Evolutionary mobile robot using breeder genetic algorithms

Dissertant

al-Khafaji, Tefool Husayn

University

University of Technology

Faculty

-

Department

Computer Sciences Department

University Country

Iraq

Degree

Ph.D.

Degree Date

2006

English Abstract

The field of evolutionary robotics is an interesting research area that deals with building Neuro-controller systems by evolving individuals depending on the Darwinian principles of natural selection and survival of the fittest.

This area of research aimed to build general purpose robot control systems that can be used with real robots rather than simulated agents.

It also proposes that the robot is strongly coupled with its environment and affected by which.

Usually, genetic algorithms (GAs) are used as the evolutionary component that evolves a population of controllers (usually neural networks) with the ultimate goal finding a controller to tackle the given task.

The field of evolutionary robotics stems its importance from the need to build intelligent robots that can learn to behave autonomously in unexpected situations in unknown and unpredictable environments.

This feature is important to be gained by any reliable autonomous robot to can be depended on in performing tasks like discovering dangerous places (mines fields, whorls surfaces, volcanoes, etc.), helping disabled people, cleaning, and many others tasks.

In this dissertation Breeder Genetic Algorithm (BGA) is experimented to be the evolutionary tool in an evolutionary mobile robot system.

BGAs share aspects with traditional GAs and evolution strategies.

They share evolution strategies the selection strategy and relying on real parameters rather than their coding.

On the other hand, they share GAs the consideration that the recombination operator is the main explorative component and mutation as a secondary operation.

A robot simulator is specially designed to test and implement the controller system.

This simulator has two main benefits : 1.

to specify sensors' activations at specific situations ; 2.

to implement the action that the controller decides to take at that situation.

The controller (referred to as EMROV as abbreviation to Evolutionary Mobile Robot for Obstacle Avoidance I is tested in performing the task of obstacle avoidance.

Also the system is tested when two tasks are wanted to be modulated.

Homing is chosen to be modulated with obstacle avoidance task.

The system is implemented using C++ programming language.

The system exhibits good performance levels at the experiments made.

The system is tested at several environments with and without learning in addition to evolution.

The results are documented throughout the dissertation.

Main Subjects

Information Technology and Computer Science

Topics

• Computing
• Algorithms

• APA
• MLA
• AMA

Language

English

Data Type

Arab Theses

Record ID

BIM-305967