Nicola Milano

Abstract

This thesis addresses the study of evolutionary methods to achieve robustness and evolvability in artificial systems. Chapter 1 introduces the research area, reviews the state of the art, discusses promising research directions, and presents the two major scientific objectives of the thesis. The first objective, which is covered in Chapter 2, is to verify and exploit the impact of the environmental variations on the evolvability of an artificial system. This is accomplished through the use of two experimental setup: the simulation of digital circuits and the simulation of a robotic agent situated in an external environment. Digital circuits are used to considers the variation as internal to the system, modelled as fault in circuit gates; agent-based simulation instead consider the variation in the external environment where the robot performs. The second objective, which is targeted in Chapter 3, presents the design of a new algorithm and a more efficient selection mechanism that exploits the characteristics of robustness and neutrality of the digital circuit domain. Due to its relative simplicity quantitative measures of phenotypic complexity, robustness and evolvability are obtained. Such information on the search space composition is then used to design a novel evolutionary algorithm that outperforms previously methods and to propose a selection mechanism that takes into account the phenotypic complexity of the genotypes. Chapter 4 summarizes the results obtained and describes the major contributions of the thesis.

Keywords

Artificial evolution, Neural networks, Cartesian genetic programming, Embodied agents

Document Type

Thesis

Publication Date

2021

Download

Share

COinS