and division of labor between foraging and resting. The interaction rules were upgraded for individual robots. The results of the experiments show that both division of labor and task allocation emerged at a group level. The robot group also demonstrates the ability to collectively perceive the changes in the environment and to guide the group toward foraging performance optimization.

Access Setting

Dissertation-Open Access

Share

COinS